{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.989247311827957,
  "eval_steps": 500,
  "global_step": 1300,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 21.59375,
      "epoch": 0.0030721966205837174,
      "grad_norm": 18.146560668945312,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": 0.1353,
      "reward": 0.5462811142206192,
      "reward_std": 0.20919979363679886,
      "reward_std_all": 0.2262057512998581,
      "rewards-max/forensics_sinkhorn_reward": 0.9101973474025726,
      "rewards-max/forensics_strict_edge_reward": 0.06767599159593374,
      "rewards-max/forensics_tiou_reward": 0.7984493374824524,
      "rewards/forensics_sinkhorn_reward": 0.4543914496898651,
      "rewards/forensics_strict_edge_reward": 0.0037262411422034347,
      "rewards/forensics_tiou_reward": 0.20149360224604607,
      "step": 1
    },
    {
      "completion_length": 22.3125,
      "epoch": 0.006144393241167435,
      "grad_norm": 15.60630989074707,
      "kl": 0.0023193359375,
      "learning_rate": 9.992307692307693e-07,
      "loss": -0.0445,
      "reward": 0.603609710931778,
      "reward_std": 0.2174774333834648,
      "reward_std_all": 0.20837509632110596,
      "rewards-max/forensics_sinkhorn_reward": 0.8703066110610962,
      "rewards-max/forensics_strict_edge_reward": 0.026384840719401836,
      "rewards-max/forensics_tiou_reward": 0.7771201133728027,
      "rewards/forensics_sinkhorn_reward": 0.5115953385829926,
      "rewards/forensics_strict_edge_reward": 0.0011129422928206623,
      "rewards/forensics_tiou_reward": 0.22939951717853546,
      "step": 2
    },
    {
      "completion_length": 22.3125,
      "epoch": 0.009216589861751152,
      "grad_norm": 6.263747215270996,
      "kl": 0.001399993896484375,
      "learning_rate": 9.984615384615384e-07,
      "loss": 0.1789,
      "reward": 0.5772744417190552,
      "reward_std": 0.22607571631669998,
      "reward_std_all": 0.22278043627738953,
      "rewards-max/forensics_sinkhorn_reward": 0.9006702303886414,
      "rewards-max/forensics_strict_edge_reward": 0.27267758920788765,
      "rewards-max/forensics_tiou_reward": 0.7823465168476105,
      "rewards/forensics_sinkhorn_reward": 0.46084265410900116,
      "rewards/forensics_strict_edge_reward": 0.012500784825533628,
      "rewards/forensics_tiou_reward": 0.20496424287557602,
      "step": 3
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.01228878648233487,
      "grad_norm": 5.492680072784424,
      "kl": 0.00222015380859375,
      "learning_rate": 9.976923076923077e-07,
      "loss": 0.0618,
      "reward": 0.5977490544319153,
      "reward_std": 0.21978794783353806,
      "reward_std_all": 0.211220882833004,
      "rewards-max/forensics_sinkhorn_reward": 0.8207923471927643,
      "rewards-max/forensics_strict_edge_reward": 0.04574807739118114,
      "rewards-max/forensics_tiou_reward": 0.6275305449962616,
      "rewards/forensics_sinkhorn_reward": 0.4763678014278412,
      "rewards/forensics_strict_edge_reward": 0.0019283224464743398,
      "rewards/forensics_tiou_reward": 0.1937377005815506,
      "step": 4
    },
    {
      "completion_length": 23.25,
      "epoch": 0.015360983102918587,
      "grad_norm": 5.738350868225098,
      "kl": 0.0084991455078125,
      "learning_rate": 9.969230769230768e-07,
      "loss": 0.0914,
      "reward": 0.5935109257698059,
      "reward_std": 0.17575131356716156,
      "reward_std_all": 0.18328332155942917,
      "rewards-max/forensics_sinkhorn_reward": 0.7815071642398834,
      "rewards-max/forensics_strict_edge_reward": 0.0023898350773379207,
      "rewards-max/forensics_tiou_reward": 0.5722115933895111,
      "rewards/forensics_sinkhorn_reward": 0.4501051902770996,
      "rewards/forensics_strict_edge_reward": 0.00010783540710690431,
      "rewards/forensics_tiou_reward": 0.1652725711464882,
      "step": 5
    },
    {
      "completion_length": 22.46875,
      "epoch": 0.018433179723502304,
      "grad_norm": 4.5898308753967285,
      "kl": 0.01568603515625,
      "learning_rate": 9.961538461538461e-07,
      "loss": 0.1282,
      "reward": 0.5890757441520691,
      "reward_std": 0.19398128986358643,
      "reward_std_all": 0.19860997796058655,
      "rewards-max/forensics_sinkhorn_reward": 0.8495948910713196,
      "rewards-max/forensics_strict_edge_reward": 0.009700306691229343,
      "rewards-max/forensics_tiou_reward": 0.6566298007965088,
      "rewards/forensics_sinkhorn_reward": 0.454777330160141,
      "rewards/forensics_strict_edge_reward": 0.00040925166103988886,
      "rewards/forensics_tiou_reward": 0.1658662110567093,
      "step": 6
    },
    {
      "completion_length": 23.03125,
      "epoch": 0.021505376344086023,
      "grad_norm": 5.314596176147461,
      "kl": 0.03814697265625,
      "learning_rate": 9.953846153846154e-07,
      "loss": 0.2018,
      "reward": 0.5839033126831055,
      "reward_std": 0.17346890270709991,
      "reward_std_all": 0.1915440931916237,
      "rewards-max/forensics_sinkhorn_reward": 0.8803270161151886,
      "rewards-max/forensics_strict_edge_reward": 0.42113904654979706,
      "rewards-max/forensics_tiou_reward": 0.7818833589553833,
      "rewards/forensics_sinkhorn_reward": 0.4648725986480713,
      "rewards/forensics_strict_edge_reward": 0.018044481985270977,
      "rewards/forensics_tiou_reward": 0.19740928709506989,
      "step": 7
    },
    {
      "completion_length": 23.0625,
      "epoch": 0.02457757296466974,
      "grad_norm": 5.760272979736328,
      "kl": 0.02667236328125,
      "learning_rate": 9.946153846153846e-07,
      "loss": 0.2126,
      "reward": 0.5836149156093597,
      "reward_std": 0.12668076157569885,
      "reward_std_all": 0.16896754503250122,
      "rewards-max/forensics_sinkhorn_reward": 0.8795074820518494,
      "rewards-max/forensics_strict_edge_reward": 0.014040044043213129,
      "rewards-max/forensics_tiou_reward": 0.718321293592453,
      "rewards/forensics_sinkhorn_reward": 0.4756494164466858,
      "rewards/forensics_strict_edge_reward": 0.0007101816590875387,
      "rewards/forensics_tiou_reward": 0.1541489064693451,
      "step": 8
    },
    {
      "completion_length": 22.453125,
      "epoch": 0.027649769585253458,
      "grad_norm": 6.163335800170898,
      "kl": 0.04150390625,
      "learning_rate": 9.938461538461539e-07,
      "loss": 0.1918,
      "reward": 0.6585423350334167,
      "reward_std": 0.12904922664165497,
      "reward_std_all": 0.1369374804198742,
      "rewards-max/forensics_sinkhorn_reward": 0.896562933921814,
      "rewards-max/forensics_strict_edge_reward": 0.052019173628650606,
      "rewards-max/forensics_tiou_reward": 0.7453477084636688,
      "rewards/forensics_sinkhorn_reward": 0.5374647676944733,
      "rewards/forensics_strict_edge_reward": 0.002297825901223405,
      "rewards/forensics_tiou_reward": 0.242266945540905,
      "step": 9
    },
    {
      "completion_length": 23.015625,
      "epoch": 0.030721966205837174,
      "grad_norm": 4.075323581695557,
      "kl": 0.1025390625,
      "learning_rate": 9.93076923076923e-07,
      "loss": 0.1742,
      "reward": 0.6144940555095673,
      "reward_std": 0.15181313455104828,
      "reward_std_all": 0.1672765538096428,
      "rewards-max/forensics_sinkhorn_reward": 0.9401176869869232,
      "rewards-max/forensics_strict_edge_reward": 0.07594180421438068,
      "rewards-max/forensics_tiou_reward": 0.8596353530883789,
      "rewards/forensics_sinkhorn_reward": 0.4854152500629425,
      "rewards/forensics_strict_edge_reward": 0.0035782791819656268,
      "rewards/forensics_tiou_reward": 0.20319797843694687,
      "step": 10
    },
    {
      "completion_length": 24.25,
      "epoch": 0.03379416282642089,
      "grad_norm": 5.225701332092285,
      "kl": 0.16845703125,
      "learning_rate": 9.923076923076923e-07,
      "loss": 0.0576,
      "reward": 0.6263064444065094,
      "reward_std": 0.1793917492032051,
      "reward_std_all": 0.17517630755901337,
      "rewards-max/forensics_sinkhorn_reward": 0.9019253253936768,
      "rewards-max/forensics_strict_edge_reward": 0.04419824481010437,
      "rewards-max/forensics_tiou_reward": 0.7638852000236511,
      "rewards/forensics_sinkhorn_reward": 0.49936220049858093,
      "rewards/forensics_strict_edge_reward": 0.0019181165262125432,
      "rewards/forensics_tiou_reward": 0.21799243986606598,
      "step": 11
    },
    {
      "completion_length": 24.359375,
      "epoch": 0.03686635944700461,
      "grad_norm": 5.002464771270752,
      "kl": 0.166015625,
      "learning_rate": 9.915384615384616e-07,
      "loss": 0.0762,
      "reward": 0.6258647739887238,
      "reward_std": 0.1684976890683174,
      "reward_std_all": 0.16965282708406448,
      "rewards-max/forensics_sinkhorn_reward": 0.9325146973133087,
      "rewards-max/forensics_strict_edge_reward": 0.07144753076136112,
      "rewards-max/forensics_tiou_reward": 0.8434546887874603,
      "rewards/forensics_sinkhorn_reward": 0.4884527325630188,
      "rewards/forensics_strict_edge_reward": 0.0030527999624609947,
      "rewards/forensics_tiou_reward": 0.21853097528219223,
      "step": 12
    },
    {
      "completion_length": 26.28125,
      "epoch": 0.039938556067588324,
      "grad_norm": 5.383829593658447,
      "kl": 0.205078125,
      "learning_rate": 9.907692307692307e-07,
      "loss": 0.1568,
      "reward": 0.6022998094558716,
      "reward_std": 0.19551489502191544,
      "reward_std_all": 0.19292201846837997,
      "rewards-max/forensics_sinkhorn_reward": 0.8081215620040894,
      "rewards-max/forensics_strict_edge_reward": 0.035460053593851626,
      "rewards-max/forensics_tiou_reward": 0.5680249333381653,
      "rewards/forensics_sinkhorn_reward": 0.47061264514923096,
      "rewards/forensics_strict_edge_reward": 0.0014844348261249252,
      "rewards/forensics_tiou_reward": 0.18180298060178757,
      "step": 13
    },
    {
      "completion_length": 26.890625,
      "epoch": 0.043010752688172046,
      "grad_norm": 4.849427700042725,
      "kl": 0.20068359375,
      "learning_rate": 9.9e-07,
      "loss": 0.1525,
      "reward": 0.611253410577774,
      "reward_std": 0.1658443547785282,
      "reward_std_all": 0.17137330025434494,
      "rewards-max/forensics_sinkhorn_reward": 0.9193753302097321,
      "rewards-max/forensics_strict_edge_reward": 0.0038512124701810535,
      "rewards-max/forensics_tiou_reward": 0.795384556055069,
      "rewards/forensics_sinkhorn_reward": 0.485385000705719,
      "rewards/forensics_strict_edge_reward": 0.00024995841226882476,
      "rewards/forensics_tiou_reward": 0.18528565764427185,
      "step": 14
    },
    {
      "completion_length": 27.984375,
      "epoch": 0.04608294930875576,
      "grad_norm": 3.9280714988708496,
      "kl": 0.177734375,
      "learning_rate": 9.892307692307692e-07,
      "loss": 0.2526,
      "reward": 0.6149695515632629,
      "reward_std": 0.17771771550178528,
      "reward_std_all": 0.1814841777086258,
      "rewards-max/forensics_sinkhorn_reward": 0.8410426676273346,
      "rewards-max/forensics_strict_edge_reward": 0.0011578568010008894,
      "rewards-max/forensics_tiou_reward": 0.6398727893829346,
      "rewards/forensics_sinkhorn_reward": 0.48249006271362305,
      "rewards/forensics_strict_edge_reward": 4.91175223942264e-05,
      "rewards/forensics_tiou_reward": 0.2011963278055191,
      "step": 15
    },
    {
      "completion_length": 24.59375,
      "epoch": 0.04915514592933948,
      "grad_norm": 3.7814526557922363,
      "kl": 0.20947265625,
      "learning_rate": 9.884615384615385e-07,
      "loss": 0.2744,
      "reward": 0.6532995104789734,
      "reward_std": 0.14766999334096909,
      "reward_std_all": 0.14683809876441956,
      "rewards-max/forensics_sinkhorn_reward": 0.9161403775215149,
      "rewards-max/forensics_strict_edge_reward": 0.006520973518490791,
      "rewards-max/forensics_tiou_reward": 0.789638340473175,
      "rewards/forensics_sinkhorn_reward": 0.555908739566803,
      "rewards/forensics_strict_edge_reward": 0.00042933808254019823,
      "rewards/forensics_tiou_reward": 0.2773989513516426,
      "step": 16
    },
    {
      "completion_length": 23.4375,
      "epoch": 0.05222734254992319,
      "grad_norm": 5.4512529373168945,
      "kl": 0.25341796875,
      "learning_rate": 9.876923076923076e-07,
      "loss": 0.0438,
      "reward": 0.6071439683437347,
      "reward_std": 0.16086484491825104,
      "reward_std_all": 0.16942999511957169,
      "rewards-max/forensics_sinkhorn_reward": 0.8508546650409698,
      "rewards-max/forensics_strict_edge_reward": 0.16069063544273376,
      "rewards-max/forensics_tiou_reward": 0.7031525373458862,
      "rewards/forensics_sinkhorn_reward": 0.4864690601825714,
      "rewards/forensics_strict_edge_reward": 0.006721127778291702,
      "rewards/forensics_tiou_reward": 0.2133760303258896,
      "step": 17
    },
    {
      "completion_length": 26.75,
      "epoch": 0.055299539170506916,
      "grad_norm": 4.823437690734863,
      "kl": 0.255859375,
      "learning_rate": 9.86923076923077e-07,
      "loss": 0.1873,
      "reward": 0.6627356112003326,
      "reward_std": 0.12350479513406754,
      "reward_std_all": 0.12782582640647888,
      "rewards-max/forensics_sinkhorn_reward": 0.8975256681442261,
      "rewards-max/forensics_strict_edge_reward": 0.18648181855678558,
      "rewards-max/forensics_tiou_reward": 0.738258421421051,
      "rewards/forensics_sinkhorn_reward": 0.5270382761955261,
      "rewards/forensics_strict_edge_reward": 0.00841903779655695,
      "rewards/forensics_tiou_reward": 0.2595607563853264,
      "step": 18
    },
    {
      "completion_length": 23.9375,
      "epoch": 0.05837173579109063,
      "grad_norm": 5.119932174682617,
      "kl": 0.2109375,
      "learning_rate": 9.86153846153846e-07,
      "loss": 0.1651,
      "reward": 0.6255657076835632,
      "reward_std": 0.13367101177573204,
      "reward_std_all": 0.1508273333311081,
      "rewards-max/forensics_sinkhorn_reward": 0.9516340494155884,
      "rewards-max/forensics_strict_edge_reward": 0.018350413884036243,
      "rewards-max/forensics_tiou_reward": 0.8948493301868439,
      "rewards/forensics_sinkhorn_reward": 0.49403099715709686,
      "rewards/forensics_strict_edge_reward": 0.0011419749789638445,
      "rewards/forensics_tiou_reward": 0.2332862913608551,
      "step": 19
    },
    {
      "completion_length": 23.9375,
      "epoch": 0.06144393241167435,
      "grad_norm": 4.599627494812012,
      "kl": 0.2275390625,
      "learning_rate": 9.853846153846153e-07,
      "loss": 0.0849,
      "reward": 0.6831024289131165,
      "reward_std": 0.09370995685458183,
      "reward_std_all": 0.10439354553818703,
      "rewards-max/forensics_sinkhorn_reward": 0.9355421364307404,
      "rewards-max/forensics_strict_edge_reward": 0.11253642477095127,
      "rewards-max/forensics_tiou_reward": 0.8501511216163635,
      "rewards/forensics_sinkhorn_reward": 0.5763214230537415,
      "rewards/forensics_strict_edge_reward": 0.0049818872939795256,
      "rewards/forensics_tiou_reward": 0.22680387645959854,
      "step": 20
    },
    {
      "completion_length": 30.046875,
      "epoch": 0.06451612903225806,
      "grad_norm": 3.6531009674072266,
      "kl": 0.203125,
      "learning_rate": 9.846153846153847e-07,
      "loss": 0.2559,
      "reward": 0.6329624652862549,
      "reward_std": 0.13196272030472755,
      "reward_std_all": 0.14518187567591667,
      "rewards-max/forensics_sinkhorn_reward": 0.8638041019439697,
      "rewards-max/forensics_strict_edge_reward": 0.03886437232722528,
      "rewards-max/forensics_tiou_reward": 0.7795758247375488,
      "rewards/forensics_sinkhorn_reward": 0.44499748945236206,
      "rewards/forensics_strict_edge_reward": 0.0016198581288335845,
      "rewards/forensics_tiou_reward": 0.2287597581744194,
      "step": 21
    },
    {
      "completion_length": 25.765625,
      "epoch": 0.06758832565284179,
      "grad_norm": 5.458347320556641,
      "kl": 0.22509765625,
      "learning_rate": 9.838461538461538e-07,
      "loss": 0.1875,
      "reward": 0.6574777662754059,
      "reward_std": 0.13423921540379524,
      "reward_std_all": 0.13443952053785324,
      "rewards-max/forensics_sinkhorn_reward": 0.8348661065101624,
      "rewards-max/forensics_strict_edge_reward": 0.0032595657149840918,
      "rewards-max/forensics_tiou_reward": 0.6150646805763245,
      "rewards/forensics_sinkhorn_reward": 0.5107719749212265,
      "rewards/forensics_strict_edge_reward": 0.00018167558859261135,
      "rewards/forensics_tiou_reward": 0.18373855203390121,
      "step": 22
    },
    {
      "completion_length": 24.890625,
      "epoch": 0.0706605222734255,
      "grad_norm": 4.474020481109619,
      "kl": 0.2607421875,
      "learning_rate": 9.83076923076923e-07,
      "loss": 0.1652,
      "reward": 0.6954609155654907,
      "reward_std": 0.12533185631036758,
      "reward_std_all": 0.12001248821616173,
      "rewards-max/forensics_sinkhorn_reward": 0.9214763641357422,
      "rewards-max/forensics_strict_edge_reward": 0.044346646405756474,
      "rewards-max/forensics_tiou_reward": 0.8129966259002686,
      "rewards/forensics_sinkhorn_reward": 0.5687742233276367,
      "rewards/forensics_strict_edge_reward": 0.0022304722224362195,
      "rewards/forensics_tiou_reward": 0.29352113604545593,
      "step": 23
    },
    {
      "completion_length": 28.109375,
      "epoch": 0.07373271889400922,
      "grad_norm": 3.4764187335968018,
      "kl": 0.234375,
      "learning_rate": 9.823076923076922e-07,
      "loss": 0.1749,
      "reward": 0.6431558728218079,
      "reward_std": 0.1335897594690323,
      "reward_std_all": 0.1422906294465065,
      "rewards-max/forensics_sinkhorn_reward": 0.869425356388092,
      "rewards-max/forensics_strict_edge_reward": 0.03274567611515522,
      "rewards-max/forensics_tiou_reward": 0.7205378711223602,
      "rewards/forensics_sinkhorn_reward": 0.5019095242023468,
      "rewards/forensics_strict_edge_reward": 0.002210689359344542,
      "rewards/forensics_tiou_reward": 0.21781378984451294,
      "step": 24
    },
    {
      "completion_length": 25.25,
      "epoch": 0.07680491551459294,
      "grad_norm": 5.046910762786865,
      "kl": 0.27734375,
      "learning_rate": 9.815384615384615e-07,
      "loss": 0.2691,
      "reward": 0.6191011071205139,
      "reward_std": 0.17522384598851204,
      "reward_std_all": 0.17803925275802612,
      "rewards-max/forensics_sinkhorn_reward": 0.9146412909030914,
      "rewards-max/forensics_strict_edge_reward": 0.04163654148578644,
      "rewards-max/forensics_tiou_reward": 0.8193241059780121,
      "rewards/forensics_sinkhorn_reward": 0.48722898960113525,
      "rewards/forensics_strict_edge_reward": 0.0018004863522946835,
      "rewards/forensics_tiou_reward": 0.2553655430674553,
      "step": 25
    },
    {
      "completion_length": 28.109375,
      "epoch": 0.07987711213517665,
      "grad_norm": 4.988758563995361,
      "kl": 0.248046875,
      "learning_rate": 9.807692307692306e-07,
      "loss": 0.2282,
      "reward": 0.6433919668197632,
      "reward_std": 0.08635412156581879,
      "reward_std_all": 0.11944019049406052,
      "rewards-max/forensics_sinkhorn_reward": 0.8523464798927307,
      "rewards-max/forensics_strict_edge_reward": 0.00019683288701344281,
      "rewards-max/forensics_tiou_reward": 0.6708400845527649,
      "rewards/forensics_sinkhorn_reward": 0.4758220314979553,
      "rewards/forensics_strict_edge_reward": 1.0697721108954283e-05,
      "rewards/forensics_tiou_reward": 0.18483681976795197,
      "step": 26
    },
    {
      "completion_length": 24.8125,
      "epoch": 0.08294930875576037,
      "grad_norm": 4.263630390167236,
      "kl": 0.24755859375,
      "learning_rate": 9.8e-07,
      "loss": 0.0914,
      "reward": 0.6651565432548523,
      "reward_std": 0.1363380029797554,
      "reward_std_all": 0.1361115723848343,
      "rewards-max/forensics_sinkhorn_reward": 0.9283618330955505,
      "rewards-max/forensics_strict_edge_reward": 0.017286939546465874,
      "rewards-max/forensics_tiou_reward": 0.8372900485992432,
      "rewards/forensics_sinkhorn_reward": 0.5459594428539276,
      "rewards/forensics_strict_edge_reward": 0.0007477532781194896,
      "rewards/forensics_tiou_reward": 0.25155576318502426,
      "step": 27
    },
    {
      "completion_length": 26.96875,
      "epoch": 0.08602150537634409,
      "grad_norm": 4.012750625610352,
      "kl": 0.24951171875,
      "learning_rate": 9.792307692307693e-07,
      "loss": 0.1257,
      "reward": 0.603148341178894,
      "reward_std": 0.17276041954755783,
      "reward_std_all": 0.17812344431877136,
      "rewards-max/forensics_sinkhorn_reward": 0.8775750994682312,
      "rewards-max/forensics_strict_edge_reward": 0.20378620049450547,
      "rewards-max/forensics_tiou_reward": 0.7016532123088837,
      "rewards/forensics_sinkhorn_reward": 0.44862326979637146,
      "rewards/forensics_strict_edge_reward": 0.01621444710326614,
      "rewards/forensics_tiou_reward": 0.19206619262695312,
      "step": 28
    },
    {
      "completion_length": 24.125,
      "epoch": 0.0890937019969278,
      "grad_norm": 4.337585926055908,
      "kl": 0.2412109375,
      "learning_rate": 9.784615384615384e-07,
      "loss": 0.1492,
      "reward": 0.6361353397369385,
      "reward_std": 0.12304724752902985,
      "reward_std_all": 0.13791215419769287,
      "rewards-max/forensics_sinkhorn_reward": 0.8654220402240753,
      "rewards-max/forensics_strict_edge_reward": 0.003069945058086887,
      "rewards-max/forensics_tiou_reward": 0.677327424287796,
      "rewards/forensics_sinkhorn_reward": 0.5245951116085052,
      "rewards/forensics_strict_edge_reward": 0.0002064336940748035,
      "rewards/forensics_tiou_reward": 0.215591698884964,
      "step": 29
    },
    {
      "completion_length": 23.6875,
      "epoch": 0.09216589861751152,
      "grad_norm": 3.780148983001709,
      "kl": 0.2548828125,
      "learning_rate": 9.776923076923077e-07,
      "loss": 0.1269,
      "reward": 0.6971118152141571,
      "reward_std": 0.11602437496185303,
      "reward_std_all": 0.11395644024014473,
      "rewards-max/forensics_sinkhorn_reward": 0.8636206984519958,
      "rewards-max/forensics_strict_edge_reward": 0.15689257474878104,
      "rewards-max/forensics_tiou_reward": 0.7009704411029816,
      "rewards/forensics_sinkhorn_reward": 0.5653184056282043,
      "rewards/forensics_strict_edge_reward": 0.011050556619011331,
      "rewards/forensics_tiou_reward": 0.2491288036108017,
      "step": 30
    },
    {
      "completion_length": 25.984375,
      "epoch": 0.09523809523809523,
      "grad_norm": 4.758464336395264,
      "kl": 0.2705078125,
      "learning_rate": 9.769230769230768e-07,
      "loss": 0.1122,
      "reward": 0.6796302497386932,
      "reward_std": 0.11415428295731544,
      "reward_std_all": 0.11606819182634354,
      "rewards-max/forensics_sinkhorn_reward": 0.881220132112503,
      "rewards-max/forensics_strict_edge_reward": 0.027609472366748378,
      "rewards-max/forensics_tiou_reward": 0.7145146727561951,
      "rewards/forensics_sinkhorn_reward": 0.5165582299232483,
      "rewards/forensics_strict_edge_reward": 0.0011515891887938778,
      "rewards/forensics_tiou_reward": 0.22577021270990372,
      "step": 31
    },
    {
      "completion_length": 24.5625,
      "epoch": 0.09831029185867896,
      "grad_norm": 5.121673107147217,
      "kl": 0.26171875,
      "learning_rate": 9.761538461538461e-07,
      "loss": 0.2293,
      "reward": 0.6485134661197662,
      "reward_std": 0.06093694269657135,
      "reward_std_all": 0.10342670232057571,
      "rewards-max/forensics_sinkhorn_reward": 0.8182037174701691,
      "rewards-max/forensics_strict_edge_reward": 0.01724346145056188,
      "rewards-max/forensics_tiou_reward": 0.6590011119842529,
      "rewards/forensics_sinkhorn_reward": 0.5210434198379517,
      "rewards/forensics_strict_edge_reward": 0.0008654790581203997,
      "rewards/forensics_tiou_reward": 0.1929239183664322,
      "step": 32
    },
    {
      "completion_length": 22.140625,
      "epoch": 0.10138248847926268,
      "grad_norm": 4.483846187591553,
      "kl": 0.228515625,
      "learning_rate": 9.753846153846154e-07,
      "loss": 0.2125,
      "reward": 0.6863099038600922,
      "reward_std": 0.09932651743292809,
      "reward_std_all": 0.10541805252432823,
      "rewards-max/forensics_sinkhorn_reward": 0.9503282308578491,
      "rewards-max/forensics_strict_edge_reward": 0.04688570275902748,
      "rewards-max/forensics_tiou_reward": 0.8806364238262177,
      "rewards/forensics_sinkhorn_reward": 0.5710239410400391,
      "rewards/forensics_strict_edge_reward": 0.001985614711884409,
      "rewards/forensics_tiou_reward": 0.24460162222385406,
      "step": 33
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.10445468509984639,
      "grad_norm": 5.831628799438477,
      "kl": 0.2314453125,
      "learning_rate": 9.746153846153845e-07,
      "loss": 0.1403,
      "reward": 0.626410961151123,
      "reward_std": 0.15915706753730774,
      "reward_std_all": 0.16301418095827103,
      "rewards-max/forensics_sinkhorn_reward": 0.8873429298400879,
      "rewards-max/forensics_strict_edge_reward": 0.017540502827614546,
      "rewards-max/forensics_tiou_reward": 0.7407320439815521,
      "rewards/forensics_sinkhorn_reward": 0.4989311397075653,
      "rewards/forensics_strict_edge_reward": 0.0008158170967362821,
      "rewards/forensics_tiou_reward": 0.1967586651444435,
      "step": 34
    },
    {
      "completion_length": 22.9375,
      "epoch": 0.10752688172043011,
      "grad_norm": 4.531584739685059,
      "kl": 0.27734375,
      "learning_rate": 9.738461538461539e-07,
      "loss": 0.2169,
      "reward": 0.6366090178489685,
      "reward_std": 0.14142496138811111,
      "reward_std_all": 0.14767279475927353,
      "rewards-max/forensics_sinkhorn_reward": 0.8759974837303162,
      "rewards-max/forensics_strict_edge_reward": 0.02122372854501009,
      "rewards-max/forensics_tiou_reward": 0.6764084696769714,
      "rewards/forensics_sinkhorn_reward": 0.5407557487487793,
      "rewards/forensics_strict_edge_reward": 0.0009695739718154073,
      "rewards/forensics_tiou_reward": 0.2251693606376648,
      "step": 35
    },
    {
      "completion_length": 22.109375,
      "epoch": 0.11059907834101383,
      "grad_norm": 5.716493606567383,
      "kl": 0.23388671875,
      "learning_rate": 9.73076923076923e-07,
      "loss": 0.1976,
      "reward": 0.6522999703884125,
      "reward_std": 0.11162745952606201,
      "reward_std_all": 0.13185536116361618,
      "rewards-max/forensics_sinkhorn_reward": 0.9465036690235138,
      "rewards-max/forensics_strict_edge_reward": 0.1331665739417076,
      "rewards-max/forensics_tiou_reward": 0.8917142748832703,
      "rewards/forensics_sinkhorn_reward": 0.5525940358638763,
      "rewards/forensics_strict_edge_reward": 0.006516592111438513,
      "rewards/forensics_tiou_reward": 0.23694618791341782,
      "step": 36
    },
    {
      "completion_length": 23.359375,
      "epoch": 0.11367127496159754,
      "grad_norm": 4.784858226776123,
      "kl": 0.216796875,
      "learning_rate": 9.723076923076923e-07,
      "loss": 0.0579,
      "reward": 0.7164542973041534,
      "reward_std": 0.08815527707338333,
      "reward_std_all": 0.08809789642691612,
      "rewards-max/forensics_sinkhorn_reward": 0.9323951900005341,
      "rewards-max/forensics_strict_edge_reward": 0.04389955662190914,
      "rewards-max/forensics_tiou_reward": 0.8128550946712494,
      "rewards/forensics_sinkhorn_reward": 0.6196878552436829,
      "rewards/forensics_strict_edge_reward": 0.002082064631395042,
      "rewards/forensics_tiou_reward": 0.3200177699327469,
      "step": 37
    },
    {
      "completion_length": 25.046875,
      "epoch": 0.11674347158218126,
      "grad_norm": 3.5947937965393066,
      "kl": 0.22607421875,
      "learning_rate": 9.715384615384616e-07,
      "loss": 0.1568,
      "reward": 0.6936796009540558,
      "reward_std": 0.11759147047996521,
      "reward_std_all": 0.114802535623312,
      "rewards-max/forensics_sinkhorn_reward": 0.8874393701553345,
      "rewards-max/forensics_strict_edge_reward": 0.18031708896160126,
      "rewards-max/forensics_tiou_reward": 0.7538883686065674,
      "rewards/forensics_sinkhorn_reward": 0.5771413445472717,
      "rewards/forensics_strict_edge_reward": 0.007620531367138028,
      "rewards/forensics_tiou_reward": 0.27652980387210846,
      "step": 38
    },
    {
      "completion_length": 23.078125,
      "epoch": 0.11981566820276497,
      "grad_norm": 3.6605939865112305,
      "kl": 0.232421875,
      "learning_rate": 9.707692307692307e-07,
      "loss": 0.1159,
      "reward": 0.6962751150131226,
      "reward_std": 0.10022244602441788,
      "reward_std_all": 0.10310691222548485,
      "rewards-max/forensics_sinkhorn_reward": 0.9037206470966339,
      "rewards-max/forensics_strict_edge_reward": 0.03410390194039792,
      "rewards-max/forensics_tiou_reward": 0.7628375589847565,
      "rewards/forensics_sinkhorn_reward": 0.596336841583252,
      "rewards/forensics_strict_edge_reward": 0.001476450572226895,
      "rewards/forensics_tiou_reward": 0.25905323028564453,
      "step": 39
    },
    {
      "completion_length": 22.765625,
      "epoch": 0.1228878648233487,
      "grad_norm": 5.418461322784424,
      "kl": 0.24658203125,
      "learning_rate": 9.7e-07,
      "loss": 0.1961,
      "reward": 0.6717543005943298,
      "reward_std": 0.14227843284606934,
      "reward_std_all": 0.13719970360398293,
      "rewards-max/forensics_sinkhorn_reward": 0.943099319934845,
      "rewards-max/forensics_strict_edge_reward": 0.08912976988358423,
      "rewards-max/forensics_tiou_reward": 0.8617689907550812,
      "rewards/forensics_sinkhorn_reward": 0.5839158892631531,
      "rewards/forensics_strict_edge_reward": 0.005463194604089949,
      "rewards/forensics_tiou_reward": 0.31967486441135406,
      "step": 40
    },
    {
      "completion_length": 21.8125,
      "epoch": 0.1259600614439324,
      "grad_norm": 6.331235885620117,
      "kl": 0.234375,
      "learning_rate": 9.692307692307691e-07,
      "loss": 0.1149,
      "reward": 0.6746267974376678,
      "reward_std": 0.12860628217458725,
      "reward_std_all": 0.12742450833320618,
      "rewards-max/forensics_sinkhorn_reward": 0.9293773174285889,
      "rewards-max/forensics_strict_edge_reward": 0.09262130595743656,
      "rewards-max/forensics_tiou_reward": 0.8498240411281586,
      "rewards/forensics_sinkhorn_reward": 0.5822104215621948,
      "rewards/forensics_strict_edge_reward": 0.0109390530269593,
      "rewards/forensics_tiou_reward": 0.2778351455926895,
      "step": 41
    },
    {
      "completion_length": 23.25,
      "epoch": 0.12903225806451613,
      "grad_norm": 4.789433002471924,
      "kl": 0.2255859375,
      "learning_rate": 9.684615384615385e-07,
      "loss": 0.0678,
      "reward": 0.7018954157829285,
      "reward_std": 0.11148649454116821,
      "reward_std_all": 0.10754229128360748,
      "rewards-max/forensics_sinkhorn_reward": 0.9186494946479797,
      "rewards-max/forensics_strict_edge_reward": 0.004764646990224719,
      "rewards-max/forensics_tiou_reward": 0.810422271490097,
      "rewards/forensics_sinkhorn_reward": 0.593712329864502,
      "rewards/forensics_strict_edge_reward": 0.0002524163937778212,
      "rewards/forensics_tiou_reward": 0.27382056415081024,
      "step": 42
    },
    {
      "completion_length": 25.140625,
      "epoch": 0.13210445468509985,
      "grad_norm": 3.5653481483459473,
      "kl": 0.2158203125,
      "learning_rate": 9.676923076923078e-07,
      "loss": 0.1652,
      "reward": 0.6904636025428772,
      "reward_std": 0.1182481236755848,
      "reward_std_all": 0.11639522761106491,
      "rewards-max/forensics_sinkhorn_reward": 0.927133172750473,
      "rewards-max/forensics_strict_edge_reward": 0.06811708537861705,
      "rewards-max/forensics_tiou_reward": 0.8434119522571564,
      "rewards/forensics_sinkhorn_reward": 0.5665176212787628,
      "rewards/forensics_strict_edge_reward": 0.002927440495113842,
      "rewards/forensics_tiou_reward": 0.2955707013607025,
      "step": 43
    },
    {
      "completion_length": 24.46875,
      "epoch": 0.13517665130568357,
      "grad_norm": 5.997074127197266,
      "kl": 0.25048828125,
      "learning_rate": 9.669230769230769e-07,
      "loss": 0.184,
      "reward": 0.6540962159633636,
      "reward_std": 0.12383687123656273,
      "reward_std_all": 0.1339123174548149,
      "rewards-max/forensics_sinkhorn_reward": 0.9629330933094025,
      "rewards-max/forensics_strict_edge_reward": 0.06894902558997273,
      "rewards-max/forensics_tiou_reward": 0.9133732318878174,
      "rewards/forensics_sinkhorn_reward": 0.526799887418747,
      "rewards/forensics_strict_edge_reward": 0.0030562529136659577,
      "rewards/forensics_tiou_reward": 0.24012598395347595,
      "step": 44
    },
    {
      "completion_length": 25.984375,
      "epoch": 0.1382488479262673,
      "grad_norm": 4.438931941986084,
      "kl": 0.2080078125,
      "learning_rate": 9.66153846153846e-07,
      "loss": 0.0951,
      "reward": 0.682368814945221,
      "reward_std": 0.11907115951180458,
      "reward_std_all": 0.11833671107888222,
      "rewards-max/forensics_sinkhorn_reward": 0.9316262602806091,
      "rewards-max/forensics_strict_edge_reward": 0.06162173207849264,
      "rewards-max/forensics_tiou_reward": 0.8390675187110901,
      "rewards/forensics_sinkhorn_reward": 0.5247065275907516,
      "rewards/forensics_strict_edge_reward": 0.004897169237665366,
      "rewards/forensics_tiou_reward": 0.2576906085014343,
      "step": 45
    },
    {
      "completion_length": 23.703125,
      "epoch": 0.141321044546851,
      "grad_norm": 4.281137466430664,
      "kl": 0.185546875,
      "learning_rate": 9.653846153846153e-07,
      "loss": 0.2056,
      "reward": 0.6761369109153748,
      "reward_std": 0.13219496607780457,
      "reward_std_all": 0.12867975607514381,
      "rewards-max/forensics_sinkhorn_reward": 0.9517321288585663,
      "rewards-max/forensics_strict_edge_reward": 0.2545395717024803,
      "rewards-max/forensics_tiou_reward": 0.9053380787372589,
      "rewards/forensics_sinkhorn_reward": 0.5678829848766327,
      "rewards/forensics_strict_edge_reward": 0.012881679460406303,
      "rewards/forensics_tiou_reward": 0.2771346867084503,
      "step": 46
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.1443932411674347,
      "grad_norm": 4.223880767822266,
      "kl": 0.20068359375,
      "learning_rate": 9.646153846153846e-07,
      "loss": 0.0792,
      "reward": 0.6691817343235016,
      "reward_std": 0.14014269411563873,
      "reward_std_all": 0.13679108023643494,
      "rewards-max/forensics_sinkhorn_reward": 0.8827645778656006,
      "rewards-max/forensics_strict_edge_reward": 0.0200788548681885,
      "rewards-max/forensics_tiou_reward": 0.730121910572052,
      "rewards/forensics_sinkhorn_reward": 0.5661272257566452,
      "rewards/forensics_strict_edge_reward": 0.000912042916752398,
      "rewards/forensics_tiou_reward": 0.2380741685628891,
      "step": 47
    },
    {
      "completion_length": 26.59375,
      "epoch": 0.14746543778801843,
      "grad_norm": 4.00600528717041,
      "kl": 0.19287109375,
      "learning_rate": 9.638461538461537e-07,
      "loss": 0.1976,
      "reward": 0.6885904371738434,
      "reward_std": 0.1368827000260353,
      "reward_std_all": 0.129411481320858,
      "rewards-max/forensics_sinkhorn_reward": 0.9567147791385651,
      "rewards-max/forensics_strict_edge_reward": 0.3095208704471588,
      "rewards-max/forensics_tiou_reward": 0.9175335168838501,
      "rewards/forensics_sinkhorn_reward": 0.5964789092540741,
      "rewards/forensics_strict_edge_reward": 0.013244772097095847,
      "rewards/forensics_tiou_reward": 0.328886479139328,
      "step": 48
    },
    {
      "completion_length": 26.328125,
      "epoch": 0.15053763440860216,
      "grad_norm": 3.1951377391815186,
      "kl": 0.17822265625,
      "learning_rate": 9.63076923076923e-07,
      "loss": 0.1118,
      "reward": 0.709286242723465,
      "reward_std": 0.11702793836593628,
      "reward_std_all": 0.11155247688293457,
      "rewards-max/forensics_sinkhorn_reward": 0.9572377502918243,
      "rewards-max/forensics_strict_edge_reward": 0.18679149961099029,
      "rewards-max/forensics_tiou_reward": 0.9107157289981842,
      "rewards/forensics_sinkhorn_reward": 0.5997984409332275,
      "rewards/forensics_strict_edge_reward": 0.007894023205153644,
      "rewards/forensics_tiou_reward": 0.33803023397922516,
      "step": 49
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.15360983102918588,
      "grad_norm": 5.540486812591553,
      "kl": 0.251953125,
      "learning_rate": 9.623076923076922e-07,
      "loss": 0.1463,
      "reward": 0.6482558250427246,
      "reward_std": 0.12634561955928802,
      "reward_std_all": 0.13491751998662949,
      "rewards-max/forensics_sinkhorn_reward": 0.8862074911594391,
      "rewards-max/forensics_strict_edge_reward": 0.023611271870322526,
      "rewards-max/forensics_tiou_reward": 0.7288515567779541,
      "rewards/forensics_sinkhorn_reward": 0.531125009059906,
      "rewards/forensics_strict_edge_reward": 0.0015615371376043186,
      "rewards/forensics_tiou_reward": 0.22952765226364136,
      "step": 50
    },
    {
      "completion_length": 21.953125,
      "epoch": 0.15668202764976957,
      "grad_norm": 3.7839488983154297,
      "kl": 0.2080078125,
      "learning_rate": 9.615384615384615e-07,
      "loss": 0.2749,
      "reward": 0.6415822803974152,
      "reward_std": 0.14038866758346558,
      "reward_std_all": 0.14771518111228943,
      "rewards-max/forensics_sinkhorn_reward": 0.9299024641513824,
      "rewards-max/forensics_strict_edge_reward": 0.05733165495257708,
      "rewards-max/forensics_tiou_reward": 0.8063192665576935,
      "rewards/forensics_sinkhorn_reward": 0.5351508855819702,
      "rewards/forensics_strict_edge_reward": 0.0026867529611536156,
      "rewards/forensics_tiou_reward": 0.2007523477077484,
      "step": 51
    },
    {
      "completion_length": 25.09375,
      "epoch": 0.1597542242703533,
      "grad_norm": 4.020214557647705,
      "kl": 0.212890625,
      "learning_rate": 9.607692307692308e-07,
      "loss": 0.166,
      "reward": 0.6208245158195496,
      "reward_std": 0.12591831386089325,
      "reward_std_all": 0.14831425994634628,
      "rewards-max/forensics_sinkhorn_reward": 0.9343450963497162,
      "rewards-max/forensics_strict_edge_reward": 0.036214907973771915,
      "rewards-max/forensics_tiou_reward": 0.8469762206077576,
      "rewards/forensics_sinkhorn_reward": 0.5091472566127777,
      "rewards/forensics_strict_edge_reward": 0.0015606709603162017,
      "rewards/forensics_tiou_reward": 0.22282514721155167,
      "step": 52
    },
    {
      "completion_length": 25.234375,
      "epoch": 0.16282642089093702,
      "grad_norm": 3.652981996536255,
      "kl": 0.216796875,
      "learning_rate": 9.6e-07,
      "loss": 0.1344,
      "reward": 0.662017822265625,
      "reward_std": 0.13527057692408562,
      "reward_std_all": 0.13740131631493568,
      "rewards-max/forensics_sinkhorn_reward": 0.939992368221283,
      "rewards-max/forensics_strict_edge_reward": 0.002190087892813608,
      "rewards-max/forensics_tiou_reward": 0.8586106300354004,
      "rewards/forensics_sinkhorn_reward": 0.5475860089063644,
      "rewards/forensics_strict_edge_reward": 0.00014211130837793462,
      "rewards/forensics_tiou_reward": 0.273717999458313,
      "step": 53
    },
    {
      "completion_length": 24.421875,
      "epoch": 0.16589861751152074,
      "grad_norm": 4.520156383514404,
      "kl": 0.2353515625,
      "learning_rate": 9.592307692307692e-07,
      "loss": 0.1237,
      "reward": 0.6965833902359009,
      "reward_std": 0.11241995543241501,
      "reward_std_all": 0.11167731508612633,
      "rewards-max/forensics_sinkhorn_reward": 0.9219892024993896,
      "rewards-max/forensics_strict_edge_reward": 0.21785599132999778,
      "rewards-max/forensics_tiou_reward": 0.8030133247375488,
      "rewards/forensics_sinkhorn_reward": 0.60343998670578,
      "rewards/forensics_strict_edge_reward": 0.009161495145235676,
      "rewards/forensics_tiou_reward": 0.2994960770010948,
      "step": 54
    },
    {
      "completion_length": 24.390625,
      "epoch": 0.16897081413210446,
      "grad_norm": 4.728549957275391,
      "kl": 0.24609375,
      "learning_rate": 9.584615384615383e-07,
      "loss": 0.2309,
      "reward": 0.7121903002262115,
      "reward_std": 0.10021183267235756,
      "reward_std_all": 0.09917710721492767,
      "rewards-max/forensics_sinkhorn_reward": 0.9164549112319946,
      "rewards-max/forensics_strict_edge_reward": 0.04065893776714802,
      "rewards-max/forensics_tiou_reward": 0.7952257692813873,
      "rewards/forensics_sinkhorn_reward": 0.5933344960212708,
      "rewards/forensics_strict_edge_reward": 0.001808517030440271,
      "rewards/forensics_tiou_reward": 0.30104804039001465,
      "step": 55
    },
    {
      "completion_length": 23.578125,
      "epoch": 0.17204301075268819,
      "grad_norm": 4.621572017669678,
      "kl": 0.216796875,
      "learning_rate": 9.576923076923077e-07,
      "loss": -0.0443,
      "reward": 0.5781574547290802,
      "reward_std": 0.13143355026841164,
      "reward_std_all": 0.1669558435678482,
      "rewards-max/forensics_sinkhorn_reward": 0.8956639766693115,
      "rewards-max/forensics_strict_edge_reward": 0.1082392914686352,
      "rewards-max/forensics_tiou_reward": 0.7660687863826752,
      "rewards/forensics_sinkhorn_reward": 0.46816299855709076,
      "rewards/forensics_strict_edge_reward": 0.005669844773365185,
      "rewards/forensics_tiou_reward": 0.18065637350082397,
      "step": 56
    },
    {
      "completion_length": 21.9375,
      "epoch": 0.17511520737327188,
      "grad_norm": 3.9565727710723877,
      "kl": 0.20263671875,
      "learning_rate": 9.569230769230768e-07,
      "loss": 0.1024,
      "reward": 0.7005904614925385,
      "reward_std": 0.10823103040456772,
      "reward_std_all": 0.10713116079568863,
      "rewards-max/forensics_sinkhorn_reward": 0.947025716304779,
      "rewards-max/forensics_strict_edge_reward": 0.15634983777999878,
      "rewards-max/forensics_tiou_reward": 0.8916805386543274,
      "rewards/forensics_sinkhorn_reward": 0.6046976447105408,
      "rewards/forensics_strict_edge_reward": 0.006652422482147813,
      "rewards/forensics_tiou_reward": 0.28274014592170715,
      "step": 57
    },
    {
      "completion_length": 25.4375,
      "epoch": 0.1781874039938556,
      "grad_norm": 3.5124404430389404,
      "kl": 0.2392578125,
      "learning_rate": 9.56153846153846e-07,
      "loss": 0.218,
      "reward": 0.6876277923583984,
      "reward_std": 0.11255007609724998,
      "reward_std_all": 0.11415921896696091,
      "rewards-max/forensics_sinkhorn_reward": 0.9556393623352051,
      "rewards-max/forensics_strict_edge_reward": 0.10908503457903862,
      "rewards-max/forensics_tiou_reward": 0.9037748873233795,
      "rewards/forensics_sinkhorn_reward": 0.5841823816299438,
      "rewards/forensics_strict_edge_reward": 0.005445161601528525,
      "rewards/forensics_tiou_reward": 0.3437400162220001,
      "step": 58
    },
    {
      "completion_length": 24.171875,
      "epoch": 0.18125960061443933,
      "grad_norm": 5.916181564331055,
      "kl": 0.19140625,
      "learning_rate": 9.553846153846154e-07,
      "loss": 0.2144,
      "reward": 0.7038968503475189,
      "reward_std": 0.08388814330101013,
      "reward_std_all": 0.09154055640101433,
      "rewards-max/forensics_sinkhorn_reward": 0.9049029648303986,
      "rewards-max/forensics_strict_edge_reward": 0.02181563270460174,
      "rewards-max/forensics_tiou_reward": 0.798069030046463,
      "rewards/forensics_sinkhorn_reward": 0.5685240030288696,
      "rewards/forensics_strict_edge_reward": 0.0009117559568494471,
      "rewards/forensics_tiou_reward": 0.28518857061862946,
      "step": 59
    },
    {
      "completion_length": 21.03125,
      "epoch": 0.18433179723502305,
      "grad_norm": 4.08756160736084,
      "kl": 0.1845703125,
      "learning_rate": 9.546153846153845e-07,
      "loss": 0.265,
      "reward": 0.6878965198993683,
      "reward_std": 0.10821586102247238,
      "reward_std_all": 0.11257505416870117,
      "rewards-max/forensics_sinkhorn_reward": 0.8939732611179352,
      "rewards-max/forensics_strict_edge_reward": 0.00891306926496327,
      "rewards-max/forensics_tiou_reward": 0.7666912078857422,
      "rewards/forensics_sinkhorn_reward": 0.5713842511177063,
      "rewards/forensics_strict_edge_reward": 0.0005783725209766999,
      "rewards/forensics_tiou_reward": 0.2630963921546936,
      "step": 60
    },
    {
      "completion_length": 25.578125,
      "epoch": 0.18740399385560677,
      "grad_norm": 4.396356582641602,
      "kl": 0.25732421875,
      "learning_rate": 9.538461538461538e-07,
      "loss": 0.1814,
      "reward": 0.6941646933555603,
      "reward_std": 0.09326538443565369,
      "reward_std_all": 0.09755650162696838,
      "rewards-max/forensics_sinkhorn_reward": 0.8758002519607544,
      "rewards-max/forensics_strict_edge_reward": 0.00878611032385379,
      "rewards-max/forensics_tiou_reward": 0.6832731366157532,
      "rewards/forensics_sinkhorn_reward": 0.5545637905597687,
      "rewards/forensics_strict_edge_reward": 0.0003730930729943793,
      "rewards/forensics_tiou_reward": 0.22625187039375305,
      "step": 61
    },
    {
      "completion_length": 24.0625,
      "epoch": 0.19047619047619047,
      "grad_norm": 4.651092052459717,
      "kl": 0.259765625,
      "learning_rate": 9.53076923076923e-07,
      "loss": 0.1665,
      "reward": 0.7059653103351593,
      "reward_std": 0.08621376566588879,
      "reward_std_all": 0.09103208407759666,
      "rewards-max/forensics_sinkhorn_reward": 0.874738872051239,
      "rewards-max/forensics_strict_edge_reward": 0.1314297430217266,
      "rewards-max/forensics_tiou_reward": 0.7025148868560791,
      "rewards/forensics_sinkhorn_reward": 0.5760381519794464,
      "rewards/forensics_strict_edge_reward": 0.005826393608003855,
      "rewards/forensics_tiou_reward": 0.25847596675157547,
      "step": 62
    },
    {
      "completion_length": 23.71875,
      "epoch": 0.1935483870967742,
      "grad_norm": 5.851619720458984,
      "kl": 0.20556640625,
      "learning_rate": 9.523076923076923e-07,
      "loss": 0.071,
      "reward": 0.585685670375824,
      "reward_std": 0.15125028789043427,
      "reward_std_all": 0.18103855848312378,
      "rewards-max/forensics_sinkhorn_reward": 0.8112735748291016,
      "rewards-max/forensics_strict_edge_reward": 0.05599361308850348,
      "rewards-max/forensics_tiou_reward": 0.6266713440418243,
      "rewards/forensics_sinkhorn_reward": 0.4557487517595291,
      "rewards/forensics_strict_edge_reward": 0.0034187280289188493,
      "rewards/forensics_tiou_reward": 0.18509086593985558,
      "step": 63
    },
    {
      "completion_length": 25.515625,
      "epoch": 0.1966205837173579,
      "grad_norm": 4.566220760345459,
      "kl": 0.185546875,
      "learning_rate": 9.515384615384615e-07,
      "loss": 0.1731,
      "reward": 0.6228783428668976,
      "reward_std": 0.12943247333168983,
      "reward_std_all": 0.1520153060555458,
      "rewards-max/forensics_sinkhorn_reward": 0.91466224193573,
      "rewards-max/forensics_strict_edge_reward": 0.17441173130646348,
      "rewards-max/forensics_tiou_reward": 0.8197680711746216,
      "rewards/forensics_sinkhorn_reward": 0.49235713481903076,
      "rewards/forensics_strict_edge_reward": 0.009041067445650697,
      "rewards/forensics_tiou_reward": 0.2134733349084854,
      "step": 64
    },
    {
      "completion_length": 21.53125,
      "epoch": 0.19969278033794163,
      "grad_norm": 3.894688129425049,
      "kl": 0.21044921875,
      "learning_rate": 9.507692307692308e-07,
      "loss": 0.1692,
      "reward": 0.7068492472171783,
      "reward_std": 0.09316233545541763,
      "reward_std_all": 0.09664830192923546,
      "rewards-max/forensics_sinkhorn_reward": 0.9304423630237579,
      "rewards-max/forensics_strict_edge_reward": 0.07511642016470432,
      "rewards-max/forensics_tiou_reward": 0.8416775465011597,
      "rewards/forensics_sinkhorn_reward": 0.6133661270141602,
      "rewards/forensics_strict_edge_reward": 0.004915842437185347,
      "rewards/forensics_tiou_reward": 0.2969782054424286,
      "step": 65
    },
    {
      "completion_length": 21.03125,
      "epoch": 0.20276497695852536,
      "grad_norm": 3.3687283992767334,
      "kl": 0.279296875,
      "learning_rate": 9.499999999999999e-07,
      "loss": 0.1304,
      "reward": 0.6702835857868195,
      "reward_std": 0.10666775330901146,
      "reward_std_all": 0.1190370675176382,
      "rewards-max/forensics_sinkhorn_reward": 0.9304315447807312,
      "rewards-max/forensics_strict_edge_reward": 0.04998040571808815,
      "rewards-max/forensics_tiou_reward": 0.8498913049697876,
      "rewards/forensics_sinkhorn_reward": 0.5805872827768326,
      "rewards/forensics_strict_edge_reward": 0.002139938878826797,
      "rewards/forensics_tiou_reward": 0.2835254520177841,
      "step": 66
    },
    {
      "completion_length": 23.5,
      "epoch": 0.20583717357910905,
      "grad_norm": 3.8908674716949463,
      "kl": 0.19677734375,
      "learning_rate": 9.492307692307691e-07,
      "loss": 0.1286,
      "reward": 0.6535028219223022,
      "reward_std": 0.18202261626720428,
      "reward_std_all": 0.17062987387180328,
      "rewards-max/forensics_sinkhorn_reward": 0.8442940711975098,
      "rewards-max/forensics_strict_edge_reward": 0.07283941097557545,
      "rewards-max/forensics_tiou_reward": 0.7431508600711823,
      "rewards/forensics_sinkhorn_reward": 0.5329144448041916,
      "rewards/forensics_strict_edge_reward": 0.0037355145323090255,
      "rewards/forensics_tiou_reward": 0.2698228359222412,
      "step": 67
    },
    {
      "completion_length": 20.734375,
      "epoch": 0.20890937019969277,
      "grad_norm": 4.50151252746582,
      "kl": 0.2158203125,
      "learning_rate": 9.484615384615384e-07,
      "loss": 0.1008,
      "reward": 0.6831488609313965,
      "reward_std": 0.1103062704205513,
      "reward_std_all": 0.11303595826029778,
      "rewards-max/forensics_sinkhorn_reward": 0.9245857298374176,
      "rewards-max/forensics_strict_edge_reward": 0.07008161582052708,
      "rewards-max/forensics_tiou_reward": 0.8387579917907715,
      "rewards/forensics_sinkhorn_reward": 0.5752860307693481,
      "rewards/forensics_strict_edge_reward": 0.003221528386347927,
      "rewards/forensics_tiou_reward": 0.2609328478574753,
      "step": 68
    },
    {
      "completion_length": 23.203125,
      "epoch": 0.2119815668202765,
      "grad_norm": 4.471990585327148,
      "kl": 0.25732421875,
      "learning_rate": 9.476923076923077e-07,
      "loss": 0.0621,
      "reward": 0.7060373425483704,
      "reward_std": 0.11511185020208359,
      "reward_std_all": 0.10960255935788155,
      "rewards-max/forensics_sinkhorn_reward": 0.9853708148002625,
      "rewards-max/forensics_strict_edge_reward": 0.22505103051662445,
      "rewards-max/forensics_tiou_reward": 0.9884827733039856,
      "rewards/forensics_sinkhorn_reward": 0.6207389533519745,
      "rewards/forensics_strict_edge_reward": 0.01484599310788326,
      "rewards/forensics_tiou_reward": 0.33714742958545685,
      "step": 69
    },
    {
      "completion_length": 21.71875,
      "epoch": 0.21505376344086022,
      "grad_norm": 5.198028087615967,
      "kl": 0.2265625,
      "learning_rate": 9.46923076923077e-07,
      "loss": 0.1509,
      "reward": 0.6720921099185944,
      "reward_std": 0.08053306303918362,
      "reward_std_all": 0.0999673344194889,
      "rewards-max/forensics_sinkhorn_reward": 0.8196378350257874,
      "rewards-max/forensics_strict_edge_reward": 0.0010051997378468513,
      "rewards-max/forensics_tiou_reward": 0.6122701466083527,
      "rewards/forensics_sinkhorn_reward": 0.5515009164810181,
      "rewards/forensics_strict_edge_reward": 4.680805614043493e-05,
      "rewards/forensics_tiou_reward": 0.22962268441915512,
      "step": 70
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.21812596006144394,
      "grad_norm": 5.276992321014404,
      "kl": 0.21435546875,
      "learning_rate": 9.461538461538461e-07,
      "loss": 0.16,
      "reward": 0.6744497120380402,
      "reward_std": 0.1312561333179474,
      "reward_std_all": 0.13018210977315903,
      "rewards-max/forensics_sinkhorn_reward": 0.9265616238117218,
      "rewards-max/forensics_strict_edge_reward": 0.06076258816756308,
      "rewards-max/forensics_tiou_reward": 0.8287568986415863,
      "rewards/forensics_sinkhorn_reward": 0.5887489914894104,
      "rewards/forensics_strict_edge_reward": 0.004085289721842855,
      "rewards/forensics_tiou_reward": 0.31411366164684296,
      "step": 71
    },
    {
      "completion_length": 23.640625,
      "epoch": 0.22119815668202766,
      "grad_norm": 5.203954219818115,
      "kl": 0.22314453125,
      "learning_rate": 9.453846153846153e-07,
      "loss": -0.0068,
      "reward": 0.6342475414276123,
      "reward_std": 0.12663494795560837,
      "reward_std_all": 0.14386483281850815,
      "rewards-max/forensics_sinkhorn_reward": 0.9722029566764832,
      "rewards-max/forensics_strict_edge_reward": 0.2879621237516403,
      "rewards-max/forensics_tiou_reward": 0.950395256280899,
      "rewards/forensics_sinkhorn_reward": 0.5153399705886841,
      "rewards/forensics_strict_edge_reward": 0.012807569233700633,
      "rewards/forensics_tiou_reward": 0.24881017208099365,
      "step": 72
    },
    {
      "completion_length": 23.171875,
      "epoch": 0.22427035330261136,
      "grad_norm": 4.530182838439941,
      "kl": 0.25390625,
      "learning_rate": 9.446153846153846e-07,
      "loss": 0.1095,
      "reward": 0.6973670125007629,
      "reward_std": 0.09896047785878181,
      "reward_std_all": 0.10323981568217278,
      "rewards-max/forensics_sinkhorn_reward": 0.9053173363208771,
      "rewards-max/forensics_strict_edge_reward": 0.20266746822744608,
      "rewards-max/forensics_tiou_reward": 0.7655304670333862,
      "rewards/forensics_sinkhorn_reward": 0.6016183197498322,
      "rewards/forensics_strict_edge_reward": 0.008807172474917024,
      "rewards/forensics_tiou_reward": 0.30303676426410675,
      "step": 73
    },
    {
      "completion_length": 23.1875,
      "epoch": 0.22734254992319508,
      "grad_norm": 3.963338613510132,
      "kl": 0.20458984375,
      "learning_rate": 9.438461538461538e-07,
      "loss": 0.0381,
      "reward": 0.7310169339179993,
      "reward_std": 0.08960521221160889,
      "reward_std_all": 0.08605397492647171,
      "rewards-max/forensics_sinkhorn_reward": 0.9496181905269623,
      "rewards-max/forensics_strict_edge_reward": 0.2682615122757852,
      "rewards-max/forensics_tiou_reward": 0.8688009679317474,
      "rewards/forensics_sinkhorn_reward": 0.635641872882843,
      "rewards/forensics_strict_edge_reward": 0.012270821258425713,
      "rewards/forensics_tiou_reward": 0.3637392669916153,
      "step": 74
    },
    {
      "completion_length": 25.5,
      "epoch": 0.2304147465437788,
      "grad_norm": 3.3903372287750244,
      "kl": 0.24755859375,
      "learning_rate": 9.43076923076923e-07,
      "loss": 0.196,
      "reward": 0.6552876830101013,
      "reward_std": 0.10704804211854935,
      "reward_std_all": 0.1231292337179184,
      "rewards-max/forensics_sinkhorn_reward": 0.8817733824253082,
      "rewards-max/forensics_strict_edge_reward": 0.10405873134732246,
      "rewards-max/forensics_tiou_reward": 0.7827593684196472,
      "rewards/forensics_sinkhorn_reward": 0.5199103802442551,
      "rewards/forensics_strict_edge_reward": 0.004821747250389308,
      "rewards/forensics_tiou_reward": 0.2496447190642357,
      "step": 75
    },
    {
      "completion_length": 22.09375,
      "epoch": 0.23348694316436253,
      "grad_norm": 3.989757537841797,
      "kl": 0.197265625,
      "learning_rate": 9.423076923076923e-07,
      "loss": 0.0003,
      "reward": 0.6641891300678253,
      "reward_std": 0.11606794595718384,
      "reward_std_all": 0.1272040605545044,
      "rewards-max/forensics_sinkhorn_reward": 0.9479692578315735,
      "rewards-max/forensics_strict_edge_reward": 0.10611733794212341,
      "rewards-max/forensics_tiou_reward": 0.8827633857727051,
      "rewards/forensics_sinkhorn_reward": 0.5596444010734558,
      "rewards/forensics_strict_edge_reward": 0.004585970193147659,
      "rewards/forensics_tiou_reward": 0.2726346179842949,
      "step": 76
    },
    {
      "completion_length": 21.359375,
      "epoch": 0.23655913978494625,
      "grad_norm": 4.514549255371094,
      "kl": 0.2001953125,
      "learning_rate": 9.415384615384615e-07,
      "loss": 0.1689,
      "reward": 0.7103777527809143,
      "reward_std": 0.07412565685808659,
      "reward_std_all": 0.08163403533399105,
      "rewards-max/forensics_sinkhorn_reward": 0.9722699522972107,
      "rewards-max/forensics_strict_edge_reward": 0.18639911711215973,
      "rewards-max/forensics_tiou_reward": 0.9530121386051178,
      "rewards/forensics_sinkhorn_reward": 0.5993285477161407,
      "rewards/forensics_strict_edge_reward": 0.008012991398572922,
      "rewards/forensics_tiou_reward": 0.2843260318040848,
      "step": 77
    },
    {
      "completion_length": 24.609375,
      "epoch": 0.23963133640552994,
      "grad_norm": 3.3020992279052734,
      "kl": 0.21435546875,
      "learning_rate": 9.407692307692308e-07,
      "loss": 0.1026,
      "reward": 0.67348912358284,
      "reward_std": 0.12198536098003387,
      "reward_std_all": 0.12196852266788483,
      "rewards-max/forensics_sinkhorn_reward": 0.9323428869247437,
      "rewards-max/forensics_strict_edge_reward": 0.006471386644989252,
      "rewards-max/forensics_tiou_reward": 0.8362113237380981,
      "rewards/forensics_sinkhorn_reward": 0.5548929870128632,
      "rewards/forensics_strict_edge_reward": 0.00029004485986661166,
      "rewards/forensics_tiou_reward": 0.2543860524892807,
      "step": 78
    },
    {
      "completion_length": 24.40625,
      "epoch": 0.24270353302611367,
      "grad_norm": 5.187069416046143,
      "kl": 0.24169921875,
      "learning_rate": 9.399999999999999e-07,
      "loss": 0.2391,
      "reward": 0.6807314157485962,
      "reward_std": 0.11176726594567299,
      "reward_std_all": 0.11496254801750183,
      "rewards-max/forensics_sinkhorn_reward": 0.899617999792099,
      "rewards-max/forensics_strict_edge_reward": 0.24017135612666607,
      "rewards-max/forensics_tiou_reward": 0.7266903221607208,
      "rewards/forensics_sinkhorn_reward": 0.5580211877822876,
      "rewards/forensics_strict_edge_reward": 0.011007884371792898,
      "rewards/forensics_tiou_reward": 0.2394769936800003,
      "step": 79
    },
    {
      "completion_length": 20.625,
      "epoch": 0.2457757296466974,
      "grad_norm": 5.882438659667969,
      "kl": 0.2392578125,
      "learning_rate": 9.392307692307691e-07,
      "loss": 0.0997,
      "reward": 0.7006378769874573,
      "reward_std": 0.10064864531159401,
      "reward_std_all": 0.10424996539950371,
      "rewards-max/forensics_sinkhorn_reward": 0.9353851079940796,
      "rewards-max/forensics_strict_edge_reward": 0.045362033574519955,
      "rewards-max/forensics_tiou_reward": 0.8633131384849548,
      "rewards/forensics_sinkhorn_reward": 0.5985448658466339,
      "rewards/forensics_strict_edge_reward": 0.0037332704965820085,
      "rewards/forensics_tiou_reward": 0.24234553426504135,
      "step": 80
    },
    {
      "completion_length": 22.734375,
      "epoch": 0.2488479262672811,
      "grad_norm": 4.930925369262695,
      "kl": 0.240234375,
      "learning_rate": 9.384615384615384e-07,
      "loss": 0.1242,
      "reward": 0.7010170519351959,
      "reward_std": 0.09345346316695213,
      "reward_std_all": 0.09960589557886124,
      "rewards-max/forensics_sinkhorn_reward": 0.9123463332653046,
      "rewards-max/forensics_strict_edge_reward": 0.0304800383746624,
      "rewards-max/forensics_tiou_reward": 0.7916099429130554,
      "rewards/forensics_sinkhorn_reward": 0.6022785305976868,
      "rewards/forensics_strict_edge_reward": 0.0014825110993115231,
      "rewards/forensics_tiou_reward": 0.25478217750787735,
      "step": 81
    },
    {
      "completion_length": 24.71875,
      "epoch": 0.2519201228878648,
      "grad_norm": 3.8977222442626953,
      "kl": 0.205078125,
      "learning_rate": 9.376923076923076e-07,
      "loss": 0.1916,
      "reward": 0.650360643863678,
      "reward_std": 0.12250041216611862,
      "reward_std_all": 0.13610364496707916,
      "rewards-max/forensics_sinkhorn_reward": 0.89544877409935,
      "rewards-max/forensics_strict_edge_reward": 0.0031242581680999137,
      "rewards-max/forensics_tiou_reward": 0.7084740400314331,
      "rewards/forensics_sinkhorn_reward": 0.527858555316925,
      "rewards/forensics_strict_edge_reward": 0.00013304084677656647,
      "rewards/forensics_tiou_reward": 0.24520666897296906,
      "step": 82
    },
    {
      "completion_length": 22.546875,
      "epoch": 0.25499231950844853,
      "grad_norm": 4.606162071228027,
      "kl": 0.2470703125,
      "learning_rate": 9.36923076923077e-07,
      "loss": 0.263,
      "reward": 0.6957539021968842,
      "reward_std": 0.10029266402125359,
      "reward_std_all": 0.10185574367642403,
      "rewards-max/forensics_sinkhorn_reward": 0.9158609509468079,
      "rewards-max/forensics_strict_edge_reward": 0.26467981189489365,
      "rewards-max/forensics_tiou_reward": 0.8076854646205902,
      "rewards/forensics_sinkhorn_reward": 0.578770250082016,
      "rewards/forensics_strict_edge_reward": 0.015395773574709892,
      "rewards/forensics_tiou_reward": 0.3071186989545822,
      "step": 83
    },
    {
      "completion_length": 21.65625,
      "epoch": 0.25806451612903225,
      "grad_norm": 4.2795844078063965,
      "kl": 0.1904296875,
      "learning_rate": 9.361538461538461e-07,
      "loss": 0.1083,
      "reward": 0.7223220467567444,
      "reward_std": 0.07857278920710087,
      "reward_std_all": 0.07881012931466103,
      "rewards-max/forensics_sinkhorn_reward": 0.9155991673469543,
      "rewards-max/forensics_strict_edge_reward": 0.031091660261154175,
      "rewards-max/forensics_tiou_reward": 0.7987775504589081,
      "rewards/forensics_sinkhorn_reward": 0.6403865814208984,
      "rewards/forensics_strict_edge_reward": 0.0031710366893094033,
      "rewards/forensics_tiou_reward": 0.32969075441360474,
      "step": 84
    },
    {
      "completion_length": 23.671875,
      "epoch": 0.261136712749616,
      "grad_norm": 3.399611473083496,
      "kl": 0.2431640625,
      "learning_rate": 9.353846153846153e-07,
      "loss": 0.2012,
      "reward": 0.7063974142074585,
      "reward_std": 0.10491450130939484,
      "reward_std_all": 0.10137666389346123,
      "rewards-max/forensics_sinkhorn_reward": 0.940449595451355,
      "rewards-max/forensics_strict_edge_reward": 0.019354375544935465,
      "rewards-max/forensics_tiou_reward": 0.8721063435077667,
      "rewards/forensics_sinkhorn_reward": 0.578271359205246,
      "rewards/forensics_strict_edge_reward": 0.0012340914327069186,
      "rewards/forensics_tiou_reward": 0.2902797609567642,
      "step": 85
    },
    {
      "completion_length": 25.796875,
      "epoch": 0.2642089093701997,
      "grad_norm": 3.846431255340576,
      "kl": 0.20751953125,
      "learning_rate": 9.346153846153846e-07,
      "loss": 0.1744,
      "reward": 0.6532310545444489,
      "reward_std": 0.11432422325015068,
      "reward_std_all": 0.1282445341348648,
      "rewards-max/forensics_sinkhorn_reward": 0.8791036009788513,
      "rewards-max/forensics_strict_edge_reward": 0.32139622047543526,
      "rewards-max/forensics_tiou_reward": 0.7774899303913116,
      "rewards/forensics_sinkhorn_reward": 0.5183253884315491,
      "rewards/forensics_strict_edge_reward": 0.013570784823969007,
      "rewards/forensics_tiou_reward": 0.22017168998718262,
      "step": 86
    },
    {
      "completion_length": 22.203125,
      "epoch": 0.2672811059907834,
      "grad_norm": 4.672788619995117,
      "kl": 0.20751953125,
      "learning_rate": 9.338461538461538e-07,
      "loss": 0.2996,
      "reward": 0.6803396344184875,
      "reward_std": 0.11970315128564835,
      "reward_std_all": 0.12107149884104729,
      "rewards-max/forensics_sinkhorn_reward": 0.8355065584182739,
      "rewards-max/forensics_strict_edge_reward": 0.06334447680274025,
      "rewards-max/forensics_tiou_reward": 0.6730405986309052,
      "rewards/forensics_sinkhorn_reward": 0.5432592630386353,
      "rewards/forensics_strict_edge_reward": 0.00265413559100125,
      "rewards/forensics_tiou_reward": 0.20947065204381943,
      "step": 87
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.27035330261136714,
      "grad_norm": 4.01052188873291,
      "kl": 0.1923828125,
      "learning_rate": 9.33076923076923e-07,
      "loss": 0.1685,
      "reward": 0.6619938611984253,
      "reward_std": 0.12360595539212227,
      "reward_std_all": 0.12794580310583115,
      "rewards-max/forensics_sinkhorn_reward": 0.8984107077121735,
      "rewards-max/forensics_strict_edge_reward": 0.04597866628319025,
      "rewards-max/forensics_tiou_reward": 0.7985767722129822,
      "rewards/forensics_sinkhorn_reward": 0.5299981832504272,
      "rewards/forensics_strict_edge_reward": 0.0021890165226068348,
      "rewards/forensics_tiou_reward": 0.23819081485271454,
      "step": 88
    },
    {
      "completion_length": 24.046875,
      "epoch": 0.27342549923195086,
      "grad_norm": 4.554994106292725,
      "kl": 0.15625,
      "learning_rate": 9.323076923076922e-07,
      "loss": 0.0854,
      "reward": 0.6765183508396149,
      "reward_std": 0.09415092691779137,
      "reward_std_all": 0.11023081094026566,
      "rewards-max/forensics_sinkhorn_reward": 0.9204202592372894,
      "rewards-max/forensics_strict_edge_reward": 0.03088058321736753,
      "rewards-max/forensics_tiou_reward": 0.8170127868652344,
      "rewards/forensics_sinkhorn_reward": 0.5671239197254181,
      "rewards/forensics_strict_edge_reward": 0.0018481066217646003,
      "rewards/forensics_tiou_reward": 0.25530895590782166,
      "step": 89
    },
    {
      "completion_length": 22.265625,
      "epoch": 0.2764976958525346,
      "grad_norm": 4.124297142028809,
      "kl": 0.19970703125,
      "learning_rate": 9.315384615384615e-07,
      "loss": 0.1528,
      "reward": 0.6614440977573395,
      "reward_std": 0.11785941570997238,
      "reward_std_all": 0.1336873285472393,
      "rewards-max/forensics_sinkhorn_reward": 0.8914393186569214,
      "rewards-max/forensics_strict_edge_reward": 0.00461628899211064,
      "rewards-max/forensics_tiou_reward": 0.7287907600402832,
      "rewards/forensics_sinkhorn_reward": 0.5683229565620422,
      "rewards/forensics_strict_edge_reward": 0.0002057001220236998,
      "rewards/forensics_tiou_reward": 0.2581392526626587,
      "step": 90
    },
    {
      "completion_length": 23.5,
      "epoch": 0.27956989247311825,
      "grad_norm": 4.293308734893799,
      "kl": 0.19921875,
      "learning_rate": 9.307692307692308e-07,
      "loss": 0.1995,
      "reward": 0.6850381791591644,
      "reward_std": 0.10194173827767372,
      "reward_std_all": 0.11224841326475143,
      "rewards-max/forensics_sinkhorn_reward": 0.9375712871551514,
      "rewards-max/forensics_strict_edge_reward": 0.264460489153862,
      "rewards-max/forensics_tiou_reward": 0.8772841095924377,
      "rewards/forensics_sinkhorn_reward": 0.5769389271736145,
      "rewards/forensics_strict_edge_reward": 0.018612864310853183,
      "rewards/forensics_tiou_reward": 0.2766604647040367,
      "step": 91
    },
    {
      "completion_length": 24.78125,
      "epoch": 0.282642089093702,
      "grad_norm": 3.9535574913024902,
      "kl": 0.2119140625,
      "learning_rate": 9.3e-07,
      "loss": 0.0911,
      "reward": 0.6634361445903778,
      "reward_std": 0.09417625516653061,
      "reward_std_all": 0.11473935842514038,
      "rewards-max/forensics_sinkhorn_reward": 0.9462334513664246,
      "rewards-max/forensics_strict_edge_reward": 0.1741780280135572,
      "rewards-max/forensics_tiou_reward": 0.8641929030418396,
      "rewards/forensics_sinkhorn_reward": 0.569230318069458,
      "rewards/forensics_strict_edge_reward": 0.007839070094632916,
      "rewards/forensics_tiou_reward": 0.28998011350631714,
      "step": 92
    },
    {
      "completion_length": 25.0625,
      "epoch": 0.2857142857142857,
      "grad_norm": 3.984400749206543,
      "kl": 0.216796875,
      "learning_rate": 9.292307692307692e-07,
      "loss": 0.2244,
      "reward": 0.6838216781616211,
      "reward_std": 0.11132829636335373,
      "reward_std_all": 0.11762760207056999,
      "rewards-max/forensics_sinkhorn_reward": 0.9190828800201416,
      "rewards-max/forensics_strict_edge_reward": 0.015766605734825134,
      "rewards-max/forensics_tiou_reward": 0.8107141256332397,
      "rewards/forensics_sinkhorn_reward": 0.5826047658920288,
      "rewards/forensics_strict_edge_reward": 0.0010194483038503677,
      "rewards/forensics_tiou_reward": 0.2959672510623932,
      "step": 93
    },
    {
      "completion_length": 23.53125,
      "epoch": 0.2887864823348694,
      "grad_norm": 3.9316580295562744,
      "kl": 0.2158203125,
      "learning_rate": 9.284615384615384e-07,
      "loss": 0.1955,
      "reward": 0.670534610748291,
      "reward_std": 0.14107603207230568,
      "reward_std_all": 0.1360323615372181,
      "rewards-max/forensics_sinkhorn_reward": 0.9742431938648224,
      "rewards-max/forensics_strict_edge_reward": 0.3249748758971691,
      "rewards-max/forensics_tiou_reward": 0.9603675603866577,
      "rewards/forensics_sinkhorn_reward": 0.548993319272995,
      "rewards/forensics_strict_edge_reward": 0.014410440810024738,
      "rewards/forensics_tiou_reward": 0.3086860477924347,
      "step": 94
    },
    {
      "completion_length": 25.390625,
      "epoch": 0.29185867895545314,
      "grad_norm": 4.501471519470215,
      "kl": 0.2822265625,
      "learning_rate": 9.276923076923076e-07,
      "loss": 0.2119,
      "reward": 0.7071852087974548,
      "reward_std": 0.09689635783433914,
      "reward_std_all": 0.09579430148005486,
      "rewards-max/forensics_sinkhorn_reward": 0.8788932859897614,
      "rewards-max/forensics_strict_edge_reward": 0.07937991991639137,
      "rewards-max/forensics_tiou_reward": 0.7060965597629547,
      "rewards/forensics_sinkhorn_reward": 0.5960882604122162,
      "rewards/forensics_strict_edge_reward": 0.0057503073767293245,
      "rewards/forensics_tiou_reward": 0.2898605167865753,
      "step": 95
    },
    {
      "completion_length": 27.265625,
      "epoch": 0.29493087557603687,
      "grad_norm": 5.83609676361084,
      "kl": 0.21435546875,
      "learning_rate": 9.26923076923077e-07,
      "loss": 0.1493,
      "reward": 0.6938998401165009,
      "reward_std": 0.10180044174194336,
      "reward_std_all": 0.10300350189208984,
      "rewards-max/forensics_sinkhorn_reward": 0.9098639488220215,
      "rewards-max/forensics_strict_edge_reward": 0.006827035686001182,
      "rewards-max/forensics_tiou_reward": 0.8130742609500885,
      "rewards/forensics_sinkhorn_reward": 0.5494679510593414,
      "rewards/forensics_strict_edge_reward": 0.0004002096684416756,
      "rewards/forensics_tiou_reward": 0.3208559453487396,
      "step": 96
    },
    {
      "completion_length": 25.328125,
      "epoch": 0.2980030721966206,
      "grad_norm": 4.093597412109375,
      "kl": 0.17919921875,
      "learning_rate": 9.261538461538461e-07,
      "loss": 0.1841,
      "reward": 0.651899516582489,
      "reward_std": 0.1387939676642418,
      "reward_std_all": 0.14182935655117035,
      "rewards-max/forensics_sinkhorn_reward": 0.8175784051418304,
      "rewards-max/forensics_strict_edge_reward": 0.013778270280454308,
      "rewards-max/forensics_tiou_reward": 0.607740968465805,
      "rewards/forensics_sinkhorn_reward": 0.5032462775707245,
      "rewards/forensics_strict_edge_reward": 0.0007452611280314159,
      "rewards/forensics_tiou_reward": 0.23947419971227646,
      "step": 97
    },
    {
      "completion_length": 27.453125,
      "epoch": 0.3010752688172043,
      "grad_norm": 3.555084705352783,
      "kl": 0.21435546875,
      "learning_rate": 9.253846153846153e-07,
      "loss": 0.0318,
      "reward": 0.6994836032390594,
      "reward_std": 0.11905693262815475,
      "reward_std_all": 0.11393306031823158,
      "rewards-max/forensics_sinkhorn_reward": 0.9380309581756592,
      "rewards-max/forensics_strict_edge_reward": 0.19819949753582478,
      "rewards-max/forensics_tiou_reward": 0.8573337495326996,
      "rewards/forensics_sinkhorn_reward": 0.5504316091537476,
      "rewards/forensics_strict_edge_reward": 0.008441511890850961,
      "rewards/forensics_tiou_reward": 0.33485326170921326,
      "step": 98
    },
    {
      "completion_length": 27.40625,
      "epoch": 0.30414746543778803,
      "grad_norm": 5.300096035003662,
      "kl": 0.17529296875,
      "learning_rate": 9.246153846153846e-07,
      "loss": 0.125,
      "reward": 0.6781783998012543,
      "reward_std": 0.11172496899962425,
      "reward_std_all": 0.11712309718132019,
      "rewards-max/forensics_sinkhorn_reward": 0.8314941227436066,
      "rewards-max/forensics_strict_edge_reward": 0.1995817037532106,
      "rewards-max/forensics_tiou_reward": 0.6327329576015472,
      "rewards/forensics_sinkhorn_reward": 0.546312689781189,
      "rewards/forensics_strict_edge_reward": 0.011559725004190113,
      "rewards/forensics_tiou_reward": 0.2671235054731369,
      "step": 99
    },
    {
      "completion_length": 22.453125,
      "epoch": 0.30721966205837176,
      "grad_norm": 4.973781585693359,
      "kl": 0.18798828125,
      "learning_rate": 9.238461538461538e-07,
      "loss": 0.1026,
      "reward": 0.6678672730922699,
      "reward_std": 0.10669650882482529,
      "reward_std_all": 0.12280864268541336,
      "rewards-max/forensics_sinkhorn_reward": 0.9535457491874695,
      "rewards-max/forensics_strict_edge_reward": 0.20613250508904457,
      "rewards-max/forensics_tiou_reward": 0.9056941270828247,
      "rewards/forensics_sinkhorn_reward": 0.5728576481342316,
      "rewards/forensics_strict_edge_reward": 0.011641261633485556,
      "rewards/forensics_tiou_reward": 0.3179563581943512,
      "step": 100
    },
    {
      "completion_length": 23.921875,
      "epoch": 0.3102918586789555,
      "grad_norm": 3.989253520965576,
      "kl": 0.17431640625,
      "learning_rate": 9.230769230769231e-07,
      "loss": 0.1531,
      "reward": 0.652406245470047,
      "reward_std": 0.1499311476945877,
      "reward_std_all": 0.15242073684930801,
      "rewards-max/forensics_sinkhorn_reward": 0.9590258002281189,
      "rewards-max/forensics_strict_edge_reward": 0.3159602954983711,
      "rewards-max/forensics_tiou_reward": 0.9243376851081848,
      "rewards/forensics_sinkhorn_reward": 0.5290761590003967,
      "rewards/forensics_strict_edge_reward": 0.020611689426004887,
      "rewards/forensics_tiou_reward": 0.294885978102684,
      "step": 101
    },
    {
      "completion_length": 22.484375,
      "epoch": 0.31336405529953915,
      "grad_norm": 4.317831039428711,
      "kl": 0.2138671875,
      "learning_rate": 9.223076923076922e-07,
      "loss": 0.1305,
      "reward": 0.6250726282596588,
      "reward_std": 0.15872742235660553,
      "reward_std_all": 0.16814492642879486,
      "rewards-max/forensics_sinkhorn_reward": 0.9429434537887573,
      "rewards-max/forensics_strict_edge_reward": 0.08301437273621559,
      "rewards-max/forensics_tiou_reward": 0.8649616539478302,
      "rewards/forensics_sinkhorn_reward": 0.535449281334877,
      "rewards/forensics_strict_edge_reward": 0.006613269913941622,
      "rewards/forensics_tiou_reward": 0.2742621749639511,
      "step": 102
    },
    {
      "completion_length": 23.46875,
      "epoch": 0.31643625192012287,
      "grad_norm": 4.635815143585205,
      "kl": 0.173828125,
      "learning_rate": 9.215384615384614e-07,
      "loss": 0.0299,
      "reward": 0.7183987498283386,
      "reward_std": 0.08224273100495338,
      "reward_std_all": 0.08242032304406166,
      "rewards-max/forensics_sinkhorn_reward": 0.9744893312454224,
      "rewards-max/forensics_strict_edge_reward": 0.40866968780755997,
      "rewards-max/forensics_tiou_reward": 0.9585742354393005,
      "rewards/forensics_sinkhorn_reward": 0.6022777557373047,
      "rewards/forensics_strict_edge_reward": 0.02483038231730461,
      "rewards/forensics_tiou_reward": 0.34763891994953156,
      "step": 103
    },
    {
      "completion_length": 22.109375,
      "epoch": 0.3195084485407066,
      "grad_norm": 4.752939701080322,
      "kl": 0.29638671875,
      "learning_rate": 9.207692307692308e-07,
      "loss": 0.1517,
      "reward": 0.7082028388977051,
      "reward_std": 0.09048548340797424,
      "reward_std_all": 0.09688258916139603,
      "rewards-max/forensics_sinkhorn_reward": 0.9067874848842621,
      "rewards-max/forensics_strict_edge_reward": 0.28529953956604004,
      "rewards-max/forensics_tiou_reward": 0.8040073812007904,
      "rewards/forensics_sinkhorn_reward": 0.6390056312084198,
      "rewards/forensics_strict_edge_reward": 0.020061166491359472,
      "rewards/forensics_tiou_reward": 0.3536484092473984,
      "step": 104
    },
    {
      "completion_length": 22.359375,
      "epoch": 0.3225806451612903,
      "grad_norm": 3.97251296043396,
      "kl": 0.2255859375,
      "learning_rate": 9.2e-07,
      "loss": 0.1757,
      "reward": 0.6573615372180939,
      "reward_std": 0.1255042552947998,
      "reward_std_all": 0.13156210631132126,
      "rewards-max/forensics_sinkhorn_reward": 0.9517093598842621,
      "rewards-max/forensics_strict_edge_reward": 0.19672049209475517,
      "rewards-max/forensics_tiou_reward": 0.9020406603813171,
      "rewards/forensics_sinkhorn_reward": 0.5456351041793823,
      "rewards/forensics_strict_edge_reward": 0.008234355133026838,
      "rewards/forensics_tiou_reward": 0.2701706886291504,
      "step": 105
    },
    {
      "completion_length": 23.09375,
      "epoch": 0.32565284178187404,
      "grad_norm": 5.109591484069824,
      "kl": 0.20263671875,
      "learning_rate": 9.192307692307692e-07,
      "loss": -0.0085,
      "reward": 0.7083730697631836,
      "reward_std": 0.0849413312971592,
      "reward_std_all": 0.08790911734104156,
      "rewards-max/forensics_sinkhorn_reward": 0.9573144316673279,
      "rewards-max/forensics_strict_edge_reward": 0.16134114749729633,
      "rewards-max/forensics_tiou_reward": 0.9106801748275757,
      "rewards/forensics_sinkhorn_reward": 0.6012448370456696,
      "rewards/forensics_strict_edge_reward": 0.015887623230810277,
      "rewards/forensics_tiou_reward": 0.28564705699682236,
      "step": 106
    },
    {
      "completion_length": 22.4375,
      "epoch": 0.32872503840245776,
      "grad_norm": 5.239882946014404,
      "kl": 0.20263671875,
      "learning_rate": 9.184615384615384e-07,
      "loss": 0.1739,
      "reward": 0.7257605493068695,
      "reward_std": 0.08192195743322372,
      "reward_std_all": 0.08356864005327225,
      "rewards-max/forensics_sinkhorn_reward": 0.9565331935882568,
      "rewards-max/forensics_strict_edge_reward": 0.25740488804876804,
      "rewards-max/forensics_tiou_reward": 0.9141034483909607,
      "rewards/forensics_sinkhorn_reward": 0.6323734521865845,
      "rewards/forensics_strict_edge_reward": 0.014003062504343688,
      "rewards/forensics_tiou_reward": 0.34302452206611633,
      "step": 107
    },
    {
      "completion_length": 23.46875,
      "epoch": 0.3317972350230415,
      "grad_norm": 3.892258644104004,
      "kl": 0.1845703125,
      "learning_rate": 9.176923076923076e-07,
      "loss": 0.2185,
      "reward": 0.6968185901641846,
      "reward_std": 0.07683944329619408,
      "reward_std_all": 0.09222147986292839,
      "rewards-max/forensics_sinkhorn_reward": 0.9690417945384979,
      "rewards-max/forensics_strict_edge_reward": 0.39595161124907463,
      "rewards-max/forensics_tiou_reward": 0.9518364667892456,
      "rewards/forensics_sinkhorn_reward": 0.5995332896709442,
      "rewards/forensics_strict_edge_reward": 0.03450260467707267,
      "rewards/forensics_tiou_reward": 0.3088626265525818,
      "step": 108
    },
    {
      "completion_length": 21.078125,
      "epoch": 0.3348694316436252,
      "grad_norm": 4.788924217224121,
      "kl": 0.1953125,
      "learning_rate": 9.169230769230769e-07,
      "loss": 0.1445,
      "reward": 0.7281473278999329,
      "reward_std": 0.08571437001228333,
      "reward_std_all": 0.08271703869104385,
      "rewards-max/forensics_sinkhorn_reward": 0.9787065088748932,
      "rewards-max/forensics_strict_edge_reward": 0.5320244655013084,
      "rewards-max/forensics_tiou_reward": 0.9625748693943024,
      "rewards/forensics_sinkhorn_reward": 0.6410692632198334,
      "rewards/forensics_strict_edge_reward": 0.039628634229302406,
      "rewards/forensics_tiou_reward": 0.3614591807126999,
      "step": 109
    },
    {
      "completion_length": 23.84375,
      "epoch": 0.3379416282642089,
      "grad_norm": 3.888638734817505,
      "kl": 0.22314453125,
      "learning_rate": 9.161538461538462e-07,
      "loss": 0.0704,
      "reward": 0.682310163974762,
      "reward_std": 0.10524578392505646,
      "reward_std_all": 0.1094505563378334,
      "rewards-max/forensics_sinkhorn_reward": 0.9326315820217133,
      "rewards-max/forensics_strict_edge_reward": 0.0390629917383194,
      "rewards-max/forensics_tiou_reward": 0.8384874165058136,
      "rewards/forensics_sinkhorn_reward": 0.5769358575344086,
      "rewards/forensics_strict_edge_reward": 0.004202864307444543,
      "rewards/forensics_tiou_reward": 0.3246810883283615,
      "step": 110
    },
    {
      "completion_length": 23.703125,
      "epoch": 0.34101382488479265,
      "grad_norm": 4.4391069412231445,
      "kl": 0.23974609375,
      "learning_rate": 9.153846153846153e-07,
      "loss": 0.1828,
      "reward": 0.6633279621601105,
      "reward_std": 0.09195753559470177,
      "reward_std_all": 0.11326979473233223,
      "rewards-max/forensics_sinkhorn_reward": 0.9734886884689331,
      "rewards-max/forensics_strict_edge_reward": 0.3846648633480072,
      "rewards-max/forensics_tiou_reward": 0.9533218741416931,
      "rewards/forensics_sinkhorn_reward": 0.5479127466678619,
      "rewards/forensics_strict_edge_reward": 0.026478039100766182,
      "rewards/forensics_tiou_reward": 0.26579688489437103,
      "step": 111
    },
    {
      "completion_length": 21.546875,
      "epoch": 0.34408602150537637,
      "grad_norm": 6.036125659942627,
      "kl": 0.29736328125,
      "learning_rate": 9.146153846153846e-07,
      "loss": -0.0863,
      "reward": 0.6965771615505219,
      "reward_std": 0.07296135276556015,
      "reward_std_all": 0.0889790691435337,
      "rewards-max/forensics_sinkhorn_reward": 0.9239678680896759,
      "rewards-max/forensics_strict_edge_reward": 0.177419763058424,
      "rewards-max/forensics_tiou_reward": 0.8199724555015564,
      "rewards/forensics_sinkhorn_reward": 0.5717905461788177,
      "rewards/forensics_strict_edge_reward": 0.007800784311257303,
      "rewards/forensics_tiou_reward": 0.2513194680213928,
      "step": 112
    },
    {
      "completion_length": 21.578125,
      "epoch": 0.34715821812596004,
      "grad_norm": 4.194161891937256,
      "kl": 0.23828125,
      "learning_rate": 9.138461538461538e-07,
      "loss": 0.1454,
      "reward": 0.686409056186676,
      "reward_std": 0.1144479401409626,
      "reward_std_all": 0.11939387023448944,
      "rewards-max/forensics_sinkhorn_reward": 0.9269570708274841,
      "rewards-max/forensics_strict_edge_reward": 0.03247493240633048,
      "rewards-max/forensics_tiou_reward": 0.8495339751243591,
      "rewards/forensics_sinkhorn_reward": 0.5870908200740814,
      "rewards/forensics_strict_edge_reward": 0.0013770911673418595,
      "rewards/forensics_tiou_reward": 0.29195065796375275,
      "step": 113
    },
    {
      "completion_length": 23.296875,
      "epoch": 0.35023041474654376,
      "grad_norm": 5.4491777420043945,
      "kl": 0.30859375,
      "learning_rate": 9.130769230769231e-07,
      "loss": 0.2358,
      "reward": 0.6268754303455353,
      "reward_std": 0.13947048783302307,
      "reward_std_all": 0.15301287919282913,
      "rewards-max/forensics_sinkhorn_reward": 0.9151811599731445,
      "rewards-max/forensics_strict_edge_reward": 0.09687153506092727,
      "rewards-max/forensics_tiou_reward": 0.8123402297496796,
      "rewards/forensics_sinkhorn_reward": 0.5112425684928894,
      "rewards/forensics_strict_edge_reward": 0.005032276822021231,
      "rewards/forensics_tiou_reward": 0.23001950234174728,
      "step": 114
    },
    {
      "completion_length": 22.640625,
      "epoch": 0.3533026113671275,
      "grad_norm": 4.211780071258545,
      "kl": 0.22509765625,
      "learning_rate": 9.123076923076922e-07,
      "loss": 0.1656,
      "reward": 0.6635256409645081,
      "reward_std": 0.07484827935695648,
      "reward_std_all": 0.10386835038661957,
      "rewards-max/forensics_sinkhorn_reward": 0.8850461542606354,
      "rewards-max/forensics_strict_edge_reward": 0.2933268314413908,
      "rewards-max/forensics_tiou_reward": 0.7838974595069885,
      "rewards/forensics_sinkhorn_reward": 0.5624166429042816,
      "rewards/forensics_strict_edge_reward": 0.012651888720796478,
      "rewards/forensics_tiou_reward": 0.26331179589033127,
      "step": 115
    },
    {
      "completion_length": 24.6875,
      "epoch": 0.3563748079877112,
      "grad_norm": 4.107433319091797,
      "kl": 0.24072265625,
      "learning_rate": 9.115384615384614e-07,
      "loss": 0.071,
      "reward": 0.6237091422080994,
      "reward_std": 0.1316344141960144,
      "reward_std_all": 0.1505041867494583,
      "rewards-max/forensics_sinkhorn_reward": 0.9689725339412689,
      "rewards-max/forensics_strict_edge_reward": 0.11172731569968164,
      "rewards-max/forensics_tiou_reward": 0.9450499713420868,
      "rewards/forensics_sinkhorn_reward": 0.5015500634908676,
      "rewards/forensics_strict_edge_reward": 0.004659419252675434,
      "rewards/forensics_tiou_reward": 0.2273254245519638,
      "step": 116
    },
    {
      "completion_length": 24.0625,
      "epoch": 0.35944700460829493,
      "grad_norm": 4.962100028991699,
      "kl": 0.23486328125,
      "learning_rate": 9.107692307692308e-07,
      "loss": 0.098,
      "reward": 0.7036144137382507,
      "reward_std": 0.07453915104269981,
      "reward_std_all": 0.08679182082414627,
      "rewards-max/forensics_sinkhorn_reward": 0.9006543457508087,
      "rewards-max/forensics_strict_edge_reward": 0.10308091901242733,
      "rewards-max/forensics_tiou_reward": 0.7766641080379486,
      "rewards/forensics_sinkhorn_reward": 0.5939966440200806,
      "rewards/forensics_strict_edge_reward": 0.004794749605935067,
      "rewards/forensics_tiou_reward": 0.3016921728849411,
      "step": 117
    },
    {
      "completion_length": 22.546875,
      "epoch": 0.36251920122887865,
      "grad_norm": 5.092795372009277,
      "kl": 0.25439453125,
      "learning_rate": 9.1e-07,
      "loss": 0.1933,
      "reward": 0.7114239037036896,
      "reward_std": 0.11220360919833183,
      "reward_std_all": 0.10645899176597595,
      "rewards-max/forensics_sinkhorn_reward": 0.9427308440208435,
      "rewards-max/forensics_strict_edge_reward": 0.19366364739835262,
      "rewards-max/forensics_tiou_reward": 0.8576912581920624,
      "rewards/forensics_sinkhorn_reward": 0.6161253750324249,
      "rewards/forensics_strict_edge_reward": 0.013935673981904984,
      "rewards/forensics_tiou_reward": 0.32266204059123993,
      "step": 118
    },
    {
      "completion_length": 29.875,
      "epoch": 0.3655913978494624,
      "grad_norm": 3.601071834564209,
      "kl": 0.23681640625,
      "learning_rate": 9.092307692307692e-07,
      "loss": 0.1378,
      "reward": 0.6409783363342285,
      "reward_std": 0.0847595315426588,
      "reward_std_all": 0.11867161467671394,
      "rewards-max/forensics_sinkhorn_reward": 0.9269739091396332,
      "rewards-max/forensics_strict_edge_reward": 0.2578680074075237,
      "rewards-max/forensics_tiou_reward": 0.8359212875366211,
      "rewards/forensics_sinkhorn_reward": 0.5152475237846375,
      "rewards/forensics_strict_edge_reward": 0.011469966673757881,
      "rewards/forensics_tiou_reward": 0.2608402296900749,
      "step": 119
    },
    {
      "completion_length": 24.4375,
      "epoch": 0.3686635944700461,
      "grad_norm": 4.326385498046875,
      "kl": 0.24609375,
      "learning_rate": 9.084615384615384e-07,
      "loss": -0.0272,
      "reward": 0.6934029459953308,
      "reward_std": 0.11264162510633469,
      "reward_std_all": 0.11123588308691978,
      "rewards-max/forensics_sinkhorn_reward": 0.8994802236557007,
      "rewards-max/forensics_strict_edge_reward": 0.11540399894511211,
      "rewards-max/forensics_tiou_reward": 0.7862301468849182,
      "rewards/forensics_sinkhorn_reward": 0.5614398121833801,
      "rewards/forensics_strict_edge_reward": 0.0056555419266715035,
      "rewards/forensics_tiou_reward": 0.30814455449581146,
      "step": 120
    },
    {
      "completion_length": 22.28125,
      "epoch": 0.3717357910906298,
      "grad_norm": 5.080354690551758,
      "kl": 0.23828125,
      "learning_rate": 9.076923076923076e-07,
      "loss": 0.1839,
      "reward": 0.717615157365799,
      "reward_std": 0.08568545803427696,
      "reward_std_all": 0.0849427692592144,
      "rewards-max/forensics_sinkhorn_reward": 0.8656897842884064,
      "rewards-max/forensics_strict_edge_reward": 0.04672353807836771,
      "rewards-max/forensics_tiou_reward": 0.72898069024086,
      "rewards/forensics_sinkhorn_reward": 0.6242617666721344,
      "rewards/forensics_strict_edge_reward": 0.0023126022424548864,
      "rewards/forensics_tiou_reward": 0.285324864089489,
      "step": 121
    },
    {
      "completion_length": 23.875,
      "epoch": 0.37480798771121354,
      "grad_norm": 5.101120471954346,
      "kl": 0.279296875,
      "learning_rate": 9.069230769230769e-07,
      "loss": 0.1382,
      "reward": 0.6541453003883362,
      "reward_std": 0.11235585808753967,
      "reward_std_all": 0.13306356966495514,
      "rewards-max/forensics_sinkhorn_reward": 0.9697747528553009,
      "rewards-max/forensics_strict_edge_reward": 0.29456247575581074,
      "rewards-max/forensics_tiou_reward": 0.9466418921947479,
      "rewards/forensics_sinkhorn_reward": 0.5258007496595383,
      "rewards/forensics_strict_edge_reward": 0.020241256701410748,
      "rewards/forensics_tiou_reward": 0.3105051666498184,
      "step": 122
    },
    {
      "completion_length": 23.1875,
      "epoch": 0.3778801843317972,
      "grad_norm": 4.576303958892822,
      "kl": 0.2177734375,
      "learning_rate": 9.061538461538461e-07,
      "loss": 0.1519,
      "reward": 0.6545627117156982,
      "reward_std": 0.1440073475241661,
      "reward_std_all": 0.14789243042469025,
      "rewards-max/forensics_sinkhorn_reward": 0.9125995635986328,
      "rewards-max/forensics_strict_edge_reward": 0.14022861677221954,
      "rewards-max/forensics_tiou_reward": 0.8054374754428864,
      "rewards/forensics_sinkhorn_reward": 0.528709888458252,
      "rewards/forensics_strict_edge_reward": 0.013435382927127648,
      "rewards/forensics_tiou_reward": 0.29799818247556686,
      "step": 123
    },
    {
      "completion_length": 23.6875,
      "epoch": 0.38095238095238093,
      "grad_norm": 5.415949821472168,
      "kl": 0.271484375,
      "learning_rate": 9.053846153846153e-07,
      "loss": 0.0296,
      "reward": 0.6811282336711884,
      "reward_std": 0.09330116212368011,
      "reward_std_all": 0.1069151982665062,
      "rewards-max/forensics_sinkhorn_reward": 0.913462221622467,
      "rewards-max/forensics_strict_edge_reward": 0.28051102021709085,
      "rewards-max/forensics_tiou_reward": 0.8236248195171356,
      "rewards/forensics_sinkhorn_reward": 0.5665398240089417,
      "rewards/forensics_strict_edge_reward": 0.01203817268833518,
      "rewards/forensics_tiou_reward": 0.30067354440689087,
      "step": 124
    },
    {
      "completion_length": 25.09375,
      "epoch": 0.38402457757296465,
      "grad_norm": 6.224219799041748,
      "kl": 0.24755859375,
      "learning_rate": 9.046153846153846e-07,
      "loss": 0.1772,
      "reward": 0.6482094824314117,
      "reward_std": 0.13580892980098724,
      "reward_std_all": 0.1413438431918621,
      "rewards-max/forensics_sinkhorn_reward": 0.8786899745464325,
      "rewards-max/forensics_strict_edge_reward": 0.059675346419680864,
      "rewards-max/forensics_tiou_reward": 0.7217077016830444,
      "rewards/forensics_sinkhorn_reward": 0.5013777613639832,
      "rewards/forensics_strict_edge_reward": 0.002531174090108834,
      "rewards/forensics_tiou_reward": 0.2575315460562706,
      "step": 125
    },
    {
      "completion_length": 27.34375,
      "epoch": 0.3870967741935484,
      "grad_norm": 4.329051494598389,
      "kl": 0.26318359375,
      "learning_rate": 9.038461538461538e-07,
      "loss": 0.2258,
      "reward": 0.6536787450313568,
      "reward_std": 0.11023084446787834,
      "reward_std_all": 0.12622301280498505,
      "rewards-max/forensics_sinkhorn_reward": 0.9198334217071533,
      "rewards-max/forensics_strict_edge_reward": 0.07003513164818287,
      "rewards-max/forensics_tiou_reward": 0.8234404027462006,
      "rewards/forensics_sinkhorn_reward": 0.5111680626869202,
      "rewards/forensics_strict_edge_reward": 0.0031394254183396697,
      "rewards/forensics_tiou_reward": 0.3043002784252167,
      "step": 126
    },
    {
      "completion_length": 25.125,
      "epoch": 0.3901689708141321,
      "grad_norm": 3.9919703006744385,
      "kl": 0.2099609375,
      "learning_rate": 9.030769230769231e-07,
      "loss": 0.2299,
      "reward": 0.6957383751869202,
      "reward_std": 0.0953773595392704,
      "reward_std_all": 0.10314103215932846,
      "rewards-max/forensics_sinkhorn_reward": 0.9133339524269104,
      "rewards-max/forensics_strict_edge_reward": 0.14381973072886467,
      "rewards-max/forensics_tiou_reward": 0.8122845888137817,
      "rewards/forensics_sinkhorn_reward": 0.5837920606136322,
      "rewards/forensics_strict_edge_reward": 0.009784935973584652,
      "rewards/forensics_tiou_reward": 0.3133707791566849,
      "step": 127
    },
    {
      "completion_length": 26.546875,
      "epoch": 0.3932411674347158,
      "grad_norm": 3.7950279712677,
      "kl": 0.24853515625,
      "learning_rate": 9.023076923076922e-07,
      "loss": 0.1177,
      "reward": 0.7107680439949036,
      "reward_std": 0.10332817956805229,
      "reward_std_all": 0.10003652796149254,
      "rewards-max/forensics_sinkhorn_reward": 0.9551940858364105,
      "rewards-max/forensics_strict_edge_reward": 0.20139022218063474,
      "rewards-max/forensics_tiou_reward": 0.8971261978149414,
      "rewards/forensics_sinkhorn_reward": 0.5883759260177612,
      "rewards/forensics_strict_edge_reward": 0.010282587696565315,
      "rewards/forensics_tiou_reward": 0.3592185527086258,
      "step": 128
    },
    {
      "completion_length": 25.25,
      "epoch": 0.39631336405529954,
      "grad_norm": 4.80145263671875,
      "kl": 0.2138671875,
      "learning_rate": 9.015384615384614e-07,
      "loss": 0.1548,
      "reward": 0.6269001662731171,
      "reward_std": 0.17354001104831696,
      "reward_std_all": 0.17094184458255768,
      "rewards-max/forensics_sinkhorn_reward": 0.898750364780426,
      "rewards-max/forensics_strict_edge_reward": 0.11759518086910248,
      "rewards-max/forensics_tiou_reward": 0.783599466085434,
      "rewards/forensics_sinkhorn_reward": 0.47995007038116455,
      "rewards/forensics_strict_edge_reward": 0.007380595430731773,
      "rewards/forensics_tiou_reward": 0.25659196078777313,
      "step": 129
    },
    {
      "completion_length": 21.65625,
      "epoch": 0.39938556067588327,
      "grad_norm": 4.892286777496338,
      "kl": 0.2724609375,
      "learning_rate": 9.007692307692307e-07,
      "loss": 0.0429,
      "reward": 0.6742261350154877,
      "reward_std": 0.07751192152500153,
      "reward_std_all": 0.10112189501523972,
      "rewards-max/forensics_sinkhorn_reward": 0.8217279016971588,
      "rewards-max/forensics_strict_edge_reward": 0.18719066493213177,
      "rewards-max/forensics_tiou_reward": 0.7088338434696198,
      "rewards/forensics_sinkhorn_reward": 0.5549533069133759,
      "rewards/forensics_strict_edge_reward": 0.010550589940976351,
      "rewards/forensics_tiou_reward": 0.24002937972545624,
      "step": 130
    },
    {
      "completion_length": 24.328125,
      "epoch": 0.402457757296467,
      "grad_norm": 4.276612758636475,
      "kl": 0.2275390625,
      "learning_rate": 9e-07,
      "loss": 0.1886,
      "reward": 0.7034057974815369,
      "reward_std": 0.10461972653865814,
      "reward_std_all": 0.10183832421898842,
      "rewards-max/forensics_sinkhorn_reward": 0.9824081361293793,
      "rewards-max/forensics_strict_edge_reward": 0.38799008168280125,
      "rewards-max/forensics_tiou_reward": 0.9768793880939484,
      "rewards/forensics_sinkhorn_reward": 0.627097100019455,
      "rewards/forensics_strict_edge_reward": 0.016549982887227088,
      "rewards/forensics_tiou_reward": 0.36035096645355225,
      "step": 131
    },
    {
      "completion_length": 23.609375,
      "epoch": 0.4055299539170507,
      "grad_norm": 5.792685508728027,
      "kl": 0.22265625,
      "learning_rate": 8.992307692307693e-07,
      "loss": 0.0671,
      "reward": 0.6770969331264496,
      "reward_std": 0.1454283893108368,
      "reward_std_all": 0.1387171447277069,
      "rewards-max/forensics_sinkhorn_reward": 0.8779067099094391,
      "rewards-max/forensics_strict_edge_reward": 0.0033945991308428347,
      "rewards-max/forensics_tiou_reward": 0.7297439575195312,
      "rewards/forensics_sinkhorn_reward": 0.5374642312526703,
      "rewards/forensics_strict_edge_reward": 0.00018132830882677808,
      "rewards/forensics_tiou_reward": 0.24161741137504578,
      "step": 132
    },
    {
      "completion_length": 26.40625,
      "epoch": 0.40860215053763443,
      "grad_norm": 4.126456260681152,
      "kl": 0.2529296875,
      "learning_rate": 8.984615384615384e-07,
      "loss": 0.1127,
      "reward": 0.6608079373836517,
      "reward_std": 0.11241042613983154,
      "reward_std_all": 0.12303540855646133,
      "rewards-max/forensics_sinkhorn_reward": 0.9662142395973206,
      "rewards-max/forensics_strict_edge_reward": 0.28095155395567417,
      "rewards-max/forensics_tiou_reward": 0.9383784532546997,
      "rewards/forensics_sinkhorn_reward": 0.5461407899856567,
      "rewards/forensics_strict_edge_reward": 0.012903545750305057,
      "rewards/forensics_tiou_reward": 0.3197692036628723,
      "step": 133
    },
    {
      "completion_length": 23.953125,
      "epoch": 0.4116743471582181,
      "grad_norm": 4.5338521003723145,
      "kl": 0.265625,
      "learning_rate": 8.976923076923076e-07,
      "loss": 0.1448,
      "reward": 0.6681766211986542,
      "reward_std": 0.10188189893960953,
      "reward_std_all": 0.11448444426059723,
      "rewards-max/forensics_sinkhorn_reward": 0.9464017748832703,
      "rewards-max/forensics_strict_edge_reward": 0.2462407574057579,
      "rewards-max/forensics_tiou_reward": 0.8873580098152161,
      "rewards/forensics_sinkhorn_reward": 0.5522295832633972,
      "rewards/forensics_strict_edge_reward": 0.012125790119171143,
      "rewards/forensics_tiou_reward": 0.27776169776916504,
      "step": 134
    },
    {
      "completion_length": 22.5,
      "epoch": 0.4147465437788018,
      "grad_norm": 5.732400417327881,
      "kl": 0.22216796875,
      "learning_rate": 8.969230769230769e-07,
      "loss": 0.2473,
      "reward": 0.6449492573738098,
      "reward_std": 0.14012929797172546,
      "reward_std_all": 0.14882763475179672,
      "rewards-max/forensics_sinkhorn_reward": 0.8807512819766998,
      "rewards-max/forensics_strict_edge_reward": 0.17363484343513846,
      "rewards-max/forensics_tiou_reward": 0.8048329055309296,
      "rewards/forensics_sinkhorn_reward": 0.5376958250999451,
      "rewards/forensics_strict_edge_reward": 0.007242896215757355,
      "rewards/forensics_tiou_reward": 0.23037414997816086,
      "step": 135
    },
    {
      "completion_length": 22.59375,
      "epoch": 0.41781874039938555,
      "grad_norm": 4.349639892578125,
      "kl": 0.25830078125,
      "learning_rate": 8.961538461538461e-07,
      "loss": 0.1704,
      "reward": 0.6987101435661316,
      "reward_std": 0.0965159647166729,
      "reward_std_all": 0.09884136915206909,
      "rewards-max/forensics_sinkhorn_reward": 0.8948231637477875,
      "rewards-max/forensics_strict_edge_reward": 0.2039441093802452,
      "rewards-max/forensics_tiou_reward": 0.7879446744918823,
      "rewards/forensics_sinkhorn_reward": 0.6133834719657898,
      "rewards/forensics_strict_edge_reward": 0.01271396828815341,
      "rewards/forensics_tiou_reward": 0.2923775613307953,
      "step": 136
    },
    {
      "completion_length": 22.609375,
      "epoch": 0.42089093701996927,
      "grad_norm": 5.800956726074219,
      "kl": 0.22314453125,
      "learning_rate": 8.953846153846152e-07,
      "loss": 0.2096,
      "reward": 0.7022919058799744,
      "reward_std": 0.11120996624231339,
      "reward_std_all": 0.10657554119825363,
      "rewards-max/forensics_sinkhorn_reward": 0.954442024230957,
      "rewards-max/forensics_strict_edge_reward": 0.24453946948051453,
      "rewards-max/forensics_tiou_reward": 0.912382036447525,
      "rewards/forensics_sinkhorn_reward": 0.5677136480808258,
      "rewards/forensics_strict_edge_reward": 0.017409736290574074,
      "rewards/forensics_tiou_reward": 0.30827394127845764,
      "step": 137
    },
    {
      "completion_length": 23.921875,
      "epoch": 0.423963133640553,
      "grad_norm": 3.694828510284424,
      "kl": 0.2431640625,
      "learning_rate": 8.946153846153846e-07,
      "loss": 0.1718,
      "reward": 0.7240963578224182,
      "reward_std": 0.10568437725305557,
      "reward_std_all": 0.09795583412051201,
      "rewards-max/forensics_sinkhorn_reward": 0.9805586338043213,
      "rewards-max/forensics_strict_edge_reward": 0.34085880406200886,
      "rewards-max/forensics_tiou_reward": 0.9745890498161316,
      "rewards/forensics_sinkhorn_reward": 0.6285966336727142,
      "rewards/forensics_strict_edge_reward": 0.014998781611211598,
      "rewards/forensics_tiou_reward": 0.39867742359638214,
      "step": 138
    },
    {
      "completion_length": 21.8125,
      "epoch": 0.4270353302611367,
      "grad_norm": 4.471842288970947,
      "kl": 0.177734375,
      "learning_rate": 8.938461538461538e-07,
      "loss": 0.1932,
      "reward": 0.6942915320396423,
      "reward_std": 0.11817297711968422,
      "reward_std_all": 0.11672546714544296,
      "rewards-max/forensics_sinkhorn_reward": 0.9258284568786621,
      "rewards-max/forensics_strict_edge_reward": 0.049147823825478554,
      "rewards-max/forensics_tiou_reward": 0.8390234112739563,
      "rewards/forensics_sinkhorn_reward": 0.5882952511310577,
      "rewards/forensics_strict_edge_reward": 0.0032932902686297894,
      "rewards/forensics_tiou_reward": 0.32506997883319855,
      "step": 139
    },
    {
      "completion_length": 23.578125,
      "epoch": 0.43010752688172044,
      "grad_norm": 3.9474425315856934,
      "kl": 0.234375,
      "learning_rate": 8.930769230769231e-07,
      "loss": 0.0882,
      "reward": 0.6778824925422668,
      "reward_std": 0.10679807513952255,
      "reward_std_all": 0.11277594417333603,
      "rewards-max/forensics_sinkhorn_reward": 0.9217803180217743,
      "rewards-max/forensics_strict_edge_reward": 0.15571584236568015,
      "rewards-max/forensics_tiou_reward": 0.7906446754932404,
      "rewards/forensics_sinkhorn_reward": 0.5453579723834991,
      "rewards/forensics_strict_edge_reward": 0.006657965837007396,
      "rewards/forensics_tiou_reward": 0.26650601625442505,
      "step": 140
    },
    {
      "completion_length": 22.609375,
      "epoch": 0.43317972350230416,
      "grad_norm": 4.792158603668213,
      "kl": 0.2294921875,
      "learning_rate": 8.923076923076923e-07,
      "loss": 0.1356,
      "reward": 0.6963842511177063,
      "reward_std": 0.09285787120461464,
      "reward_std_all": 0.09842776134610176,
      "rewards-max/forensics_sinkhorn_reward": 0.944133460521698,
      "rewards-max/forensics_strict_edge_reward": 0.13634620374068618,
      "rewards-max/forensics_tiou_reward": 0.8827459514141083,
      "rewards/forensics_sinkhorn_reward": 0.5708144903182983,
      "rewards/forensics_strict_edge_reward": 0.00766995208687149,
      "rewards/forensics_tiou_reward": 0.3143882304430008,
      "step": 141
    },
    {
      "completion_length": 26.234375,
      "epoch": 0.4362519201228879,
      "grad_norm": 4.958629131317139,
      "kl": 0.2353515625,
      "learning_rate": 8.915384615384614e-07,
      "loss": 0.204,
      "reward": 0.7011412978172302,
      "reward_std": 0.10777286812663078,
      "reward_std_all": 0.10538806766271591,
      "rewards-max/forensics_sinkhorn_reward": 0.9711102843284607,
      "rewards-max/forensics_strict_edge_reward": 0.2559391278773546,
      "rewards-max/forensics_tiou_reward": 0.9542979300022125,
      "rewards/forensics_sinkhorn_reward": 0.5961432158946991,
      "rewards/forensics_strict_edge_reward": 0.015229815850034356,
      "rewards/forensics_tiou_reward": 0.3242039531469345,
      "step": 142
    },
    {
      "completion_length": 23.375,
      "epoch": 0.4393241167434716,
      "grad_norm": 4.871788024902344,
      "kl": 0.23046875,
      "learning_rate": 8.907692307692307e-07,
      "loss": 0.2127,
      "reward": 0.6626656651496887,
      "reward_std": 0.12130428105592728,
      "reward_std_all": 0.13419946283102036,
      "rewards-max/forensics_sinkhorn_reward": 0.9430137276649475,
      "rewards-max/forensics_strict_edge_reward": 0.20348969734914135,
      "rewards-max/forensics_tiou_reward": 0.8729286193847656,
      "rewards/forensics_sinkhorn_reward": 0.5714111626148224,
      "rewards/forensics_strict_edge_reward": 0.01586975360987708,
      "rewards/forensics_tiou_reward": 0.30851344764232635,
      "step": 143
    },
    {
      "completion_length": 25.140625,
      "epoch": 0.4423963133640553,
      "grad_norm": 4.217568397521973,
      "kl": 0.21044921875,
      "learning_rate": 8.9e-07,
      "loss": 0.102,
      "reward": 0.6746624708175659,
      "reward_std": 0.08828376233577728,
      "reward_std_all": 0.10627438873052597,
      "rewards-max/forensics_sinkhorn_reward": 0.9542522728443146,
      "rewards-max/forensics_strict_edge_reward": 0.35818374156951904,
      "rewards-max/forensics_tiou_reward": 0.9187540411949158,
      "rewards/forensics_sinkhorn_reward": 0.5653981417417526,
      "rewards/forensics_strict_edge_reward": 0.02373104728758335,
      "rewards/forensics_tiou_reward": 0.3165574073791504,
      "step": 144
    },
    {
      "completion_length": 24.625,
      "epoch": 0.445468509984639,
      "grad_norm": 4.449778079986572,
      "kl": 0.4658203125,
      "learning_rate": 8.892307692307693e-07,
      "loss": 0.2035,
      "reward": 0.6721921563148499,
      "reward_std": 0.12760038673877716,
      "reward_std_all": 0.13126518577337265,
      "rewards-max/forensics_sinkhorn_reward": 0.965978592634201,
      "rewards-max/forensics_strict_edge_reward": 0.0768802291713655,
      "rewards-max/forensics_tiou_reward": 0.9408950507640839,
      "rewards/forensics_sinkhorn_reward": 0.5537117719650269,
      "rewards/forensics_strict_edge_reward": 0.006919116305653006,
      "rewards/forensics_tiou_reward": 0.314882755279541,
      "step": 145
    },
    {
      "completion_length": 22.4375,
      "epoch": 0.4485407066052227,
      "grad_norm": 4.217392444610596,
      "kl": 0.2138671875,
      "learning_rate": 8.884615384615384e-07,
      "loss": 0.2166,
      "reward": 0.6636734008789062,
      "reward_std": 0.09037844464182854,
      "reward_std_all": 0.11409731954336166,
      "rewards-max/forensics_sinkhorn_reward": 0.9199443161487579,
      "rewards-max/forensics_strict_edge_reward": 0.14827086648438126,
      "rewards-max/forensics_tiou_reward": 0.8216037154197693,
      "rewards/forensics_sinkhorn_reward": 0.5251879692077637,
      "rewards/forensics_strict_edge_reward": 0.008820847899187356,
      "rewards/forensics_tiou_reward": 0.2484678104519844,
      "step": 146
    },
    {
      "completion_length": 24.21875,
      "epoch": 0.45161290322580644,
      "grad_norm": 4.399779319763184,
      "kl": 0.2744140625,
      "learning_rate": 8.876923076923076e-07,
      "loss": 0.1798,
      "reward": 0.6959438025951385,
      "reward_std": 0.06960267201066017,
      "reward_std_all": 0.08925118669867516,
      "rewards-max/forensics_sinkhorn_reward": 0.9677084386348724,
      "rewards-max/forensics_strict_edge_reward": 0.2783775255084038,
      "rewards-max/forensics_tiou_reward": 0.9446767270565033,
      "rewards/forensics_sinkhorn_reward": 0.5678699910640717,
      "rewards/forensics_strict_edge_reward": 0.018246880266815424,
      "rewards/forensics_tiou_reward": 0.345008984208107,
      "step": 147
    },
    {
      "completion_length": 25.3125,
      "epoch": 0.45468509984639016,
      "grad_norm": 4.44042444229126,
      "kl": 0.22607421875,
      "learning_rate": 8.869230769230769e-07,
      "loss": 0.1206,
      "reward": 0.7348821759223938,
      "reward_std": 0.09215330705046654,
      "reward_std_all": 0.08559831604361534,
      "rewards-max/forensics_sinkhorn_reward": 0.9467328786849976,
      "rewards-max/forensics_strict_edge_reward": 0.30796711426228285,
      "rewards-max/forensics_tiou_reward": 0.8794458210468292,
      "rewards/forensics_sinkhorn_reward": 0.6478594541549683,
      "rewards/forensics_strict_edge_reward": 0.023634315701201558,
      "rewards/forensics_tiou_reward": 0.4027377814054489,
      "step": 148
    },
    {
      "completion_length": 23.78125,
      "epoch": 0.4577572964669739,
      "grad_norm": 4.064465045928955,
      "kl": 0.20458984375,
      "learning_rate": 8.861538461538461e-07,
      "loss": 0.1924,
      "reward": 0.6841771602630615,
      "reward_std": 0.1000148169696331,
      "reward_std_all": 0.10770749300718307,
      "rewards-max/forensics_sinkhorn_reward": 0.9255901575088501,
      "rewards-max/forensics_strict_edge_reward": 0.1144559197127819,
      "rewards-max/forensics_tiou_reward": 0.8447557091712952,
      "rewards/forensics_sinkhorn_reward": 0.5606777667999268,
      "rewards/forensics_strict_edge_reward": 0.005072564817965031,
      "rewards/forensics_tiou_reward": 0.29787617921829224,
      "step": 149
    },
    {
      "completion_length": 23.828125,
      "epoch": 0.4608294930875576,
      "grad_norm": 6.945254325866699,
      "kl": 0.23046875,
      "learning_rate": 8.853846153846154e-07,
      "loss": 0.1815,
      "reward": 0.7232116460800171,
      "reward_std": 0.09074846282601357,
      "reward_std_all": 0.08692970499396324,
      "rewards-max/forensics_sinkhorn_reward": 0.9832765758037567,
      "rewards-max/forensics_strict_edge_reward": 0.6282242834568024,
      "rewards-max/forensics_tiou_reward": 0.9833943843841553,
      "rewards/forensics_sinkhorn_reward": 0.6498846113681793,
      "rewards/forensics_strict_edge_reward": 0.0293005658313632,
      "rewards/forensics_tiou_reward": 0.39905086159706116,
      "step": 150
    },
    {
      "completion_length": 26.265625,
      "epoch": 0.46390168970814133,
      "grad_norm": 4.5512166023254395,
      "kl": 0.240234375,
      "learning_rate": 8.846153846153846e-07,
      "loss": 0.2493,
      "reward": 0.7008373141288757,
      "reward_std": 0.11713284626603127,
      "reward_std_all": 0.1118561327457428,
      "rewards-max/forensics_sinkhorn_reward": 0.938999742269516,
      "rewards-max/forensics_strict_edge_reward": 0.05180455558001995,
      "rewards-max/forensics_tiou_reward": 0.878886342048645,
      "rewards/forensics_sinkhorn_reward": 0.5765085518360138,
      "rewards/forensics_strict_edge_reward": 0.002261372748762369,
      "rewards/forensics_tiou_reward": 0.3236112743616104,
      "step": 151
    },
    {
      "completion_length": 25.625,
      "epoch": 0.46697388632872505,
      "grad_norm": 5.500548362731934,
      "kl": 0.341796875,
      "learning_rate": 8.838461538461538e-07,
      "loss": 0.0915,
      "reward": 0.6704597473144531,
      "reward_std": 0.10062473639845848,
      "reward_std_all": 0.12025376036763191,
      "rewards-max/forensics_sinkhorn_reward": 0.9609903693199158,
      "rewards-max/forensics_strict_edge_reward": 0.15656040981411934,
      "rewards-max/forensics_tiou_reward": 0.9228677153587341,
      "rewards/forensics_sinkhorn_reward": 0.5803723633289337,
      "rewards/forensics_strict_edge_reward": 0.010807888582348824,
      "rewards/forensics_tiou_reward": 0.3214395344257355,
      "step": 152
    },
    {
      "completion_length": 24.484375,
      "epoch": 0.4700460829493088,
      "grad_norm": 4.845566749572754,
      "kl": 0.24169921875,
      "learning_rate": 8.830769230769231e-07,
      "loss": 0.0262,
      "reward": 0.6819782853126526,
      "reward_std": 0.13529227674007416,
      "reward_std_all": 0.1317717656493187,
      "rewards-max/forensics_sinkhorn_reward": 0.9410677850246429,
      "rewards-max/forensics_strict_edge_reward": 0.031222350895404816,
      "rewards-max/forensics_tiou_reward": 0.8851908445358276,
      "rewards/forensics_sinkhorn_reward": 0.5653812289237976,
      "rewards/forensics_strict_edge_reward": 0.0020749172545038164,
      "rewards/forensics_tiou_reward": 0.3295438140630722,
      "step": 153
    },
    {
      "completion_length": 23.921875,
      "epoch": 0.4731182795698925,
      "grad_norm": 5.519564151763916,
      "kl": 0.2705078125,
      "learning_rate": 8.823076923076923e-07,
      "loss": 0.0261,
      "reward": 0.6485837996006012,
      "reward_std": 0.08361990004777908,
      "reward_std_all": 0.11715476959943771,
      "rewards-max/forensics_sinkhorn_reward": 0.9255356192588806,
      "rewards-max/forensics_strict_edge_reward": 0.17467453703284264,
      "rewards-max/forensics_tiou_reward": 0.8487852215766907,
      "rewards/forensics_sinkhorn_reward": 0.5531803220510483,
      "rewards/forensics_strict_edge_reward": 0.009055446367710829,
      "rewards/forensics_tiou_reward": 0.2491452693939209,
      "step": 154
    },
    {
      "completion_length": 22.65625,
      "epoch": 0.47619047619047616,
      "grad_norm": 5.0446624755859375,
      "kl": 0.22021484375,
      "learning_rate": 8.815384615384614e-07,
      "loss": 0.1289,
      "reward": 0.7490330636501312,
      "reward_std": 0.06994971632957458,
      "reward_std_all": 0.06633937358856201,
      "rewards-max/forensics_sinkhorn_reward": 0.9619845449924469,
      "rewards-max/forensics_strict_edge_reward": 0.14613009244203568,
      "rewards-max/forensics_tiou_reward": 0.9279092848300934,
      "rewards/forensics_sinkhorn_reward": 0.6651309728622437,
      "rewards/forensics_strict_edge_reward": 0.00957674952223897,
      "rewards/forensics_tiou_reward": 0.379882276058197,
      "step": 155
    },
    {
      "completion_length": 23.15625,
      "epoch": 0.4792626728110599,
      "grad_norm": 5.216426372528076,
      "kl": 0.25146484375,
      "learning_rate": 8.807692307692307e-07,
      "loss": -0.0575,
      "reward": 0.7170002460479736,
      "reward_std": 0.06855271756649017,
      "reward_std_all": 0.0802885890007019,
      "rewards-max/forensics_sinkhorn_reward": 0.9676449298858643,
      "rewards-max/forensics_strict_edge_reward": 0.22547354741254821,
      "rewards-max/forensics_tiou_reward": 0.9295676648616791,
      "rewards/forensics_sinkhorn_reward": 0.6128226220607758,
      "rewards/forensics_strict_edge_reward": 0.009517764698102837,
      "rewards/forensics_tiou_reward": 0.3495504707098007,
      "step": 156
    },
    {
      "completion_length": 24.375,
      "epoch": 0.4823348694316436,
      "grad_norm": 4.550693035125732,
      "kl": 0.23095703125,
      "learning_rate": 8.799999999999999e-07,
      "loss": 0.1971,
      "reward": 0.7240906357765198,
      "reward_std": 0.07537032105028629,
      "reward_std_all": 0.08135461620986462,
      "rewards-max/forensics_sinkhorn_reward": 0.9396239519119263,
      "rewards-max/forensics_strict_edge_reward": 0.11352772824466228,
      "rewards-max/forensics_tiou_reward": 0.871220201253891,
      "rewards/forensics_sinkhorn_reward": 0.6307456195354462,
      "rewards/forensics_strict_edge_reward": 0.004993458511307836,
      "rewards/forensics_tiou_reward": 0.35720258951187134,
      "step": 157
    },
    {
      "completion_length": 22.78125,
      "epoch": 0.48540706605222733,
      "grad_norm": 4.5417985916137695,
      "kl": 0.27734375,
      "learning_rate": 8.792307692307693e-07,
      "loss": 0.0478,
      "reward": 0.7232173085212708,
      "reward_std": 0.07486564666032791,
      "reward_std_all": 0.07681325823068619,
      "rewards-max/forensics_sinkhorn_reward": 0.9379238486289978,
      "rewards-max/forensics_strict_edge_reward": 0.28674357384443283,
      "rewards-max/forensics_tiou_reward": 0.8640895187854767,
      "rewards/forensics_sinkhorn_reward": 0.608306884765625,
      "rewards/forensics_strict_edge_reward": 0.018570775166153908,
      "rewards/forensics_tiou_reward": 0.34443874657154083,
      "step": 158
    },
    {
      "completion_length": 22.0,
      "epoch": 0.48847926267281105,
      "grad_norm": 4.129448890686035,
      "kl": 0.30859375,
      "learning_rate": 8.784615384615385e-07,
      "loss": 0.1077,
      "reward": 0.7226527035236359,
      "reward_std": 0.08825143612921238,
      "reward_std_all": 0.08652303740382195,
      "rewards-max/forensics_sinkhorn_reward": 0.9678252637386322,
      "rewards-max/forensics_strict_edge_reward": 0.35060252249240875,
      "rewards-max/forensics_tiou_reward": 0.931475967168808,
      "rewards/forensics_sinkhorn_reward": 0.6270008385181427,
      "rewards/forensics_strict_edge_reward": 0.015403814613819122,
      "rewards/forensics_tiou_reward": 0.3194589912891388,
      "step": 159
    },
    {
      "completion_length": 25.21875,
      "epoch": 0.4915514592933948,
      "grad_norm": 5.117344379425049,
      "kl": 0.2890625,
      "learning_rate": 8.776923076923076e-07,
      "loss": 0.0741,
      "reward": 0.7270995676517487,
      "reward_std": 0.08493782207369804,
      "reward_std_all": 0.08234677463769913,
      "rewards-max/forensics_sinkhorn_reward": 0.8114226162433624,
      "rewards-max/forensics_strict_edge_reward": 0.031856439309194684,
      "rewards-max/forensics_tiou_reward": 0.6256107687950134,
      "rewards/forensics_sinkhorn_reward": 0.6000070571899414,
      "rewards/forensics_strict_edge_reward": 0.002623651278554462,
      "rewards/forensics_tiou_reward": 0.27261078357696533,
      "step": 160
    },
    {
      "completion_length": 22.140625,
      "epoch": 0.4946236559139785,
      "grad_norm": 5.035660266876221,
      "kl": 0.28515625,
      "learning_rate": 8.769230769230769e-07,
      "loss": 0.1456,
      "reward": 0.7057152390480042,
      "reward_std": 0.0724222008138895,
      "reward_std_all": 0.08232743293046951,
      "rewards-max/forensics_sinkhorn_reward": 0.8569810390472412,
      "rewards-max/forensics_strict_edge_reward": 0.26924880035221577,
      "rewards-max/forensics_tiou_reward": 0.766091912984848,
      "rewards/forensics_sinkhorn_reward": 0.5904027819633484,
      "rewards/forensics_strict_edge_reward": 0.011323092738166451,
      "rewards/forensics_tiou_reward": 0.28453126549720764,
      "step": 161
    },
    {
      "completion_length": 22.25,
      "epoch": 0.4976958525345622,
      "grad_norm": 5.187171459197998,
      "kl": 0.3046875,
      "learning_rate": 8.761538461538461e-07,
      "loss": 0.0124,
      "reward": 0.6499526798725128,
      "reward_std": 0.14054929837584496,
      "reward_std_all": 0.14276979491114616,
      "rewards-max/forensics_sinkhorn_reward": 0.9117198884487152,
      "rewards-max/forensics_strict_edge_reward": 0.10861724801361561,
      "rewards-max/forensics_tiou_reward": 0.8051945567131042,
      "rewards/forensics_sinkhorn_reward": 0.5624239444732666,
      "rewards/forensics_strict_edge_reward": 0.00659868324873969,
      "rewards/forensics_tiou_reward": 0.2794494926929474,
      "step": 162
    },
    {
      "completion_length": 24.109375,
      "epoch": 0.500768049155146,
      "grad_norm": 5.018589019775391,
      "kl": 0.25,
      "learning_rate": 8.753846153846154e-07,
      "loss": 0.1483,
      "reward": 0.7157326340675354,
      "reward_std": 0.06303860805928707,
      "reward_std_all": 0.07831531763076782,
      "rewards-max/forensics_sinkhorn_reward": 0.9339728355407715,
      "rewards-max/forensics_strict_edge_reward": 0.0704179797321558,
      "rewards-max/forensics_tiou_reward": 0.8651198744773865,
      "rewards/forensics_sinkhorn_reward": 0.603749543428421,
      "rewards/forensics_strict_edge_reward": 0.004108201363123953,
      "rewards/forensics_tiou_reward": 0.3415427803993225,
      "step": 163
    },
    {
      "completion_length": 22.703125,
      "epoch": 0.5038402457757296,
      "grad_norm": 4.4569268226623535,
      "kl": 0.26953125,
      "learning_rate": 8.746153846153845e-07,
      "loss": 0.1771,
      "reward": 0.6943131685256958,
      "reward_std": 0.11322323232889175,
      "reward_std_all": 0.1109497994184494,
      "rewards-max/forensics_sinkhorn_reward": 0.8934144675731659,
      "rewards-max/forensics_strict_edge_reward": 0.07246137224137783,
      "rewards-max/forensics_tiou_reward": 0.7792040109634399,
      "rewards/forensics_sinkhorn_reward": 0.5837436318397522,
      "rewards/forensics_strict_edge_reward": 0.0046471161767840385,
      "rewards/forensics_tiou_reward": 0.288415864109993,
      "step": 164
    },
    {
      "completion_length": 23.09375,
      "epoch": 0.5069124423963134,
      "grad_norm": 5.045226573944092,
      "kl": 0.2724609375,
      "learning_rate": 8.738461538461538e-07,
      "loss": 0.0647,
      "reward": 0.6938222646713257,
      "reward_std": 0.07846295461058617,
      "reward_std_all": 0.09316589683294296,
      "rewards-max/forensics_sinkhorn_reward": 0.8956686556339264,
      "rewards-max/forensics_strict_edge_reward": 0.18395296942981076,
      "rewards-max/forensics_tiou_reward": 0.786674439907074,
      "rewards/forensics_sinkhorn_reward": 0.5815131068229675,
      "rewards/forensics_strict_edge_reward": 0.01748156198198103,
      "rewards/forensics_tiou_reward": 0.2846509665250778,
      "step": 165
    },
    {
      "completion_length": 23.421875,
      "epoch": 0.5099846390168971,
      "grad_norm": 4.46400785446167,
      "kl": 0.265625,
      "learning_rate": 8.730769230769231e-07,
      "loss": 0.0527,
      "reward": 0.7509584128856659,
      "reward_std": 0.049903564155101776,
      "reward_std_all": 0.053984226658940315,
      "rewards-max/forensics_sinkhorn_reward": 0.9685367047786713,
      "rewards-max/forensics_strict_edge_reward": 0.18959669023752213,
      "rewards-max/forensics_tiou_reward": 0.9394062459468842,
      "rewards/forensics_sinkhorn_reward": 0.693334311246872,
      "rewards/forensics_strict_edge_reward": 0.01707451546099037,
      "rewards/forensics_tiou_reward": 0.4610304534435272,
      "step": 166
    },
    {
      "completion_length": 21.359375,
      "epoch": 0.5130568356374808,
      "grad_norm": 5.999330997467041,
      "kl": 0.24658203125,
      "learning_rate": 8.723076923076923e-07,
      "loss": 0.0778,
      "reward": 0.66765096783638,
      "reward_std": 0.1192675344645977,
      "reward_std_all": 0.12485218793153763,
      "rewards-max/forensics_sinkhorn_reward": 0.9549171328544617,
      "rewards-max/forensics_strict_edge_reward": 0.21754248067736626,
      "rewards-max/forensics_tiou_reward": 0.9169151186943054,
      "rewards/forensics_sinkhorn_reward": 0.5730946958065033,
      "rewards/forensics_strict_edge_reward": 0.013476673513650894,
      "rewards/forensics_tiou_reward": 0.305097296833992,
      "step": 167
    },
    {
      "completion_length": 23.875,
      "epoch": 0.5161290322580645,
      "grad_norm": 5.026996612548828,
      "kl": 0.3193359375,
      "learning_rate": 8.715384615384614e-07,
      "loss": 0.0106,
      "reward": 0.705724447965622,
      "reward_std": 0.06349419802427292,
      "reward_std_all": 0.07938066869974136,
      "rewards-max/forensics_sinkhorn_reward": 0.8863535225391388,
      "rewards-max/forensics_strict_edge_reward": 0.27734443731606007,
      "rewards-max/forensics_tiou_reward": 0.81759113073349,
      "rewards/forensics_sinkhorn_reward": 0.5980517864227295,
      "rewards/forensics_strict_edge_reward": 0.0117677201051265,
      "rewards/forensics_tiou_reward": 0.3370661288499832,
      "step": 168
    },
    {
      "completion_length": 24.65625,
      "epoch": 0.5192012288786483,
      "grad_norm": 5.565403938293457,
      "kl": 0.33984375,
      "learning_rate": 8.707692307692307e-07,
      "loss": 0.1203,
      "reward": 0.6719127297401428,
      "reward_std": 0.09668517112731934,
      "reward_std_all": 0.11105294898152351,
      "rewards-max/forensics_sinkhorn_reward": 0.9107454717159271,
      "rewards-max/forensics_strict_edge_reward": 0.31077447533607483,
      "rewards-max/forensics_tiou_reward": 0.8072304427623749,
      "rewards/forensics_sinkhorn_reward": 0.5601408779621124,
      "rewards/forensics_strict_edge_reward": 0.018427591305226088,
      "rewards/forensics_tiou_reward": 0.29624584317207336,
      "step": 169
    },
    {
      "completion_length": 21.546875,
      "epoch": 0.522273425499232,
      "grad_norm": 5.201751232147217,
      "kl": 0.28125,
      "learning_rate": 8.699999999999999e-07,
      "loss": 0.035,
      "reward": 0.6807001233100891,
      "reward_std": 0.09993725642561913,
      "reward_std_all": 0.11480408906936646,
      "rewards-max/forensics_sinkhorn_reward": 0.8952707648277283,
      "rewards-max/forensics_strict_edge_reward": 0.009764846414327621,
      "rewards-max/forensics_tiou_reward": 0.7390807569026947,
      "rewards/forensics_sinkhorn_reward": 0.5781146883964539,
      "rewards/forensics_strict_edge_reward": 0.0006222765077836812,
      "rewards/forensics_tiou_reward": 0.34124647080898285,
      "step": 170
    },
    {
      "completion_length": 23.484375,
      "epoch": 0.5253456221198156,
      "grad_norm": 3.722832202911377,
      "kl": 0.2490234375,
      "learning_rate": 8.692307692307692e-07,
      "loss": 0.1933,
      "reward": 0.7233347296714783,
      "reward_std": 0.07963572442531586,
      "reward_std_all": 0.08040521293878555,
      "rewards-max/forensics_sinkhorn_reward": 0.9599876999855042,
      "rewards-max/forensics_strict_edge_reward": 0.13356080651283264,
      "rewards-max/forensics_tiou_reward": 0.9059800803661346,
      "rewards/forensics_sinkhorn_reward": 0.6376855671405792,
      "rewards/forensics_strict_edge_reward": 0.007502663880586624,
      "rewards/forensics_tiou_reward": 0.3642522245645523,
      "step": 171
    },
    {
      "completion_length": 25.15625,
      "epoch": 0.5284178187403994,
      "grad_norm": 12.140303611755371,
      "kl": 0.6025390625,
      "learning_rate": 8.684615384615385e-07,
      "loss": 0.2026,
      "reward": 0.7288776934146881,
      "reward_std": 0.04824382625520229,
      "reward_std_all": 0.0636291392147541,
      "rewards-max/forensics_sinkhorn_reward": 0.9011913537979126,
      "rewards-max/forensics_strict_edge_reward": 0.12272865325212479,
      "rewards-max/forensics_tiou_reward": 0.7667510509490967,
      "rewards/forensics_sinkhorn_reward": 0.5847241878509521,
      "rewards/forensics_strict_edge_reward": 0.008377352263778448,
      "rewards/forensics_tiou_reward": 0.34366393089294434,
      "step": 172
    },
    {
      "completion_length": 23.546875,
      "epoch": 0.5314900153609831,
      "grad_norm": 5.6332526206970215,
      "kl": 0.357421875,
      "learning_rate": 8.676923076923076e-07,
      "loss": 0.1163,
      "reward": 0.6601263284683228,
      "reward_std": 0.09216349944472313,
      "reward_std_all": 0.11434592679142952,
      "rewards-max/forensics_sinkhorn_reward": 0.949179470539093,
      "rewards-max/forensics_strict_edge_reward": 0.04457263310905546,
      "rewards-max/forensics_tiou_reward": 0.8878839313983917,
      "rewards/forensics_sinkhorn_reward": 0.5351188778877258,
      "rewards/forensics_strict_edge_reward": 0.003934826343538589,
      "rewards/forensics_tiou_reward": 0.26014629006385803,
      "step": 173
    },
    {
      "completion_length": 23.609375,
      "epoch": 0.5345622119815668,
      "grad_norm": 4.640859127044678,
      "kl": 0.291015625,
      "learning_rate": 8.669230769230769e-07,
      "loss": -0.0309,
      "reward": 0.6770159006118774,
      "reward_std": 0.09861288592219353,
      "reward_std_all": 0.1080603115260601,
      "rewards-max/forensics_sinkhorn_reward": 0.9765845537185669,
      "rewards-max/forensics_strict_edge_reward": 0.38056008517742157,
      "rewards-max/forensics_tiou_reward": 0.967382937669754,
      "rewards/forensics_sinkhorn_reward": 0.5643880069255829,
      "rewards/forensics_strict_edge_reward": 0.03471472393721342,
      "rewards/forensics_tiou_reward": 0.3539722114801407,
      "step": 174
    },
    {
      "completion_length": 22.53125,
      "epoch": 0.5376344086021505,
      "grad_norm": 5.373100757598877,
      "kl": 0.31201171875,
      "learning_rate": 8.661538461538461e-07,
      "loss": 0.0577,
      "reward": 0.7050711214542389,
      "reward_std": 0.09913821145892143,
      "reward_std_all": 0.09804659336805344,
      "rewards-max/forensics_sinkhorn_reward": 0.8905023634433746,
      "rewards-max/forensics_strict_edge_reward": 0.025158059550449252,
      "rewards-max/forensics_tiou_reward": 0.7483513653278351,
      "rewards/forensics_sinkhorn_reward": 0.5908448994159698,
      "rewards/forensics_strict_edge_reward": 0.002103349397657439,
      "rewards/forensics_tiou_reward": 0.3078334629535675,
      "step": 175
    },
    {
      "completion_length": 25.375,
      "epoch": 0.5407066052227343,
      "grad_norm": 5.483180046081543,
      "kl": 0.267578125,
      "learning_rate": 8.653846153846154e-07,
      "loss": 0.1101,
      "reward": 0.7006265223026276,
      "reward_std": 0.09108537063002586,
      "reward_std_all": 0.09367525950074196,
      "rewards-max/forensics_sinkhorn_reward": 0.9522243142127991,
      "rewards-max/forensics_strict_edge_reward": 0.06392737291753292,
      "rewards-max/forensics_tiou_reward": 0.9079914391040802,
      "rewards/forensics_sinkhorn_reward": 0.5826828479766846,
      "rewards/forensics_strict_edge_reward": 0.003588411840610206,
      "rewards/forensics_tiou_reward": 0.3494526594877243,
      "step": 176
    },
    {
      "completion_length": 29.5,
      "epoch": 0.543778801843318,
      "grad_norm": 5.116785526275635,
      "kl": 0.3310546875,
      "learning_rate": 8.646153846153845e-07,
      "loss": 0.1846,
      "reward": 0.6411579251289368,
      "reward_std": 0.14205124974250793,
      "reward_std_all": 0.15145523846149445,
      "rewards-max/forensics_sinkhorn_reward": 0.9514627754688263,
      "rewards-max/forensics_strict_edge_reward": 0.1425763824954629,
      "rewards-max/forensics_tiou_reward": 0.8951018154621124,
      "rewards/forensics_sinkhorn_reward": 0.4900350272655487,
      "rewards/forensics_strict_edge_reward": 0.00760829052887857,
      "rewards/forensics_tiou_reward": 0.3175799697637558,
      "step": 177
    },
    {
      "completion_length": 27.671875,
      "epoch": 0.5468509984639017,
      "grad_norm": 6.424310207366943,
      "kl": 0.3203125,
      "learning_rate": 8.638461538461537e-07,
      "loss": 0.1431,
      "reward": 0.6318187415599823,
      "reward_std": 0.1116851456463337,
      "reward_std_all": 0.137628685683012,
      "rewards-max/forensics_sinkhorn_reward": 0.9304232001304626,
      "rewards-max/forensics_strict_edge_reward": 0.2506886227056384,
      "rewards-max/forensics_tiou_reward": 0.8634383976459503,
      "rewards/forensics_sinkhorn_reward": 0.4985942840576172,
      "rewards/forensics_strict_edge_reward": 0.01554764376487583,
      "rewards/forensics_tiou_reward": 0.33760271966457367,
      "step": 178
    },
    {
      "completion_length": 24.734375,
      "epoch": 0.5499231950844854,
      "grad_norm": 4.673572540283203,
      "kl": 0.2353515625,
      "learning_rate": 8.630769230769231e-07,
      "loss": 0.1541,
      "reward": 0.6944327354431152,
      "reward_std": 0.09630150720477104,
      "reward_std_all": 0.10033155605196953,
      "rewards-max/forensics_sinkhorn_reward": 0.9567544162273407,
      "rewards-max/forensics_strict_edge_reward": 0.18887560069561005,
      "rewards-max/forensics_tiou_reward": 0.9210394620895386,
      "rewards/forensics_sinkhorn_reward": 0.6026080250740051,
      "rewards/forensics_strict_edge_reward": 0.014420160092413425,
      "rewards/forensics_tiou_reward": 0.3714005947113037,
      "step": 179
    },
    {
      "completion_length": 24.078125,
      "epoch": 0.5529953917050692,
      "grad_norm": 6.057821750640869,
      "kl": 0.2646484375,
      "learning_rate": 8.623076923076923e-07,
      "loss": 0.1334,
      "reward": 0.6684362590312958,
      "reward_std": 0.1052723228931427,
      "reward_std_all": 0.12192235514521599,
      "rewards-max/forensics_sinkhorn_reward": 0.9615969955921173,
      "rewards-max/forensics_strict_edge_reward": 0.48182694986462593,
      "rewards-max/forensics_tiou_reward": 0.9313267171382904,
      "rewards/forensics_sinkhorn_reward": 0.5542276352643967,
      "rewards/forensics_strict_edge_reward": 0.021709568100050092,
      "rewards/forensics_tiou_reward": 0.2957174628973007,
      "step": 180
    },
    {
      "completion_length": 27.46875,
      "epoch": 0.5560675883256528,
      "grad_norm": 4.499628067016602,
      "kl": 0.419921875,
      "learning_rate": 8.615384615384616e-07,
      "loss": 0.1793,
      "reward": 0.7108199000358582,
      "reward_std": 0.07232211902737617,
      "reward_std_all": 0.08188512176275253,
      "rewards-max/forensics_sinkhorn_reward": 0.9620414972305298,
      "rewards-max/forensics_strict_edge_reward": 0.39520856738090515,
      "rewards-max/forensics_tiou_reward": 0.9340132474899292,
      "rewards/forensics_sinkhorn_reward": 0.6003974676132202,
      "rewards/forensics_strict_edge_reward": 0.018106555566191673,
      "rewards/forensics_tiou_reward": 0.33299368619918823,
      "step": 181
    },
    {
      "completion_length": 25.359375,
      "epoch": 0.5591397849462365,
      "grad_norm": 4.5528564453125,
      "kl": 0.33251953125,
      "learning_rate": 8.607692307692307e-07,
      "loss": 0.2308,
      "reward": 0.7409724593162537,
      "reward_std": 0.08106275647878647,
      "reward_std_all": 0.07593535631895065,
      "rewards-max/forensics_sinkhorn_reward": 0.9391268193721771,
      "rewards-max/forensics_strict_edge_reward": 0.26491051726043224,
      "rewards-max/forensics_tiou_reward": 0.8682626783847809,
      "rewards/forensics_sinkhorn_reward": 0.6430345177650452,
      "rewards/forensics_strict_edge_reward": 0.015459030168130994,
      "rewards/forensics_tiou_reward": 0.42233413457870483,
      "step": 182
    },
    {
      "completion_length": 23.359375,
      "epoch": 0.5622119815668203,
      "grad_norm": 6.179098129272461,
      "kl": 0.28369140625,
      "learning_rate": 8.599999999999999e-07,
      "loss": 0.1878,
      "reward": 0.7115383148193359,
      "reward_std": 0.0804961770772934,
      "reward_std_all": 0.08571914583444595,
      "rewards-max/forensics_sinkhorn_reward": 0.8717063069343567,
      "rewards-max/forensics_strict_edge_reward": 0.03753862716257572,
      "rewards-max/forensics_tiou_reward": 0.6671188175678253,
      "rewards/forensics_sinkhorn_reward": 0.6042337715625763,
      "rewards/forensics_strict_edge_reward": 0.0016753404634073377,
      "rewards/forensics_tiou_reward": 0.32024379074573517,
      "step": 183
    },
    {
      "completion_length": 21.71875,
      "epoch": 0.565284178187404,
      "grad_norm": 5.5975728034973145,
      "kl": 0.2373046875,
      "learning_rate": 8.592307692307692e-07,
      "loss": 0.1567,
      "reward": 0.7167530655860901,
      "reward_std": 0.07831653580069542,
      "reward_std_all": 0.08190694823861122,
      "rewards-max/forensics_sinkhorn_reward": 0.9583435952663422,
      "rewards-max/forensics_strict_edge_reward": 0.21566618978977203,
      "rewards-max/forensics_tiou_reward": 0.9174966514110565,
      "rewards/forensics_sinkhorn_reward": 0.6221041977405548,
      "rewards/forensics_strict_edge_reward": 0.01870965026319027,
      "rewards/forensics_tiou_reward": 0.3026004582643509,
      "step": 184
    },
    {
      "completion_length": 24.109375,
      "epoch": 0.5683563748079877,
      "grad_norm": 6.8495707511901855,
      "kl": 0.26171875,
      "learning_rate": 8.584615384615385e-07,
      "loss": 0.1655,
      "reward": 0.7070267498493195,
      "reward_std": 0.09808700904250145,
      "reward_std_all": 0.09855572506785393,
      "rewards-max/forensics_sinkhorn_reward": 0.9581574201583862,
      "rewards-max/forensics_strict_edge_reward": 0.291122630238533,
      "rewards-max/forensics_tiou_reward": 0.9257367849349976,
      "rewards/forensics_sinkhorn_reward": 0.6114894151687622,
      "rewards/forensics_strict_edge_reward": 0.014610711019486189,
      "rewards/forensics_tiou_reward": 0.37464335560798645,
      "step": 185
    },
    {
      "completion_length": 21.515625,
      "epoch": 0.5714285714285714,
      "grad_norm": 5.157290935516357,
      "kl": 0.296875,
      "learning_rate": 8.576923076923076e-07,
      "loss": 0.1658,
      "reward": 0.7437226176261902,
      "reward_std": 0.0842093974351883,
      "reward_std_all": 0.07805406674742699,
      "rewards-max/forensics_sinkhorn_reward": 0.9301123023033142,
      "rewards-max/forensics_strict_edge_reward": 0.11585588939487934,
      "rewards-max/forensics_tiou_reward": 0.8516813516616821,
      "rewards/forensics_sinkhorn_reward": 0.6672173142433167,
      "rewards/forensics_strict_edge_reward": 0.010433299699798226,
      "rewards/forensics_tiou_reward": 0.4005502760410309,
      "step": 186
    },
    {
      "completion_length": 22.734375,
      "epoch": 0.5745007680491552,
      "grad_norm": 7.397681713104248,
      "kl": 0.291015625,
      "learning_rate": 8.569230769230769e-07,
      "loss": 0.0724,
      "reward": 0.6628270149230957,
      "reward_std": 0.07016284018754959,
      "reward_std_all": 0.10396897420287132,
      "rewards-max/forensics_sinkhorn_reward": 0.9386081397533417,
      "rewards-max/forensics_strict_edge_reward": 0.38418035581707954,
      "rewards-max/forensics_tiou_reward": 0.8757503926753998,
      "rewards/forensics_sinkhorn_reward": 0.5763638019561768,
      "rewards/forensics_strict_edge_reward": 0.02431737259030342,
      "rewards/forensics_tiou_reward": 0.36000020802021027,
      "step": 187
    },
    {
      "completion_length": 22.609375,
      "epoch": 0.5775729646697388,
      "grad_norm": 4.806649684906006,
      "kl": 0.330078125,
      "learning_rate": 8.561538461538461e-07,
      "loss": 0.164,
      "reward": 0.7111903131008148,
      "reward_std": 0.0947536751627922,
      "reward_std_all": 0.09603395313024521,
      "rewards-max/forensics_sinkhorn_reward": 0.9521118700504303,
      "rewards-max/forensics_strict_edge_reward": 0.3339256439357996,
      "rewards-max/forensics_tiou_reward": 0.9033281803131104,
      "rewards/forensics_sinkhorn_reward": 0.623638778924942,
      "rewards/forensics_strict_edge_reward": 0.022791715105995536,
      "rewards/forensics_tiou_reward": 0.33363035321235657,
      "step": 188
    },
    {
      "completion_length": 22.796875,
      "epoch": 0.5806451612903226,
      "grad_norm": 4.902263641357422,
      "kl": 0.26953125,
      "learning_rate": 8.553846153846154e-07,
      "loss": 0.0746,
      "reward": 0.7386338114738464,
      "reward_std": 0.0874734774697572,
      "reward_std_all": 0.08156840037554502,
      "rewards-max/forensics_sinkhorn_reward": 0.9729403853416443,
      "rewards-max/forensics_strict_edge_reward": 0.25579044315963984,
      "rewards-max/forensics_tiou_reward": 0.9422022104263306,
      "rewards/forensics_sinkhorn_reward": 0.642808198928833,
      "rewards/forensics_strict_edge_reward": 0.019468923623207957,
      "rewards/forensics_tiou_reward": 0.4470099210739136,
      "step": 189
    },
    {
      "completion_length": 23.890625,
      "epoch": 0.5837173579109063,
      "grad_norm": 4.136417388916016,
      "kl": 0.2783203125,
      "learning_rate": 8.546153846153846e-07,
      "loss": 0.0419,
      "reward": 0.7043465375900269,
      "reward_std": 0.08956032246351242,
      "reward_std_all": 0.09543412551283836,
      "rewards-max/forensics_sinkhorn_reward": 0.9679760336875916,
      "rewards-max/forensics_strict_edge_reward": 0.5374038442969322,
      "rewards-max/forensics_tiou_reward": 0.9425724744796753,
      "rewards/forensics_sinkhorn_reward": 0.5758253633975983,
      "rewards/forensics_strict_edge_reward": 0.03345056623220444,
      "rewards/forensics_tiou_reward": 0.3476129174232483,
      "step": 190
    },
    {
      "completion_length": 23.484375,
      "epoch": 0.5867895545314901,
      "grad_norm": 5.150000095367432,
      "kl": 0.2685546875,
      "learning_rate": 8.538461538461537e-07,
      "loss": 0.0301,
      "reward": 0.6983811855316162,
      "reward_std": 0.08474539965391159,
      "reward_std_all": 0.09235577657818794,
      "rewards-max/forensics_sinkhorn_reward": 0.9427551031112671,
      "rewards-max/forensics_strict_edge_reward": 0.5661376565694809,
      "rewards-max/forensics_tiou_reward": 0.8842841684818268,
      "rewards/forensics_sinkhorn_reward": 0.5692873895168304,
      "rewards/forensics_strict_edge_reward": 0.028120005503296852,
      "rewards/forensics_tiou_reward": 0.30300359427928925,
      "step": 191
    },
    {
      "completion_length": 25.203125,
      "epoch": 0.5898617511520737,
      "grad_norm": 5.777318000793457,
      "kl": 0.271484375,
      "learning_rate": 8.530769230769231e-07,
      "loss": 0.1718,
      "reward": 0.7271216511726379,
      "reward_std": 0.08501097559928894,
      "reward_std_all": 0.08303194120526314,
      "rewards-max/forensics_sinkhorn_reward": 0.9381106495857239,
      "rewards-max/forensics_strict_edge_reward": 0.018799281679093838,
      "rewards-max/forensics_tiou_reward": 0.8627760410308838,
      "rewards/forensics_sinkhorn_reward": 0.6440353691577911,
      "rewards/forensics_strict_edge_reward": 0.000996890405076556,
      "rewards/forensics_tiou_reward": 0.4083795100450516,
      "step": 192
    },
    {
      "completion_length": 22.265625,
      "epoch": 0.5929339477726574,
      "grad_norm": 5.472795009613037,
      "kl": 0.279296875,
      "learning_rate": 8.523076923076923e-07,
      "loss": 0.1418,
      "reward": 0.6527627110481262,
      "reward_std": 0.1394919529557228,
      "reward_std_all": 0.14100422337651253,
      "rewards-max/forensics_sinkhorn_reward": 0.941758006811142,
      "rewards-max/forensics_strict_edge_reward": 0.29928118363022804,
      "rewards-max/forensics_tiou_reward": 0.8931681513786316,
      "rewards/forensics_sinkhorn_reward": 0.5484035313129425,
      "rewards/forensics_strict_edge_reward": 0.018236498348414898,
      "rewards/forensics_tiou_reward": 0.29703207314014435,
      "step": 193
    },
    {
      "completion_length": 21.6875,
      "epoch": 0.5960061443932412,
      "grad_norm": 4.322220325469971,
      "kl": 0.23828125,
      "learning_rate": 8.515384615384616e-07,
      "loss": -0.0114,
      "reward": 0.7220044434070587,
      "reward_std": 0.03613801300525665,
      "reward_std_all": 0.06162108667194843,
      "rewards-max/forensics_sinkhorn_reward": 0.9750935137271881,
      "rewards-max/forensics_strict_edge_reward": 0.38722455501556396,
      "rewards-max/forensics_tiou_reward": 0.9586751461029053,
      "rewards/forensics_sinkhorn_reward": 0.6232890486717224,
      "rewards/forensics_strict_edge_reward": 0.01636676024645567,
      "rewards/forensics_tiou_reward": 0.4121433347463608,
      "step": 194
    },
    {
      "completion_length": 24.9375,
      "epoch": 0.5990783410138248,
      "grad_norm": 4.384520053863525,
      "kl": 0.2958984375,
      "learning_rate": 8.507692307692307e-07,
      "loss": 0.147,
      "reward": 0.6681452393531799,
      "reward_std": 0.10909057781100273,
      "reward_std_all": 0.11797568947076797,
      "rewards-max/forensics_sinkhorn_reward": 0.9479044675827026,
      "rewards-max/forensics_strict_edge_reward": 0.1864516802597791,
      "rewards-max/forensics_tiou_reward": 0.8750803172588348,
      "rewards/forensics_sinkhorn_reward": 0.5711603164672852,
      "rewards/forensics_strict_edge_reward": 0.007979137677466497,
      "rewards/forensics_tiou_reward": 0.30319371819496155,
      "step": 195
    },
    {
      "completion_length": 23.046875,
      "epoch": 0.6021505376344086,
      "grad_norm": 5.322574138641357,
      "kl": 0.328125,
      "learning_rate": 8.499999999999999e-07,
      "loss": 0.0619,
      "reward": 0.7354776859283447,
      "reward_std": 0.06702886894345284,
      "reward_std_all": 0.06785417720675468,
      "rewards-max/forensics_sinkhorn_reward": 0.9742225706577301,
      "rewards-max/forensics_strict_edge_reward": 0.5933051109313965,
      "rewards-max/forensics_tiou_reward": 0.9607788026332855,
      "rewards/forensics_sinkhorn_reward": 0.6590907871723175,
      "rewards/forensics_strict_edge_reward": 0.02981756255030632,
      "rewards/forensics_tiou_reward": 0.36889341473579407,
      "step": 196
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.6052227342549923,
      "grad_norm": 6.168060779571533,
      "kl": 0.3583984375,
      "learning_rate": 8.492307692307692e-07,
      "loss": 0.0723,
      "reward": 0.7490061819553375,
      "reward_std": 0.07727287709712982,
      "reward_std_all": 0.07173185423016548,
      "rewards-max/forensics_sinkhorn_reward": 0.9459578692913055,
      "rewards-max/forensics_strict_edge_reward": 0.12030429900914896,
      "rewards-max/forensics_tiou_reward": 0.8596477210521698,
      "rewards/forensics_sinkhorn_reward": 0.64423006772995,
      "rewards/forensics_strict_edge_reward": 0.00501337627065368,
      "rewards/forensics_tiou_reward": 0.34107331931591034,
      "step": 197
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.6082949308755761,
      "grad_norm": 4.652475357055664,
      "kl": 0.2451171875,
      "learning_rate": 8.484615384615384e-07,
      "loss": 0.1481,
      "reward": 0.7163555026054382,
      "reward_std": 0.08285531029105186,
      "reward_std_all": 0.08543640375137329,
      "rewards-max/forensics_sinkhorn_reward": 0.9708656668663025,
      "rewards-max/forensics_strict_edge_reward": 0.2578420117497444,
      "rewards-max/forensics_tiou_reward": 0.9485470354557037,
      "rewards/forensics_sinkhorn_reward": 0.6249296367168427,
      "rewards/forensics_strict_edge_reward": 0.01718377275392413,
      "rewards/forensics_tiou_reward": 0.374243900179863,
      "step": 198
    },
    {
      "completion_length": 23.21875,
      "epoch": 0.6113671274961597,
      "grad_norm": 3.9585492610931396,
      "kl": 0.25341796875,
      "learning_rate": 8.476923076923077e-07,
      "loss": 0.0519,
      "reward": 0.6868098378181458,
      "reward_std": 0.08878003060817719,
      "reward_std_all": 0.10693655535578728,
      "rewards-max/forensics_sinkhorn_reward": 0.9264138638973236,
      "rewards-max/forensics_strict_edge_reward": 0.23881489783525467,
      "rewards-max/forensics_tiou_reward": 0.8467206358909607,
      "rewards/forensics_sinkhorn_reward": 0.5762942433357239,
      "rewards/forensics_strict_edge_reward": 0.011241802712902427,
      "rewards/forensics_tiou_reward": 0.3414372205734253,
      "step": 199
    },
    {
      "completion_length": 25.09375,
      "epoch": 0.6144393241167435,
      "grad_norm": 4.304559230804443,
      "kl": 0.23486328125,
      "learning_rate": 8.469230769230769e-07,
      "loss": 0.0965,
      "reward": 0.6816214323043823,
      "reward_std": 0.09627164155244827,
      "reward_std_all": 0.10897579044103622,
      "rewards-max/forensics_sinkhorn_reward": 0.9305247366428375,
      "rewards-max/forensics_strict_edge_reward": 0.0927206389605999,
      "rewards-max/forensics_tiou_reward": 0.848559558391571,
      "rewards/forensics_sinkhorn_reward": 0.573552131652832,
      "rewards/forensics_strict_edge_reward": 0.004594432306475937,
      "rewards/forensics_tiou_reward": 0.34200720489025116,
      "step": 200
    },
    {
      "completion_length": 24.296875,
      "epoch": 0.6175115207373272,
      "grad_norm": 4.151706695556641,
      "kl": 0.24365234375,
      "learning_rate": 8.461538461538461e-07,
      "loss": 0.0537,
      "reward": 0.7117199897766113,
      "reward_std": 0.11741578578948975,
      "reward_std_all": 0.10927249118685722,
      "rewards-max/forensics_sinkhorn_reward": 0.9783377349376678,
      "rewards-max/forensics_strict_edge_reward": 0.6468778848648071,
      "rewards-max/forensics_tiou_reward": 0.9726024270057678,
      "rewards/forensics_sinkhorn_reward": 0.6031459867954254,
      "rewards/forensics_strict_edge_reward": 0.03557094931602478,
      "rewards/forensics_tiou_reward": 0.38139453530311584,
      "step": 201
    },
    {
      "completion_length": 23.265625,
      "epoch": 0.620583717357911,
      "grad_norm": 4.743753433227539,
      "kl": 0.2373046875,
      "learning_rate": 8.453846153846154e-07,
      "loss": 0.2065,
      "reward": 0.6934046149253845,
      "reward_std": 0.1028854288160801,
      "reward_std_all": 0.10901021957397461,
      "rewards-max/forensics_sinkhorn_reward": 0.9288906455039978,
      "rewards-max/forensics_strict_edge_reward": 0.047084951773285866,
      "rewards-max/forensics_tiou_reward": 0.8316220939159393,
      "rewards/forensics_sinkhorn_reward": 0.5692854523658752,
      "rewards/forensics_strict_edge_reward": 0.0035031884908676147,
      "rewards/forensics_tiou_reward": 0.3086698353290558,
      "step": 202
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.6236559139784946,
      "grad_norm": 4.352388381958008,
      "kl": 0.22607421875,
      "learning_rate": 8.446153846153846e-07,
      "loss": 0.1646,
      "reward": 0.697880744934082,
      "reward_std": 0.101195327937603,
      "reward_std_all": 0.10616061836481094,
      "rewards-max/forensics_sinkhorn_reward": 0.8431190550327301,
      "rewards-max/forensics_strict_edge_reward": 0.014175798860378563,
      "rewards-max/forensics_tiou_reward": 0.6997508704662323,
      "rewards/forensics_sinkhorn_reward": 0.5677868723869324,
      "rewards/forensics_strict_edge_reward": 0.000773446603488992,
      "rewards/forensics_tiou_reward": 0.3051315099000931,
      "step": 203
    },
    {
      "completion_length": 24.375,
      "epoch": 0.6267281105990783,
      "grad_norm": 4.620494842529297,
      "kl": 0.20654296875,
      "learning_rate": 8.438461538461537e-07,
      "loss": 0.2779,
      "reward": 0.6760421693325043,
      "reward_std": 0.09425890445709229,
      "reward_std_all": 0.10709213465452194,
      "rewards-max/forensics_sinkhorn_reward": 0.9490085542201996,
      "rewards-max/forensics_strict_edge_reward": 0.029195838375017047,
      "rewards-max/forensics_tiou_reward": 0.9009599387645721,
      "rewards/forensics_sinkhorn_reward": 0.5549156665802002,
      "rewards/forensics_strict_edge_reward": 0.00226482213838608,
      "rewards/forensics_tiou_reward": 0.3008505403995514,
      "step": 204
    },
    {
      "completion_length": 22.34375,
      "epoch": 0.6298003072196621,
      "grad_norm": 4.199692249298096,
      "kl": 0.3291015625,
      "learning_rate": 8.43076923076923e-07,
      "loss": 0.1841,
      "reward": 0.7556807398796082,
      "reward_std": 0.05452174926176667,
      "reward_std_all": 0.05339198745787144,
      "rewards-max/forensics_sinkhorn_reward": 0.9686203598976135,
      "rewards-max/forensics_strict_edge_reward": 0.33166981488466263,
      "rewards-max/forensics_tiou_reward": 0.9347500205039978,
      "rewards/forensics_sinkhorn_reward": 0.6523364186286926,
      "rewards/forensics_strict_edge_reward": 0.029102481435984373,
      "rewards/forensics_tiou_reward": 0.4179697036743164,
      "step": 205
    },
    {
      "completion_length": 25.046875,
      "epoch": 0.6328725038402457,
      "grad_norm": 4.543668746948242,
      "kl": 0.2529296875,
      "learning_rate": 8.423076923076923e-07,
      "loss": 0.2038,
      "reward": 0.6559852063655853,
      "reward_std": 0.08086803555488586,
      "reward_std_all": 0.11011653020977974,
      "rewards-max/forensics_sinkhorn_reward": 0.9247364401817322,
      "rewards-max/forensics_strict_edge_reward": 0.2312923651188612,
      "rewards-max/forensics_tiou_reward": 0.8359769582748413,
      "rewards/forensics_sinkhorn_reward": 0.5286025106906891,
      "rewards/forensics_strict_edge_reward": 0.012473267503082752,
      "rewards/forensics_tiou_reward": 0.2883238345384598,
      "step": 206
    },
    {
      "completion_length": 25.171875,
      "epoch": 0.6359447004608295,
      "grad_norm": 5.654776573181152,
      "kl": 0.294921875,
      "learning_rate": 8.415384615384616e-07,
      "loss": 0.1318,
      "reward": 0.676884263753891,
      "reward_std": 0.07681498676538467,
      "reward_std_all": 0.1008654348552227,
      "rewards-max/forensics_sinkhorn_reward": 0.9262037575244904,
      "rewards-max/forensics_strict_edge_reward": 0.05923379406794993,
      "rewards-max/forensics_tiou_reward": 0.8467687368392944,
      "rewards/forensics_sinkhorn_reward": 0.5775780379772186,
      "rewards/forensics_strict_edge_reward": 0.0024791942023512092,
      "rewards/forensics_tiou_reward": 0.3295583128929138,
      "step": 207
    },
    {
      "completion_length": 21.921875,
      "epoch": 0.6390168970814132,
      "grad_norm": 4.596075057983398,
      "kl": 0.23046875,
      "learning_rate": 8.407692307692308e-07,
      "loss": 0.1878,
      "reward": 0.7161236107349396,
      "reward_std": 0.10161100327968597,
      "reward_std_all": 0.09667078778147697,
      "rewards-max/forensics_sinkhorn_reward": 0.9301057457923889,
      "rewards-max/forensics_strict_edge_reward": 0.12337608262896538,
      "rewards-max/forensics_tiou_reward": 0.8544177114963531,
      "rewards/forensics_sinkhorn_reward": 0.6165785193443298,
      "rewards/forensics_strict_edge_reward": 0.0068836616119369864,
      "rewards/forensics_tiou_reward": 0.33952784538269043,
      "step": 208
    },
    {
      "completion_length": 23.171875,
      "epoch": 0.642089093701997,
      "grad_norm": 3.7852470874786377,
      "kl": 0.26171875,
      "learning_rate": 8.399999999999999e-07,
      "loss": 0.036,
      "reward": 0.6659691333770752,
      "reward_std": 0.09074498899281025,
      "reward_std_all": 0.10920603200793266,
      "rewards-max/forensics_sinkhorn_reward": 0.9718693792819977,
      "rewards-max/forensics_strict_edge_reward": 0.2246939551569085,
      "rewards-max/forensics_tiou_reward": 0.951463371515274,
      "rewards/forensics_sinkhorn_reward": 0.5792961716651917,
      "rewards/forensics_strict_edge_reward": 0.00966225919273711,
      "rewards/forensics_tiou_reward": 0.34826385974884033,
      "step": 209
    },
    {
      "completion_length": 24.484375,
      "epoch": 0.6451612903225806,
      "grad_norm": 4.2971110343933105,
      "kl": 0.25048828125,
      "learning_rate": 8.392307692307692e-07,
      "loss": 0.0683,
      "reward": 0.661522388458252,
      "reward_std": 0.10337726399302483,
      "reward_std_all": 0.12093698233366013,
      "rewards-max/forensics_sinkhorn_reward": 0.9751758277416229,
      "rewards-max/forensics_strict_edge_reward": 0.2925948202610016,
      "rewards-max/forensics_tiou_reward": 0.9614460468292236,
      "rewards/forensics_sinkhorn_reward": 0.5319277346134186,
      "rewards/forensics_strict_edge_reward": 0.024368907790631056,
      "rewards/forensics_tiou_reward": 0.34674379229545593,
      "step": 210
    },
    {
      "completion_length": 26.265625,
      "epoch": 0.6482334869431644,
      "grad_norm": 4.2420830726623535,
      "kl": 0.28515625,
      "learning_rate": 8.384615384615384e-07,
      "loss": 0.2004,
      "reward": 0.6419950127601624,
      "reward_std": 0.1260933205485344,
      "reward_std_all": 0.14068523794412613,
      "rewards-max/forensics_sinkhorn_reward": 0.8942658603191376,
      "rewards-max/forensics_strict_edge_reward": 0.02857796661555767,
      "rewards-max/forensics_tiou_reward": 0.7680736780166626,
      "rewards/forensics_sinkhorn_reward": 0.5346634089946747,
      "rewards/forensics_strict_edge_reward": 0.0019478751346468925,
      "rewards/forensics_tiou_reward": 0.3008715510368347,
      "step": 211
    },
    {
      "completion_length": 23.84375,
      "epoch": 0.6513056835637481,
      "grad_norm": 4.343061923980713,
      "kl": 0.39453125,
      "learning_rate": 8.376923076923076e-07,
      "loss": 0.0936,
      "reward": 0.7134978175163269,
      "reward_std": 0.08581517543643713,
      "reward_std_all": 0.08913487568497658,
      "rewards-max/forensics_sinkhorn_reward": 0.9097926914691925,
      "rewards-max/forensics_strict_edge_reward": 0.30651140958070755,
      "rewards-max/forensics_tiou_reward": 0.8092691004276276,
      "rewards/forensics_sinkhorn_reward": 0.5842432975769043,
      "rewards/forensics_strict_edge_reward": 0.028673166874796152,
      "rewards/forensics_tiou_reward": 0.3240397572517395,
      "step": 212
    },
    {
      "completion_length": 25.90625,
      "epoch": 0.6543778801843319,
      "grad_norm": 4.2077107429504395,
      "kl": 0.21142578125,
      "learning_rate": 8.369230769230769e-07,
      "loss": 0.1319,
      "reward": 0.6855466961860657,
      "reward_std": 0.10101589560508728,
      "reward_std_all": 0.10700511187314987,
      "rewards-max/forensics_sinkhorn_reward": 0.9024963080883026,
      "rewards-max/forensics_strict_edge_reward": 0.06306707055773586,
      "rewards-max/forensics_tiou_reward": 0.7971347868442535,
      "rewards/forensics_sinkhorn_reward": 0.5595648884773254,
      "rewards/forensics_strict_edge_reward": 0.005789786548120901,
      "rewards/forensics_tiou_reward": 0.296823114156723,
      "step": 213
    },
    {
      "completion_length": 26.390625,
      "epoch": 0.6574500768049155,
      "grad_norm": 4.143103122711182,
      "kl": 0.388671875,
      "learning_rate": 8.361538461538461e-07,
      "loss": 0.0963,
      "reward": 0.6283683180809021,
      "reward_std": 0.13437284529209137,
      "reward_std_all": 0.1528773084282875,
      "rewards-max/forensics_sinkhorn_reward": 0.9324646592140198,
      "rewards-max/forensics_strict_edge_reward": 0.15584840143856127,
      "rewards-max/forensics_tiou_reward": 0.8377847373485565,
      "rewards/forensics_sinkhorn_reward": 0.5283564031124115,
      "rewards/forensics_strict_edge_reward": 0.012173183316917857,
      "rewards/forensics_tiou_reward": 0.29880134761333466,
      "step": 214
    },
    {
      "completion_length": 26.109375,
      "epoch": 0.6605222734254992,
      "grad_norm": 5.375400543212891,
      "kl": 0.40625,
      "learning_rate": 8.353846153846154e-07,
      "loss": 0.0869,
      "reward": 0.6877118051052094,
      "reward_std": 0.12028838694095612,
      "reward_std_all": 0.12393098324537277,
      "rewards-max/forensics_sinkhorn_reward": 0.9726933240890503,
      "rewards-max/forensics_strict_edge_reward": 0.322879821062088,
      "rewards-max/forensics_tiou_reward": 0.9565737545490265,
      "rewards/forensics_sinkhorn_reward": 0.5910676121711731,
      "rewards/forensics_strict_edge_reward": 0.022960884496569633,
      "rewards/forensics_tiou_reward": 0.4532569348812103,
      "step": 215
    },
    {
      "completion_length": 24.265625,
      "epoch": 0.663594470046083,
      "grad_norm": 4.267817497253418,
      "kl": 0.26953125,
      "learning_rate": 8.346153846153846e-07,
      "loss": -0.0028,
      "reward": 0.7105403542518616,
      "reward_std": 0.07062531635165215,
      "reward_std_all": 0.08178874850273132,
      "rewards-max/forensics_sinkhorn_reward": 0.9582868814468384,
      "rewards-max/forensics_strict_edge_reward": 0.30914753302931786,
      "rewards-max/forensics_tiou_reward": 0.9239074289798737,
      "rewards/forensics_sinkhorn_reward": 0.6314549446105957,
      "rewards/forensics_strict_edge_reward": 0.029038281878456473,
      "rewards/forensics_tiou_reward": 0.40172484517097473,
      "step": 216
    },
    {
      "completion_length": 24.265625,
      "epoch": 0.6666666666666666,
      "grad_norm": 5.420193195343018,
      "kl": 0.2587890625,
      "learning_rate": 8.338461538461537e-07,
      "loss": 0.0817,
      "reward": 0.6286946833133698,
      "reward_std": 0.12427277117967606,
      "reward_std_all": 0.14730150252580643,
      "rewards-max/forensics_sinkhorn_reward": 0.9150215685367584,
      "rewards-max/forensics_strict_edge_reward": 0.10739240422844887,
      "rewards-max/forensics_tiou_reward": 0.8174574077129364,
      "rewards/forensics_sinkhorn_reward": 0.487652987241745,
      "rewards/forensics_strict_edge_reward": 0.008002638584002852,
      "rewards/forensics_tiou_reward": 0.2531229555606842,
      "step": 217
    },
    {
      "completion_length": 26.34375,
      "epoch": 0.6697388632872504,
      "grad_norm": 4.3831281661987305,
      "kl": 0.25830078125,
      "learning_rate": 8.33076923076923e-07,
      "loss": 0.1506,
      "reward": 0.7258153855800629,
      "reward_std": 0.08869179338216782,
      "reward_std_all": 0.08804239705204964,
      "rewards-max/forensics_sinkhorn_reward": 0.9752175211906433,
      "rewards-max/forensics_strict_edge_reward": 0.5581407397985458,
      "rewards-max/forensics_tiou_reward": 0.9570575952529907,
      "rewards/forensics_sinkhorn_reward": 0.6343160271644592,
      "rewards/forensics_strict_edge_reward": 0.032064951956272125,
      "rewards/forensics_tiou_reward": 0.4560393989086151,
      "step": 218
    },
    {
      "completion_length": 25.40625,
      "epoch": 0.6728110599078341,
      "grad_norm": 4.785116195678711,
      "kl": 0.23291015625,
      "learning_rate": 8.323076923076922e-07,
      "loss": 0.2992,
      "reward": 0.6967445611953735,
      "reward_std": 0.07833480462431908,
      "reward_std_all": 0.09259627759456635,
      "rewards-max/forensics_sinkhorn_reward": 0.900131106376648,
      "rewards-max/forensics_strict_edge_reward": 0.35889188945293427,
      "rewards-max/forensics_tiou_reward": 0.7980543673038483,
      "rewards/forensics_sinkhorn_reward": 0.5600333213806152,
      "rewards/forensics_strict_edge_reward": 0.023680374957621098,
      "rewards/forensics_tiou_reward": 0.3470681458711624,
      "step": 219
    },
    {
      "completion_length": 26.640625,
      "epoch": 0.6758832565284179,
      "grad_norm": 4.5994181632995605,
      "kl": 0.2490234375,
      "learning_rate": 8.315384615384616e-07,
      "loss": 0.1052,
      "reward": 0.6660078763961792,
      "reward_std": 0.10120843723416328,
      "reward_std_all": 0.1159728392958641,
      "rewards-max/forensics_sinkhorn_reward": 0.9393519163131714,
      "rewards-max/forensics_strict_edge_reward": 0.08968089334666729,
      "rewards-max/forensics_tiou_reward": 0.8784340918064117,
      "rewards/forensics_sinkhorn_reward": 0.5299925804138184,
      "rewards/forensics_strict_edge_reward": 0.005834830983076245,
      "rewards/forensics_tiou_reward": 0.2993753254413605,
      "step": 220
    },
    {
      "completion_length": 23.28125,
      "epoch": 0.6789554531490015,
      "grad_norm": 4.215561866760254,
      "kl": 0.2802734375,
      "learning_rate": 8.307692307692308e-07,
      "loss": 0.0838,
      "reward": 0.7408203482627869,
      "reward_std": 0.06106901727616787,
      "reward_std_all": 0.06342586502432823,
      "rewards-max/forensics_sinkhorn_reward": 0.9513482749462128,
      "rewards-max/forensics_strict_edge_reward": 0.27228695154190063,
      "rewards-max/forensics_tiou_reward": 0.9134089946746826,
      "rewards/forensics_sinkhorn_reward": 0.6836889386177063,
      "rewards/forensics_strict_edge_reward": 0.02311142859980464,
      "rewards/forensics_tiou_reward": 0.42229095101356506,
      "step": 221
    },
    {
      "completion_length": 22.421875,
      "epoch": 0.6820276497695853,
      "grad_norm": 6.1030192375183105,
      "kl": 0.3955078125,
      "learning_rate": 8.299999999999999e-07,
      "loss": 0.1326,
      "reward": 0.7176290154457092,
      "reward_std": 0.07858916744589806,
      "reward_std_all": 0.08091725409030914,
      "rewards-max/forensics_sinkhorn_reward": 0.9546390175819397,
      "rewards-max/forensics_strict_edge_reward": 0.10741388984024525,
      "rewards-max/forensics_tiou_reward": 0.9052616953849792,
      "rewards/forensics_sinkhorn_reward": 0.6081939339637756,
      "rewards/forensics_strict_edge_reward": 0.008773140842095017,
      "rewards/forensics_tiou_reward": 0.36236144602298737,
      "step": 222
    },
    {
      "completion_length": 25.015625,
      "epoch": 0.685099846390169,
      "grad_norm": 4.662515640258789,
      "kl": 0.2197265625,
      "learning_rate": 8.292307692307692e-07,
      "loss": 0.2279,
      "reward": 0.6905494332313538,
      "reward_std": 0.10712077841162682,
      "reward_std_all": 0.11026717349886894,
      "rewards-max/forensics_sinkhorn_reward": 0.8987418711185455,
      "rewards-max/forensics_strict_edge_reward": 0.01853106648195535,
      "rewards-max/forensics_tiou_reward": 0.7586184740066528,
      "rewards/forensics_sinkhorn_reward": 0.5765517950057983,
      "rewards/forensics_strict_edge_reward": 0.0013284688902786002,
      "rewards/forensics_tiou_reward": 0.274773046374321,
      "step": 223
    },
    {
      "completion_length": 25.359375,
      "epoch": 0.6881720430107527,
      "grad_norm": 4.840342998504639,
      "kl": 0.28662109375,
      "learning_rate": 8.284615384615384e-07,
      "loss": 0.0642,
      "reward": 0.7176035940647125,
      "reward_std": 0.08595551177859306,
      "reward_std_all": 0.0856085903942585,
      "rewards-max/forensics_sinkhorn_reward": 0.9543852210044861,
      "rewards-max/forensics_strict_edge_reward": 0.06335241603665054,
      "rewards-max/forensics_tiou_reward": 0.9058597385883331,
      "rewards/forensics_sinkhorn_reward": 0.6263790130615234,
      "rewards/forensics_strict_edge_reward": 0.00539549704080855,
      "rewards/forensics_tiou_reward": 0.4071870744228363,
      "step": 224
    },
    {
      "completion_length": 21.890625,
      "epoch": 0.6912442396313364,
      "grad_norm": 5.452155590057373,
      "kl": 0.25634765625,
      "learning_rate": 8.276923076923077e-07,
      "loss": -0.0108,
      "reward": 0.6756174266338348,
      "reward_std": 0.11423493549227715,
      "reward_std_all": 0.1221751719713211,
      "rewards-max/forensics_sinkhorn_reward": 0.9695922136306763,
      "rewards-max/forensics_strict_edge_reward": 0.2594909444451332,
      "rewards-max/forensics_tiou_reward": 0.9358856976032257,
      "rewards/forensics_sinkhorn_reward": 0.5954951345920563,
      "rewards/forensics_strict_edge_reward": 0.015189679805189371,
      "rewards/forensics_tiou_reward": 0.3522147387266159,
      "step": 225
    },
    {
      "completion_length": 23.34375,
      "epoch": 0.6943164362519201,
      "grad_norm": 4.587284088134766,
      "kl": 0.3154296875,
      "learning_rate": 8.269230769230768e-07,
      "loss": 0.1283,
      "reward": 0.7058395445346832,
      "reward_std": 0.10493636876344681,
      "reward_std_all": 0.10534864664077759,
      "rewards-max/forensics_sinkhorn_reward": 0.945851743221283,
      "rewards-max/forensics_strict_edge_reward": 0.18995143845677376,
      "rewards-max/forensics_tiou_reward": 0.8670287132263184,
      "rewards/forensics_sinkhorn_reward": 0.6291926503181458,
      "rewards/forensics_strict_edge_reward": 0.00946971945813857,
      "rewards/forensics_tiou_reward": 0.3755372315645218,
      "step": 226
    },
    {
      "completion_length": 22.09375,
      "epoch": 0.6973886328725039,
      "grad_norm": 4.567240238189697,
      "kl": 0.322265625,
      "learning_rate": 8.261538461538461e-07,
      "loss": 0.0817,
      "reward": 0.7480264902114868,
      "reward_std": 0.05021832324564457,
      "reward_std_all": 0.05471273139119148,
      "rewards-max/forensics_sinkhorn_reward": 0.9790293872356415,
      "rewards-max/forensics_strict_edge_reward": 0.3187941014766693,
      "rewards-max/forensics_tiou_reward": 0.9727941453456879,
      "rewards/forensics_sinkhorn_reward": 0.653082937002182,
      "rewards/forensics_strict_edge_reward": 0.01643801061436534,
      "rewards/forensics_tiou_reward": 0.38227422535419464,
      "step": 227
    },
    {
      "completion_length": 23.890625,
      "epoch": 0.7004608294930875,
      "grad_norm": 4.207234859466553,
      "kl": 0.28271484375,
      "learning_rate": 8.253846153846154e-07,
      "loss": 0.0758,
      "reward": 0.7156557142734528,
      "reward_std": 0.0805967953056097,
      "reward_std_all": 0.08638099953532219,
      "rewards-max/forensics_sinkhorn_reward": 0.9218152463436127,
      "rewards-max/forensics_strict_edge_reward": 0.1312341990415007,
      "rewards-max/forensics_tiou_reward": 0.8438805937767029,
      "rewards/forensics_sinkhorn_reward": 0.617710530757904,
      "rewards/forensics_strict_edge_reward": 0.00590831802401226,
      "rewards/forensics_tiou_reward": 0.3425441235303879,
      "step": 228
    },
    {
      "completion_length": 22.796875,
      "epoch": 0.7035330261136713,
      "grad_norm": 5.110813617706299,
      "kl": 0.330078125,
      "learning_rate": 8.246153846153846e-07,
      "loss": 0.0478,
      "reward": 0.7342382967472076,
      "reward_std": 0.07008351292461157,
      "reward_std_all": 0.07225223397836089,
      "rewards-max/forensics_sinkhorn_reward": 0.9600555598735809,
      "rewards-max/forensics_strict_edge_reward": 0.26941629871726036,
      "rewards-max/forensics_tiou_reward": 0.91963991522789,
      "rewards/forensics_sinkhorn_reward": 0.6329504549503326,
      "rewards/forensics_strict_edge_reward": 0.018141143955290318,
      "rewards/forensics_tiou_reward": 0.3930020183324814,
      "step": 229
    },
    {
      "completion_length": 22.84375,
      "epoch": 0.706605222734255,
      "grad_norm": 4.592637538909912,
      "kl": 0.3330078125,
      "learning_rate": 8.238461538461538e-07,
      "loss": 0.028,
      "reward": 0.7042916715145111,
      "reward_std": 0.07674036920070648,
      "reward_std_all": 0.08740584924817085,
      "rewards-max/forensics_sinkhorn_reward": 0.960018515586853,
      "rewards-max/forensics_strict_edge_reward": 0.24460745602846146,
      "rewards-max/forensics_tiou_reward": 0.9201466739177704,
      "rewards/forensics_sinkhorn_reward": 0.5853926241397858,
      "rewards/forensics_strict_edge_reward": 0.012556290719658136,
      "rewards/forensics_tiou_reward": 0.34499435126781464,
      "step": 230
    },
    {
      "completion_length": 21.84375,
      "epoch": 0.7096774193548387,
      "grad_norm": 4.331064701080322,
      "kl": 0.21875,
      "learning_rate": 8.23076923076923e-07,
      "loss": 0.0949,
      "reward": 0.7605343163013458,
      "reward_std": 0.0384628064930439,
      "reward_std_all": 0.04461198206990957,
      "rewards-max/forensics_sinkhorn_reward": 0.9683027863502502,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9513797461986542,
      "rewards/forensics_sinkhorn_reward": 0.6755960285663605,
      "rewards/forensics_strict_edge_reward": 0.03481096401810646,
      "rewards/forensics_tiou_reward": 0.36690396070480347,
      "step": 231
    },
    {
      "completion_length": 24.125,
      "epoch": 0.7127496159754224,
      "grad_norm": 4.790604114532471,
      "kl": 0.3056640625,
      "learning_rate": 8.223076923076922e-07,
      "loss": 0.0137,
      "reward": 0.7596235573291779,
      "reward_std": 0.043369658291339874,
      "reward_std_all": 0.04470635578036308,
      "rewards-max/forensics_sinkhorn_reward": 0.9653282165527344,
      "rewards-max/forensics_strict_edge_reward": 0.2710696943104267,
      "rewards-max/forensics_tiou_reward": 0.944340705871582,
      "rewards/forensics_sinkhorn_reward": 0.6957261264324188,
      "rewards/forensics_strict_edge_reward": 0.017636148259043694,
      "rewards/forensics_tiou_reward": 0.43202103674411774,
      "step": 232
    },
    {
      "completion_length": 23.40625,
      "epoch": 0.7158218125960062,
      "grad_norm": 4.3668084144592285,
      "kl": 0.2666015625,
      "learning_rate": 8.215384615384616e-07,
      "loss": -0.0148,
      "reward": 0.6952108442783356,
      "reward_std": 0.10812296345829964,
      "reward_std_all": 0.10932163894176483,
      "rewards-max/forensics_sinkhorn_reward": 0.8770663142204285,
      "rewards-max/forensics_strict_edge_reward": 0.09144637294230051,
      "rewards-max/forensics_tiou_reward": 0.7689408659934998,
      "rewards/forensics_sinkhorn_reward": 0.5865548253059387,
      "rewards/forensics_strict_edge_reward": 0.007675492360249336,
      "rewards/forensics_tiou_reward": 0.3529162108898163,
      "step": 233
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.7188940092165899,
      "grad_norm": 4.350690841674805,
      "kl": 0.24365234375,
      "learning_rate": 8.207692307692308e-07,
      "loss": 0.0832,
      "reward": 0.7051649391651154,
      "reward_std": 0.10926743969321251,
      "reward_std_all": 0.10528849810361862,
      "rewards-max/forensics_sinkhorn_reward": 0.9264884293079376,
      "rewards-max/forensics_strict_edge_reward": 0.16862228233367205,
      "rewards-max/forensics_tiou_reward": 0.831115335226059,
      "rewards/forensics_sinkhorn_reward": 0.6067254841327667,
      "rewards/forensics_strict_edge_reward": 0.007434824015945196,
      "rewards/forensics_tiou_reward": 0.3479065001010895,
      "step": 234
    },
    {
      "completion_length": 23.734375,
      "epoch": 0.7219662058371735,
      "grad_norm": 4.6169819831848145,
      "kl": 0.25390625,
      "learning_rate": 8.199999999999999e-07,
      "loss": 0.248,
      "reward": 0.7220579385757446,
      "reward_std": 0.09354076534509659,
      "reward_std_all": 0.0901390090584755,
      "rewards-max/forensics_sinkhorn_reward": 0.9600785374641418,
      "rewards-max/forensics_strict_edge_reward": 0.22498222440481186,
      "rewards-max/forensics_tiou_reward": 0.9150694906711578,
      "rewards/forensics_sinkhorn_reward": 0.5964250564575195,
      "rewards/forensics_strict_edge_reward": 0.012224042322486639,
      "rewards/forensics_tiou_reward": 0.3928026705980301,
      "step": 235
    },
    {
      "completion_length": 22.859375,
      "epoch": 0.7250384024577573,
      "grad_norm": 4.825022220611572,
      "kl": 0.30078125,
      "learning_rate": 8.192307692307692e-07,
      "loss": 0.1369,
      "reward": 0.7585512101650238,
      "reward_std": 0.055955372750759125,
      "reward_std_all": 0.05384375900030136,
      "rewards-max/forensics_sinkhorn_reward": 0.9720230996608734,
      "rewards-max/forensics_strict_edge_reward": 0.18652867525815964,
      "rewards-max/forensics_tiou_reward": 0.9517425000667572,
      "rewards/forensics_sinkhorn_reward": 0.6575173735618591,
      "rewards/forensics_strict_edge_reward": 0.013580030761659145,
      "rewards/forensics_tiou_reward": 0.4241022318601608,
      "step": 236
    },
    {
      "completion_length": 23.78125,
      "epoch": 0.728110599078341,
      "grad_norm": 5.7232184410095215,
      "kl": 0.3662109375,
      "learning_rate": 8.184615384615384e-07,
      "loss": 0.1521,
      "reward": 0.7042038440704346,
      "reward_std": 0.07116685807704926,
      "reward_std_all": 0.08597035706043243,
      "rewards-max/forensics_sinkhorn_reward": 0.9563885927200317,
      "rewards-max/forensics_strict_edge_reward": 0.1496355626732111,
      "rewards-max/forensics_tiou_reward": 0.9174775183200836,
      "rewards/forensics_sinkhorn_reward": 0.5866129994392395,
      "rewards/forensics_strict_edge_reward": 0.007770203927066177,
      "rewards/forensics_tiou_reward": 0.40525734424591064,
      "step": 237
    },
    {
      "completion_length": 23.09375,
      "epoch": 0.7311827956989247,
      "grad_norm": 4.458467960357666,
      "kl": 0.2333984375,
      "learning_rate": 8.176923076923077e-07,
      "loss": 0.1064,
      "reward": 0.7798974812030792,
      "reward_std": 0.03275397326797247,
      "reward_std_all": 0.031992170959711075,
      "rewards-max/forensics_sinkhorn_reward": 0.9459855854511261,
      "rewards-max/forensics_strict_edge_reward": 0.1447845734655857,
      "rewards-max/forensics_tiou_reward": 0.8839606642723083,
      "rewards/forensics_sinkhorn_reward": 0.7178933024406433,
      "rewards/forensics_strict_edge_reward": 0.011297094402834773,
      "rewards/forensics_tiou_reward": 0.4411502480506897,
      "step": 238
    },
    {
      "completion_length": 21.875,
      "epoch": 0.7342549923195084,
      "grad_norm": 6.149790287017822,
      "kl": 0.3154296875,
      "learning_rate": 8.169230769230769e-07,
      "loss": 0.0721,
      "reward": 0.7125893831253052,
      "reward_std": 0.08191617578268051,
      "reward_std_all": 0.08494698256254196,
      "rewards-max/forensics_sinkhorn_reward": 0.9384383261203766,
      "rewards-max/forensics_strict_edge_reward": 0.2502056583762169,
      "rewards-max/forensics_tiou_reward": 0.8724214136600494,
      "rewards/forensics_sinkhorn_reward": 0.6188581585884094,
      "rewards/forensics_strict_edge_reward": 0.011412408901378512,
      "rewards/forensics_tiou_reward": 0.3349562734365463,
      "step": 239
    },
    {
      "completion_length": 24.5625,
      "epoch": 0.7373271889400922,
      "grad_norm": 4.666930675506592,
      "kl": 0.271484375,
      "learning_rate": 8.16153846153846e-07,
      "loss": 0.1981,
      "reward": 0.7626184821128845,
      "reward_std": 0.041711825877428055,
      "reward_std_all": 0.04417508654296398,
      "rewards-max/forensics_sinkhorn_reward": 0.9112662672996521,
      "rewards-max/forensics_strict_edge_reward": 0.5003738112864085,
      "rewards-max/forensics_tiou_reward": 0.8355414271354675,
      "rewards/forensics_sinkhorn_reward": 0.6345337927341461,
      "rewards/forensics_strict_edge_reward": 0.041629644238128094,
      "rewards/forensics_tiou_reward": 0.42351725697517395,
      "step": 240
    },
    {
      "completion_length": 22.65625,
      "epoch": 0.7403993855606759,
      "grad_norm": 4.464959144592285,
      "kl": 0.2626953125,
      "learning_rate": 8.153846153846154e-07,
      "loss": 0.1264,
      "reward": 0.7224869430065155,
      "reward_std": 0.08526600524783134,
      "reward_std_all": 0.08909592777490616,
      "rewards-max/forensics_sinkhorn_reward": 0.9522075951099396,
      "rewards-max/forensics_strict_edge_reward": 0.40798234194517136,
      "rewards-max/forensics_tiou_reward": 0.9001908302307129,
      "rewards/forensics_sinkhorn_reward": 0.6091552674770355,
      "rewards/forensics_strict_edge_reward": 0.031483185943216085,
      "rewards/forensics_tiou_reward": 0.40090590715408325,
      "step": 241
    },
    {
      "completion_length": 24.796875,
      "epoch": 0.7434715821812596,
      "grad_norm": 6.291354656219482,
      "kl": 0.4833984375,
      "learning_rate": 8.146153846153846e-07,
      "loss": 0.0326,
      "reward": 0.6947084963321686,
      "reward_std": 0.07532565109431744,
      "reward_std_all": 0.09393961727619171,
      "rewards-max/forensics_sinkhorn_reward": 0.9223508834838867,
      "rewards-max/forensics_strict_edge_reward": 0.2823946680582594,
      "rewards-max/forensics_tiou_reward": 0.8214554786682129,
      "rewards/forensics_sinkhorn_reward": 0.5821596384048462,
      "rewards/forensics_strict_edge_reward": 0.016621240458334796,
      "rewards/forensics_tiou_reward": 0.33685946464538574,
      "step": 242
    },
    {
      "completion_length": 24.34375,
      "epoch": 0.7465437788018433,
      "grad_norm": 5.15949821472168,
      "kl": 0.322265625,
      "learning_rate": 8.138461538461538e-07,
      "loss": 0.1746,
      "reward": 0.7332266867160797,
      "reward_std": 0.060097819194197655,
      "reward_std_all": 0.06690996885299683,
      "rewards-max/forensics_sinkhorn_reward": 0.8793335258960724,
      "rewards-max/forensics_strict_edge_reward": 0.16043827682733536,
      "rewards-max/forensics_tiou_reward": 0.7404599189758301,
      "rewards/forensics_sinkhorn_reward": 0.6099964678287506,
      "rewards/forensics_strict_edge_reward": 0.011005617212504148,
      "rewards/forensics_tiou_reward": 0.3594474494457245,
      "step": 243
    },
    {
      "completion_length": 26.140625,
      "epoch": 0.7496159754224271,
      "grad_norm": 4.852986812591553,
      "kl": 0.4130859375,
      "learning_rate": 8.13076923076923e-07,
      "loss": 0.1693,
      "reward": 0.6504342257976532,
      "reward_std": 0.09778688848018646,
      "reward_std_all": 0.12376628071069717,
      "rewards-max/forensics_sinkhorn_reward": 0.9180717766284943,
      "rewards-max/forensics_strict_edge_reward": 0.10898687690496445,
      "rewards-max/forensics_tiou_reward": 0.8446061313152313,
      "rewards/forensics_sinkhorn_reward": 0.5354479849338531,
      "rewards/forensics_strict_edge_reward": 0.004614614881575108,
      "rewards/forensics_tiou_reward": 0.2876840829849243,
      "step": 244
    },
    {
      "completion_length": 24.640625,
      "epoch": 0.7526881720430108,
      "grad_norm": 4.749887943267822,
      "kl": 0.306640625,
      "learning_rate": 8.123076923076922e-07,
      "loss": 0.0898,
      "reward": 0.7367806732654572,
      "reward_std": 0.09963899478316307,
      "reward_std_all": 0.09060250967741013,
      "rewards-max/forensics_sinkhorn_reward": 0.9730114340782166,
      "rewards-max/forensics_strict_edge_reward": 0.5161091648042202,
      "rewards-max/forensics_tiou_reward": 0.9474890828132629,
      "rewards/forensics_sinkhorn_reward": 0.6545900106430054,
      "rewards/forensics_strict_edge_reward": 0.022105362615548074,
      "rewards/forensics_tiou_reward": 0.39097408950328827,
      "step": 245
    },
    {
      "completion_length": 22.234375,
      "epoch": 0.7557603686635944,
      "grad_norm": 6.510045528411865,
      "kl": 0.29296875,
      "learning_rate": 8.115384615384615e-07,
      "loss": 0.0824,
      "reward": 0.7100178897380829,
      "reward_std": 0.04658044874668121,
      "reward_std_all": 0.06954165548086166,
      "rewards-max/forensics_sinkhorn_reward": 0.9058311283588409,
      "rewards-max/forensics_strict_edge_reward": 0.4483126401901245,
      "rewards-max/forensics_tiou_reward": 0.809792548418045,
      "rewards/forensics_sinkhorn_reward": 0.630933403968811,
      "rewards/forensics_strict_edge_reward": 0.018794549629092216,
      "rewards/forensics_tiou_reward": 0.341833233833313,
      "step": 246
    },
    {
      "completion_length": 23.609375,
      "epoch": 0.7588325652841782,
      "grad_norm": 4.667747497558594,
      "kl": 0.2978515625,
      "learning_rate": 8.107692307692308e-07,
      "loss": 0.1317,
      "reward": 0.6937180459499359,
      "reward_std": 0.0995292104780674,
      "reward_std_all": 0.11139707639813423,
      "rewards-max/forensics_sinkhorn_reward": 0.9627657830715179,
      "rewards-max/forensics_strict_edge_reward": 0.23035996034741402,
      "rewards-max/forensics_tiou_reward": 0.9321535527706146,
      "rewards/forensics_sinkhorn_reward": 0.6065958142280579,
      "rewards/forensics_strict_edge_reward": 0.014627901371568441,
      "rewards/forensics_tiou_reward": 0.40975508093833923,
      "step": 247
    },
    {
      "completion_length": 26.34375,
      "epoch": 0.7619047619047619,
      "grad_norm": 4.528666019439697,
      "kl": 0.255859375,
      "learning_rate": 8.1e-07,
      "loss": 0.137,
      "reward": 0.7280974686145782,
      "reward_std": 0.057903626933693886,
      "reward_std_all": 0.06908176094293594,
      "rewards-max/forensics_sinkhorn_reward": 0.9099045097827911,
      "rewards-max/forensics_strict_edge_reward": 0.11648056376725435,
      "rewards-max/forensics_tiou_reward": 0.8084632754325867,
      "rewards/forensics_sinkhorn_reward": 0.5955948233604431,
      "rewards/forensics_strict_edge_reward": 0.005235614895354956,
      "rewards/forensics_tiou_reward": 0.37825655937194824,
      "step": 248
    },
    {
      "completion_length": 22.96875,
      "epoch": 0.7649769585253456,
      "grad_norm": 5.701565742492676,
      "kl": 0.34375,
      "learning_rate": 8.092307692307692e-07,
      "loss": 0.001,
      "reward": 0.6896715760231018,
      "reward_std": 0.05630643106997013,
      "reward_std_all": 0.08351875469088554,
      "rewards-max/forensics_sinkhorn_reward": 0.979880303144455,
      "rewards-max/forensics_strict_edge_reward": 0.22491609677672386,
      "rewards-max/forensics_tiou_reward": 0.9701422154903412,
      "rewards/forensics_sinkhorn_reward": 0.6089339256286621,
      "rewards/forensics_strict_edge_reward": 0.0153856105171144,
      "rewards/forensics_tiou_reward": 0.32517196983098984,
      "step": 249
    },
    {
      "completion_length": 24.328125,
      "epoch": 0.7680491551459293,
      "grad_norm": 5.5012407302856445,
      "kl": 0.3173828125,
      "learning_rate": 8.084615384615384e-07,
      "loss": 0.2193,
      "reward": 0.6917845606803894,
      "reward_std": 0.11666307970881462,
      "reward_std_all": 0.11517397686839104,
      "rewards-max/forensics_sinkhorn_reward": 0.9221674799919128,
      "rewards-max/forensics_strict_edge_reward": 0.13327224925160408,
      "rewards-max/forensics_tiou_reward": 0.8326038420200348,
      "rewards/forensics_sinkhorn_reward": 0.5396780371665955,
      "rewards/forensics_strict_edge_reward": 0.0090010492131114,
      "rewards/forensics_tiou_reward": 0.3298727869987488,
      "step": 250
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.7711213517665131,
      "grad_norm": 4.944151401519775,
      "kl": 0.3046875,
      "learning_rate": 8.076923076923077e-07,
      "loss": 0.0499,
      "reward": 0.7299302220344543,
      "reward_std": 0.06947961077094078,
      "reward_std_all": 0.07267212495207787,
      "rewards-max/forensics_sinkhorn_reward": 0.9580119252204895,
      "rewards-max/forensics_strict_edge_reward": 0.03862867318093777,
      "rewards-max/forensics_tiou_reward": 0.8995776474475861,
      "rewards/forensics_sinkhorn_reward": 0.6213765144348145,
      "rewards/forensics_strict_edge_reward": 0.0033152158721350133,
      "rewards/forensics_tiou_reward": 0.3882412612438202,
      "step": 251
    },
    {
      "completion_length": 22.53125,
      "epoch": 0.7741935483870968,
      "grad_norm": 5.69082498550415,
      "kl": 0.33154296875,
      "learning_rate": 8.069230769230769e-07,
      "loss": 0.1102,
      "reward": 0.7072827517986298,
      "reward_std": 0.0843629539012909,
      "reward_std_all": 0.09422048181295395,
      "rewards-max/forensics_sinkhorn_reward": 0.9321160614490509,
      "rewards-max/forensics_strict_edge_reward": 0.18668297422118485,
      "rewards-max/forensics_tiou_reward": 0.8621182441711426,
      "rewards/forensics_sinkhorn_reward": 0.6175268888473511,
      "rewards/forensics_strict_edge_reward": 0.00953321493580006,
      "rewards/forensics_tiou_reward": 0.3535712659358978,
      "step": 252
    },
    {
      "completion_length": 24.09375,
      "epoch": 0.7772657450076805,
      "grad_norm": 4.215189456939697,
      "kl": 0.318359375,
      "learning_rate": 8.06153846153846e-07,
      "loss": 0.0749,
      "reward": 0.7320781052112579,
      "reward_std": 0.04084005206823349,
      "reward_std_all": 0.056491222232580185,
      "rewards-max/forensics_sinkhorn_reward": 0.9526897966861725,
      "rewards-max/forensics_strict_edge_reward": 0.11468159779906273,
      "rewards-max/forensics_tiou_reward": 0.9163716435432434,
      "rewards/forensics_sinkhorn_reward": 0.6397728323936462,
      "rewards/forensics_strict_edge_reward": 0.009513524593785405,
      "rewards/forensics_tiou_reward": 0.3746362030506134,
      "step": 253
    },
    {
      "completion_length": 24.609375,
      "epoch": 0.7803379416282642,
      "grad_norm": 5.315303325653076,
      "kl": 0.3603515625,
      "learning_rate": 8.053846153846154e-07,
      "loss": 0.1529,
      "reward": 0.7344086170196533,
      "reward_std": 0.07625827938318253,
      "reward_std_all": 0.07583452016115189,
      "rewards-max/forensics_sinkhorn_reward": 0.8665640950202942,
      "rewards-max/forensics_strict_edge_reward": 0.35130931437015533,
      "rewards-max/forensics_tiou_reward": 0.7950733602046967,
      "rewards/forensics_sinkhorn_reward": 0.6227648258209229,
      "rewards/forensics_strict_edge_reward": 0.020094303879886866,
      "rewards/forensics_tiou_reward": 0.37785397469997406,
      "step": 254
    },
    {
      "completion_length": 26.078125,
      "epoch": 0.783410138248848,
      "grad_norm": 4.278814792633057,
      "kl": 0.26708984375,
      "learning_rate": 8.046153846153846e-07,
      "loss": 0.1207,
      "reward": 0.734731525182724,
      "reward_std": 0.05930766370147467,
      "reward_std_all": 0.06741486862301826,
      "rewards-max/forensics_sinkhorn_reward": 0.9512946605682373,
      "rewards-max/forensics_strict_edge_reward": 0.14381973259150982,
      "rewards-max/forensics_tiou_reward": 0.9038502275943756,
      "rewards/forensics_sinkhorn_reward": 0.6370856463909149,
      "rewards/forensics_strict_edge_reward": 0.006690519861876965,
      "rewards/forensics_tiou_reward": 0.39767342805862427,
      "step": 255
    },
    {
      "completion_length": 22.078125,
      "epoch": 0.7864823348694316,
      "grad_norm": 5.659913063049316,
      "kl": 0.3291015625,
      "learning_rate": 8.038461538461538e-07,
      "loss": 0.1641,
      "reward": 0.6747645139694214,
      "reward_std": 0.09338512644171715,
      "reward_std_all": 0.11113736033439636,
      "rewards-max/forensics_sinkhorn_reward": 0.9159975647926331,
      "rewards-max/forensics_strict_edge_reward": 0.11701045278459787,
      "rewards-max/forensics_tiou_reward": 0.8166438937187195,
      "rewards/forensics_sinkhorn_reward": 0.5791028439998627,
      "rewards/forensics_strict_edge_reward": 0.008397518467972986,
      "rewards/forensics_tiou_reward": 0.2770642787218094,
      "step": 256
    },
    {
      "completion_length": 25.34375,
      "epoch": 0.7895545314900153,
      "grad_norm": 5.61916971206665,
      "kl": 0.306640625,
      "learning_rate": 8.03076923076923e-07,
      "loss": 0.1,
      "reward": 0.7261924147605896,
      "reward_std": 0.0669005997478962,
      "reward_std_all": 0.07518230751156807,
      "rewards-max/forensics_sinkhorn_reward": 0.894579291343689,
      "rewards-max/forensics_strict_edge_reward": 0.14325733855366707,
      "rewards-max/forensics_tiou_reward": 0.7780874967575073,
      "rewards/forensics_sinkhorn_reward": 0.5832713544368744,
      "rewards/forensics_strict_edge_reward": 0.01139426208101213,
      "rewards/forensics_tiou_reward": 0.38341520726680756,
      "step": 257
    },
    {
      "completion_length": 23.296875,
      "epoch": 0.7926267281105991,
      "grad_norm": 4.249839782714844,
      "kl": 0.2607421875,
      "learning_rate": 8.023076923076922e-07,
      "loss": 0.1256,
      "reward": 0.6871839761734009,
      "reward_std": 0.09109894186258316,
      "reward_std_all": 0.10317542403936386,
      "rewards-max/forensics_sinkhorn_reward": 0.9331545829772949,
      "rewards-max/forensics_strict_edge_reward": 0.2048119530081749,
      "rewards-max/forensics_tiou_reward": 0.8482131659984589,
      "rewards/forensics_sinkhorn_reward": 0.5852152705192566,
      "rewards/forensics_strict_edge_reward": 0.013760003726929426,
      "rewards/forensics_tiou_reward": 0.3569076955318451,
      "step": 258
    },
    {
      "completion_length": 25.515625,
      "epoch": 0.7956989247311828,
      "grad_norm": 4.19671630859375,
      "kl": 0.3271484375,
      "learning_rate": 8.015384615384615e-07,
      "loss": 0.1494,
      "reward": 0.7038310766220093,
      "reward_std": 0.11751588247716427,
      "reward_std_all": 0.11203005537390709,
      "rewards-max/forensics_sinkhorn_reward": 0.9667329788208008,
      "rewards-max/forensics_strict_edge_reward": 0.09266299195587635,
      "rewards-max/forensics_tiou_reward": 0.9222002327442169,
      "rewards/forensics_sinkhorn_reward": 0.5975852310657501,
      "rewards/forensics_strict_edge_reward": 0.006053985736798495,
      "rewards/forensics_tiou_reward": 0.412659615278244,
      "step": 259
    },
    {
      "completion_length": 21.65625,
      "epoch": 0.7987711213517665,
      "grad_norm": 6.89665412902832,
      "kl": 0.2236328125,
      "learning_rate": 8.007692307692308e-07,
      "loss": 0.0391,
      "reward": 0.7282393276691437,
      "reward_std": 0.06392431259155273,
      "reward_std_all": 0.06796274892985821,
      "rewards-max/forensics_sinkhorn_reward": 0.9525438547134399,
      "rewards-max/forensics_strict_edge_reward": 0.5419565737247467,
      "rewards-max/forensics_tiou_reward": 0.9175592958927155,
      "rewards/forensics_sinkhorn_reward": 0.6509900093078613,
      "rewards/forensics_strict_edge_reward": 0.025888740550726652,
      "rewards/forensics_tiou_reward": 0.371433362364769,
      "step": 260
    },
    {
      "completion_length": 24.296875,
      "epoch": 0.8018433179723502,
      "grad_norm": 6.29194974899292,
      "kl": 0.2412109375,
      "learning_rate": 8e-07,
      "loss": 0.0688,
      "reward": 0.7015361189842224,
      "reward_std": 0.07445211336016655,
      "reward_std_all": 0.09211361780762672,
      "rewards-max/forensics_sinkhorn_reward": 0.9201373159885406,
      "rewards-max/forensics_strict_edge_reward": 0.012292340630665421,
      "rewards-max/forensics_tiou_reward": 0.8338068127632141,
      "rewards/forensics_sinkhorn_reward": 0.5813829898834229,
      "rewards/forensics_strict_edge_reward": 0.0012138097663410008,
      "rewards/forensics_tiou_reward": 0.3393518924713135,
      "step": 261
    },
    {
      "completion_length": 22.625,
      "epoch": 0.804915514592934,
      "grad_norm": 5.629248142242432,
      "kl": 0.333984375,
      "learning_rate": 7.992307692307692e-07,
      "loss": 0.0149,
      "reward": 0.6942814290523529,
      "reward_std": 0.07743963599205017,
      "reward_std_all": 0.0907723717391491,
      "rewards-max/forensics_sinkhorn_reward": 0.9754647016525269,
      "rewards-max/forensics_strict_edge_reward": 0.5866729617118835,
      "rewards-max/forensics_tiou_reward": 0.9653806090354919,
      "rewards/forensics_sinkhorn_reward": 0.6198190748691559,
      "rewards/forensics_strict_edge_reward": 0.05560909491032362,
      "rewards/forensics_tiou_reward": 0.4325955808162689,
      "step": 262
    },
    {
      "completion_length": 25.09375,
      "epoch": 0.8079877112135176,
      "grad_norm": 5.545593738555908,
      "kl": 0.279296875,
      "learning_rate": 7.984615384615384e-07,
      "loss": 0.0299,
      "reward": 0.6698542833328247,
      "reward_std": 0.12054170668125153,
      "reward_std_all": 0.12954851612448692,
      "rewards-max/forensics_sinkhorn_reward": 0.9569066166877747,
      "rewards-max/forensics_strict_edge_reward": 0.45420393347740173,
      "rewards-max/forensics_tiou_reward": 0.9217717349529266,
      "rewards/forensics_sinkhorn_reward": 0.5757487416267395,
      "rewards/forensics_strict_edge_reward": 0.026264396496117115,
      "rewards/forensics_tiou_reward": 0.3650319427251816,
      "step": 263
    },
    {
      "completion_length": 22.234375,
      "epoch": 0.8110599078341014,
      "grad_norm": 6.623091697692871,
      "kl": 0.302734375,
      "learning_rate": 7.976923076923077e-07,
      "loss": 0.0679,
      "reward": 0.6668092012405396,
      "reward_std": 0.06209804490208626,
      "reward_std_all": 0.09528348222374916,
      "rewards-max/forensics_sinkhorn_reward": 0.7994682490825653,
      "rewards-max/forensics_strict_edge_reward": 0.24709034818806686,
      "rewards-max/forensics_tiou_reward": 0.6519239842891693,
      "rewards/forensics_sinkhorn_reward": 0.5448404550552368,
      "rewards/forensics_strict_edge_reward": 0.010547452340688324,
      "rewards/forensics_tiou_reward": 0.258975088596344,
      "step": 264
    },
    {
      "completion_length": 23.1875,
      "epoch": 0.8141321044546851,
      "grad_norm": 6.627669811248779,
      "kl": 0.3125,
      "learning_rate": 7.969230769230769e-07,
      "loss": 0.2062,
      "reward": 0.6893832087516785,
      "reward_std": 0.07901956513524055,
      "reward_std_all": 0.09633882716298103,
      "rewards-max/forensics_sinkhorn_reward": 0.9667860865592957,
      "rewards-max/forensics_strict_edge_reward": 0.3657659441232681,
      "rewards-max/forensics_tiou_reward": 0.9399965703487396,
      "rewards/forensics_sinkhorn_reward": 0.6089177131652832,
      "rewards/forensics_strict_edge_reward": 0.018976503750309348,
      "rewards/forensics_tiou_reward": 0.33275076746940613,
      "step": 265
    },
    {
      "completion_length": 25.484375,
      "epoch": 0.8172043010752689,
      "grad_norm": 4.167696952819824,
      "kl": 0.3544921875,
      "learning_rate": 7.96153846153846e-07,
      "loss": 0.1289,
      "reward": 0.7026214599609375,
      "reward_std": 0.07625687681138515,
      "reward_std_all": 0.09204575419425964,
      "rewards-max/forensics_sinkhorn_reward": 0.9693440198898315,
      "rewards-max/forensics_strict_edge_reward": 0.28750017285346985,
      "rewards-max/forensics_tiou_reward": 0.9473532140254974,
      "rewards/forensics_sinkhorn_reward": 0.6171993017196655,
      "rewards/forensics_strict_edge_reward": 0.0135659696534276,
      "rewards/forensics_tiou_reward": 0.39984527230262756,
      "step": 266
    },
    {
      "completion_length": 24.484375,
      "epoch": 0.8202764976958525,
      "grad_norm": 5.707553863525391,
      "kl": 0.3642578125,
      "learning_rate": 7.953846153846154e-07,
      "loss": 0.0968,
      "reward": 0.6780809760093689,
      "reward_std": 0.07852715998888016,
      "reward_std_all": 0.09936553984880447,
      "rewards-max/forensics_sinkhorn_reward": 0.9290760457515717,
      "rewards-max/forensics_strict_edge_reward": 0.041347768157720566,
      "rewards-max/forensics_tiou_reward": 0.8257064521312714,
      "rewards/forensics_sinkhorn_reward": 0.5749079585075378,
      "rewards/forensics_strict_edge_reward": 0.0025778745766729116,
      "rewards/forensics_tiou_reward": 0.32397131621837616,
      "step": 267
    },
    {
      "completion_length": 22.3125,
      "epoch": 0.8233486943164362,
      "grad_norm": 5.867967128753662,
      "kl": 0.3115234375,
      "learning_rate": 7.946153846153846e-07,
      "loss": 0.0988,
      "reward": 0.72563636302948,
      "reward_std": 0.06704086996614933,
      "reward_std_all": 0.07404297403991222,
      "rewards-max/forensics_sinkhorn_reward": 0.9761504530906677,
      "rewards-max/forensics_strict_edge_reward": 0.23258191533386707,
      "rewards-max/forensics_tiou_reward": 0.9582293927669525,
      "rewards/forensics_sinkhorn_reward": 0.6216404139995575,
      "rewards/forensics_strict_edge_reward": 0.011096520349383354,
      "rewards/forensics_tiou_reward": 0.3474516272544861,
      "step": 268
    },
    {
      "completion_length": 21.390625,
      "epoch": 0.82642089093702,
      "grad_norm": 6.092067718505859,
      "kl": 0.2998046875,
      "learning_rate": 7.938461538461538e-07,
      "loss": 0.0276,
      "reward": 0.7302030622959137,
      "reward_std": 0.077725725248456,
      "reward_std_all": 0.07602378726005554,
      "rewards-max/forensics_sinkhorn_reward": 0.9838339686393738,
      "rewards-max/forensics_strict_edge_reward": 0.8540171384811401,
      "rewards-max/forensics_tiou_reward": 0.98622265458107,
      "rewards/forensics_sinkhorn_reward": 0.67858025431633,
      "rewards/forensics_strict_edge_reward": 0.06518291123211384,
      "rewards/forensics_tiou_reward": 0.4399529844522476,
      "step": 269
    },
    {
      "completion_length": 26.15625,
      "epoch": 0.8294930875576036,
      "grad_norm": 15.883108139038086,
      "kl": 1.0,
      "learning_rate": 7.930769230769231e-07,
      "loss": 0.0015,
      "reward": 0.7521250247955322,
      "reward_std": 0.020998239517211914,
      "reward_std_all": 0.038430203683674335,
      "rewards-max/forensics_sinkhorn_reward": 0.9604241847991943,
      "rewards-max/forensics_strict_edge_reward": 0.42620211839675903,
      "rewards-max/forensics_tiou_reward": 0.9316227734088898,
      "rewards/forensics_sinkhorn_reward": 0.6807838976383209,
      "rewards/forensics_strict_edge_reward": 0.04545389488339424,
      "rewards/forensics_tiou_reward": 0.4074942171573639,
      "step": 270
    },
    {
      "completion_length": 24.8125,
      "epoch": 0.8325652841781874,
      "grad_norm": 5.225461006164551,
      "kl": 0.2822265625,
      "learning_rate": 7.923076923076922e-07,
      "loss": 0.1311,
      "reward": 0.7581457495689392,
      "reward_std": 0.06092840991914272,
      "reward_std_all": 0.057650476694107056,
      "rewards-max/forensics_sinkhorn_reward": 0.8894889652729034,
      "rewards-max/forensics_strict_edge_reward": 0.5015382021665573,
      "rewards-max/forensics_tiou_reward": 0.7733494341373444,
      "rewards/forensics_sinkhorn_reward": 0.6551737487316132,
      "rewards/forensics_strict_edge_reward": 0.02176005020737648,
      "rewards/forensics_tiou_reward": 0.39180633425712585,
      "step": 271
    },
    {
      "completion_length": 25.859375,
      "epoch": 0.8356374807987711,
      "grad_norm": 6.0699992179870605,
      "kl": 0.390625,
      "learning_rate": 7.915384615384615e-07,
      "loss": -0.0656,
      "reward": 0.7130692005157471,
      "reward_std": 0.04172584135085344,
      "reward_std_all": 0.06679004989564419,
      "rewards-max/forensics_sinkhorn_reward": 0.8626649677753448,
      "rewards-max/forensics_strict_edge_reward": 0.20064860582351685,
      "rewards-max/forensics_tiou_reward": 0.7579407691955566,
      "rewards/forensics_sinkhorn_reward": 0.5848183035850525,
      "rewards/forensics_strict_edge_reward": 0.009038953576236963,
      "rewards/forensics_tiou_reward": 0.380479171872139,
      "step": 272
    },
    {
      "completion_length": 25.359375,
      "epoch": 0.8387096774193549,
      "grad_norm": 4.906902313232422,
      "kl": 0.345703125,
      "learning_rate": 7.907692307692307e-07,
      "loss": 0.2066,
      "reward": 0.7310768067836761,
      "reward_std": 0.10361723601818085,
      "reward_std_all": 0.09471219405531883,
      "rewards-max/forensics_sinkhorn_reward": 0.9193055033683777,
      "rewards-max/forensics_strict_edge_reward": 0.06706571578979492,
      "rewards-max/forensics_tiou_reward": 0.8332789838314056,
      "rewards/forensics_sinkhorn_reward": 0.6229133307933807,
      "rewards/forensics_strict_edge_reward": 0.006418063072487712,
      "rewards/forensics_tiou_reward": 0.39283302426338196,
      "step": 273
    },
    {
      "completion_length": 21.78125,
      "epoch": 0.8417818740399385,
      "grad_norm": 5.5635294914245605,
      "kl": 0.3134765625,
      "learning_rate": 7.9e-07,
      "loss": 0.0318,
      "reward": 0.6976540982723236,
      "reward_std": 0.081711046397686,
      "reward_std_all": 0.09412362426519394,
      "rewards-max/forensics_sinkhorn_reward": 0.8982018232345581,
      "rewards-max/forensics_strict_edge_reward": 0.06094524636864662,
      "rewards-max/forensics_tiou_reward": 0.7721713781356812,
      "rewards/forensics_sinkhorn_reward": 0.5786418318748474,
      "rewards/forensics_strict_edge_reward": 0.0038119893288239837,
      "rewards/forensics_tiou_reward": 0.28583380579948425,
      "step": 274
    },
    {
      "completion_length": 24.4375,
      "epoch": 0.8448540706605223,
      "grad_norm": 5.268381118774414,
      "kl": 0.2392578125,
      "learning_rate": 7.892307692307692e-07,
      "loss": 0.1268,
      "reward": 0.7370825707912445,
      "reward_std": 0.07214194908738136,
      "reward_std_all": 0.07066313922405243,
      "rewards-max/forensics_sinkhorn_reward": 0.9383246898651123,
      "rewards-max/forensics_strict_edge_reward": 0.016252015462669078,
      "rewards-max/forensics_tiou_reward": 0.8678472340106964,
      "rewards/forensics_sinkhorn_reward": 0.6345705389976501,
      "rewards/forensics_strict_edge_reward": 0.0012595038988365559,
      "rewards/forensics_tiou_reward": 0.37016355991363525,
      "step": 275
    },
    {
      "completion_length": 26.671875,
      "epoch": 0.847926267281106,
      "grad_norm": 3.868675947189331,
      "kl": 0.291015625,
      "learning_rate": 7.884615384615384e-07,
      "loss": 0.1489,
      "reward": 0.718096911907196,
      "reward_std": 0.06926103681325912,
      "reward_std_all": 0.07945095002651215,
      "rewards-max/forensics_sinkhorn_reward": 0.9260776937007904,
      "rewards-max/forensics_strict_edge_reward": 0.20559747889637947,
      "rewards-max/forensics_tiou_reward": 0.8396956324577332,
      "rewards/forensics_sinkhorn_reward": 0.6265028119087219,
      "rewards/forensics_strict_edge_reward": 0.01149987243115902,
      "rewards/forensics_tiou_reward": 0.38406355679035187,
      "step": 276
    },
    {
      "completion_length": 25.296875,
      "epoch": 0.8509984639016898,
      "grad_norm": 4.735685348510742,
      "kl": 0.287109375,
      "learning_rate": 7.876923076923077e-07,
      "loss": 0.192,
      "reward": 0.752437949180603,
      "reward_std": 0.07075788453221321,
      "reward_std_all": 0.06595722027122974,
      "rewards-max/forensics_sinkhorn_reward": 0.8961107432842255,
      "rewards-max/forensics_strict_edge_reward": 0.16526542976498604,
      "rewards-max/forensics_tiou_reward": 0.79092738032341,
      "rewards/forensics_sinkhorn_reward": 0.617615669965744,
      "rewards/forensics_strict_edge_reward": 0.007701614638790488,
      "rewards/forensics_tiou_reward": 0.35519348084926605,
      "step": 277
    },
    {
      "completion_length": 24.015625,
      "epoch": 0.8540706605222734,
      "grad_norm": 5.396667957305908,
      "kl": 0.2548828125,
      "learning_rate": 7.869230769230769e-07,
      "loss": 0.168,
      "reward": 0.699101984500885,
      "reward_std": 0.0647578276693821,
      "reward_std_all": 0.08489742875099182,
      "rewards-max/forensics_sinkhorn_reward": 0.9223718047142029,
      "rewards-max/forensics_strict_edge_reward": 0.25190944969654083,
      "rewards-max/forensics_tiou_reward": 0.8336196541786194,
      "rewards/forensics_sinkhorn_reward": 0.5793794095516205,
      "rewards/forensics_strict_edge_reward": 0.011966697871685028,
      "rewards/forensics_tiou_reward": 0.34790879487991333,
      "step": 278
    },
    {
      "completion_length": 24.71875,
      "epoch": 0.8571428571428571,
      "grad_norm": 4.6178297996521,
      "kl": 0.3251953125,
      "learning_rate": 7.861538461538461e-07,
      "loss": 0.0229,
      "reward": 0.7677347958087921,
      "reward_std": 0.04089632537215948,
      "reward_std_all": 0.041025164537131786,
      "rewards-max/forensics_sinkhorn_reward": 0.9721747636795044,
      "rewards-max/forensics_strict_edge_reward": 0.3444451615214348,
      "rewards-max/forensics_tiou_reward": 0.9537800848484039,
      "rewards/forensics_sinkhorn_reward": 0.6908732354640961,
      "rewards/forensics_strict_edge_reward": 0.017983489204198122,
      "rewards/forensics_tiou_reward": 0.4442165195941925,
      "step": 279
    },
    {
      "completion_length": 24.828125,
      "epoch": 0.8602150537634409,
      "grad_norm": 4.9666266441345215,
      "kl": 0.271484375,
      "learning_rate": 7.853846153846153e-07,
      "loss": 0.2159,
      "reward": 0.7345887422561646,
      "reward_std": 0.08365484327077866,
      "reward_std_all": 0.07972240075469017,
      "rewards-max/forensics_sinkhorn_reward": 0.9439181983470917,
      "rewards-max/forensics_strict_edge_reward": 0.2720667803660035,
      "rewards-max/forensics_tiou_reward": 0.876874178647995,
      "rewards/forensics_sinkhorn_reward": 0.6335058212280273,
      "rewards/forensics_strict_edge_reward": 0.020193190954159945,
      "rewards/forensics_tiou_reward": 0.41374771296977997,
      "step": 280
    },
    {
      "completion_length": 23.40625,
      "epoch": 0.8632872503840245,
      "grad_norm": 5.70347261428833,
      "kl": 0.3369140625,
      "learning_rate": 7.846153846153846e-07,
      "loss": 0.178,
      "reward": 0.7006502151489258,
      "reward_std": 0.08405781909823418,
      "reward_std_all": 0.09093433618545532,
      "rewards-max/forensics_sinkhorn_reward": 0.9522206783294678,
      "rewards-max/forensics_strict_edge_reward": 0.24467409402132034,
      "rewards-max/forensics_tiou_reward": 0.8992456793785095,
      "rewards/forensics_sinkhorn_reward": 0.6150844991207123,
      "rewards/forensics_strict_edge_reward": 0.01842471957206726,
      "rewards/forensics_tiou_reward": 0.37139779329299927,
      "step": 281
    },
    {
      "completion_length": 24.234375,
      "epoch": 0.8663594470046083,
      "grad_norm": 5.739495277404785,
      "kl": 0.259765625,
      "learning_rate": 7.838461538461539e-07,
      "loss": 0.0741,
      "reward": 0.719310075044632,
      "reward_std": 0.09620362147688866,
      "reward_std_all": 0.09445143863558769,
      "rewards-max/forensics_sinkhorn_reward": 0.9694356918334961,
      "rewards-max/forensics_strict_edge_reward": 0.5200863555073738,
      "rewards-max/forensics_tiou_reward": 0.9494955241680145,
      "rewards/forensics_sinkhorn_reward": 0.609334796667099,
      "rewards/forensics_strict_edge_reward": 0.04468264151364565,
      "rewards/forensics_tiou_reward": 0.4120926260948181,
      "step": 282
    },
    {
      "completion_length": 25.078125,
      "epoch": 0.869431643625192,
      "grad_norm": 4.8153181076049805,
      "kl": 0.3623046875,
      "learning_rate": 7.830769230769231e-07,
      "loss": -0.016,
      "reward": 0.7604717314243317,
      "reward_std": 0.03693707624915987,
      "reward_std_all": 0.04220361984334886,
      "rewards-max/forensics_sinkhorn_reward": 0.9808802902698517,
      "rewards-max/forensics_strict_edge_reward": 0.23684818297624588,
      "rewards-max/forensics_tiou_reward": 0.9730139970779419,
      "rewards/forensics_sinkhorn_reward": 0.7013029456138611,
      "rewards/forensics_strict_edge_reward": 0.029371734708547592,
      "rewards/forensics_tiou_reward": 0.44936278462409973,
      "step": 283
    },
    {
      "completion_length": 27.015625,
      "epoch": 0.8725038402457758,
      "grad_norm": 3.8992795944213867,
      "kl": 0.3251953125,
      "learning_rate": 7.823076923076922e-07,
      "loss": 0.0985,
      "reward": 0.6973054707050323,
      "reward_std": 0.1383657343685627,
      "reward_std_all": 0.12785978615283966,
      "rewards-max/forensics_sinkhorn_reward": 0.9238093197345734,
      "rewards-max/forensics_strict_edge_reward": 0.1970017198473215,
      "rewards-max/forensics_tiou_reward": 0.8396748006343842,
      "rewards/forensics_sinkhorn_reward": 0.5928184390068054,
      "rewards/forensics_strict_edge_reward": 0.010294231004081666,
      "rewards/forensics_tiou_reward": 0.37562063336372375,
      "step": 284
    },
    {
      "completion_length": 24.125,
      "epoch": 0.8755760368663594,
      "grad_norm": 4.1608428955078125,
      "kl": 0.2568359375,
      "learning_rate": 7.815384615384615e-07,
      "loss": 0.0584,
      "reward": 0.6477946043014526,
      "reward_std": 0.1021878831088543,
      "reward_std_all": 0.12207705900073051,
      "rewards-max/forensics_sinkhorn_reward": 0.9254492223262787,
      "rewards-max/forensics_strict_edge_reward": 0.1217162050306797,
      "rewards-max/forensics_tiou_reward": 0.846368670463562,
      "rewards/forensics_sinkhorn_reward": 0.5469700992107391,
      "rewards/forensics_strict_edge_reward": 0.0052328757010400295,
      "rewards/forensics_tiou_reward": 0.23347827047109604,
      "step": 285
    },
    {
      "completion_length": 24.640625,
      "epoch": 0.8786482334869432,
      "grad_norm": 3.909309148788452,
      "kl": 0.26318359375,
      "learning_rate": 7.807692307692307e-07,
      "loss": 0.0415,
      "reward": 0.762953907251358,
      "reward_std": 0.0482370276004076,
      "reward_std_all": 0.04729977063834667,
      "rewards-max/forensics_sinkhorn_reward": 0.9424977004528046,
      "rewards-max/forensics_strict_edge_reward": 0.2528768260963261,
      "rewards-max/forensics_tiou_reward": 0.8865786790847778,
      "rewards/forensics_sinkhorn_reward": 0.675260603427887,
      "rewards/forensics_strict_edge_reward": 0.010763448517536744,
      "rewards/forensics_tiou_reward": 0.4369168281555176,
      "step": 286
    },
    {
      "completion_length": 22.53125,
      "epoch": 0.8817204301075269,
      "grad_norm": 4.685935020446777,
      "kl": 0.255859375,
      "learning_rate": 7.799999999999999e-07,
      "loss": 0.0512,
      "reward": 0.7247497141361237,
      "reward_std": 0.06522506661713123,
      "reward_std_all": 0.07198967412114143,
      "rewards-max/forensics_sinkhorn_reward": 0.9422484636306763,
      "rewards-max/forensics_strict_edge_reward": 0.19799603521823883,
      "rewards-max/forensics_tiou_reward": 0.8860955834388733,
      "rewards/forensics_sinkhorn_reward": 0.6177974343299866,
      "rewards/forensics_strict_edge_reward": 0.01316517498344183,
      "rewards/forensics_tiou_reward": 0.34392018616199493,
      "step": 287
    },
    {
      "completion_length": 23.328125,
      "epoch": 0.8847926267281107,
      "grad_norm": 4.722728729248047,
      "kl": 0.29248046875,
      "learning_rate": 7.792307692307693e-07,
      "loss": 0.0021,
      "reward": 0.7652672231197357,
      "reward_std": 0.0563141405582428,
      "reward_std_all": 0.05286753736436367,
      "rewards-max/forensics_sinkhorn_reward": 0.9544228613376617,
      "rewards-max/forensics_strict_edge_reward": 0.16043865494430065,
      "rewards-max/forensics_tiou_reward": 0.9057233035564423,
      "rewards/forensics_sinkhorn_reward": 0.667724996805191,
      "rewards/forensics_strict_edge_reward": 0.007689413381740451,
      "rewards/forensics_tiou_reward": 0.3733496516942978,
      "step": 288
    },
    {
      "completion_length": 24.703125,
      "epoch": 0.8878648233486943,
      "grad_norm": 4.701695442199707,
      "kl": 0.345703125,
      "learning_rate": 7.784615384615384e-07,
      "loss": 0.1052,
      "reward": 0.7054867148399353,
      "reward_std": 0.053681494668126106,
      "reward_std_all": 0.07643408142030239,
      "rewards-max/forensics_sinkhorn_reward": 0.9284659922122955,
      "rewards-max/forensics_strict_edge_reward": 0.2041583598474972,
      "rewards-max/forensics_tiou_reward": 0.8497925996780396,
      "rewards/forensics_sinkhorn_reward": 0.6059238910675049,
      "rewards/forensics_strict_edge_reward": 0.015462475479580462,
      "rewards/forensics_tiou_reward": 0.36853934824466705,
      "step": 289
    },
    {
      "completion_length": 26.34375,
      "epoch": 0.890937019969278,
      "grad_norm": 9.837812423706055,
      "kl": 0.9765625,
      "learning_rate": 7.776923076923077e-07,
      "loss": 0.1456,
      "reward": 0.659755676984787,
      "reward_std": 0.1385515332221985,
      "reward_std_all": 0.14137642830610275,
      "rewards-max/forensics_sinkhorn_reward": 0.9280849695205688,
      "rewards-max/forensics_strict_edge_reward": 0.4255645051598549,
      "rewards-max/forensics_tiou_reward": 0.8210785090923309,
      "rewards/forensics_sinkhorn_reward": 0.5353768765926361,
      "rewards/forensics_strict_edge_reward": 0.01898962678387761,
      "rewards/forensics_tiou_reward": 0.33298276364803314,
      "step": 290
    },
    {
      "completion_length": 25.671875,
      "epoch": 0.8940092165898618,
      "grad_norm": 3.8514654636383057,
      "kl": 0.22509765625,
      "learning_rate": 7.769230769230769e-07,
      "loss": 0.098,
      "reward": 0.7088953256607056,
      "reward_std": 0.07658898830413818,
      "reward_std_all": 0.08514141663908958,
      "rewards-max/forensics_sinkhorn_reward": 0.9760886132717133,
      "rewards-max/forensics_strict_edge_reward": 0.48844531178474426,
      "rewards-max/forensics_tiou_reward": 0.9656158685684204,
      "rewards/forensics_sinkhorn_reward": 0.6245160698890686,
      "rewards/forensics_strict_edge_reward": 0.02334593702107668,
      "rewards/forensics_tiou_reward": 0.4081864356994629,
      "step": 291
    },
    {
      "completion_length": 24.140625,
      "epoch": 0.8970814132104454,
      "grad_norm": 4.755120277404785,
      "kl": 0.27197265625,
      "learning_rate": 7.761538461538461e-07,
      "loss": 0.1555,
      "reward": 0.7268873155117035,
      "reward_std": 0.0502210296690464,
      "reward_std_all": 0.06664086133241653,
      "rewards-max/forensics_sinkhorn_reward": 0.9679001867771149,
      "rewards-max/forensics_strict_edge_reward": 0.3457963317632675,
      "rewards-max/forensics_tiou_reward": 0.9433531165122986,
      "rewards/forensics_sinkhorn_reward": 0.6718997359275818,
      "rewards/forensics_strict_edge_reward": 0.02804142888635397,
      "rewards/forensics_tiou_reward": 0.45335087180137634,
      "step": 292
    },
    {
      "completion_length": 24.328125,
      "epoch": 0.9001536098310292,
      "grad_norm": 4.136373996734619,
      "kl": 0.2978515625,
      "learning_rate": 7.753846153846153e-07,
      "loss": 0.126,
      "reward": 0.6883220076560974,
      "reward_std": 0.07080469653010368,
      "reward_std_all": 0.09396510571241379,
      "rewards-max/forensics_sinkhorn_reward": 0.9451287090778351,
      "rewards-max/forensics_strict_edge_reward": 0.21754248067736626,
      "rewards-max/forensics_tiou_reward": 0.8828135132789612,
      "rewards/forensics_sinkhorn_reward": 0.6097986102104187,
      "rewards/forensics_strict_edge_reward": 0.023617387283593416,
      "rewards/forensics_tiou_reward": 0.41073334217071533,
      "step": 293
    },
    {
      "completion_length": 25.625,
      "epoch": 0.9032258064516129,
      "grad_norm": 4.032013893127441,
      "kl": 0.3017578125,
      "learning_rate": 7.746153846153845e-07,
      "loss": 0.2038,
      "reward": 0.7482579052448273,
      "reward_std": 0.07030130922794342,
      "reward_std_all": 0.06638358905911446,
      "rewards-max/forensics_sinkhorn_reward": 0.968904584646225,
      "rewards-max/forensics_strict_edge_reward": 0.3398134857416153,
      "rewards-max/forensics_tiou_reward": 0.9414674043655396,
      "rewards/forensics_sinkhorn_reward": 0.6591953933238983,
      "rewards/forensics_strict_edge_reward": 0.01738337194547057,
      "rewards/forensics_tiou_reward": 0.43670621514320374,
      "step": 294
    },
    {
      "completion_length": 24.5,
      "epoch": 0.9062980030721967,
      "grad_norm": 4.255565643310547,
      "kl": 0.3037109375,
      "learning_rate": 7.738461538461539e-07,
      "loss": 0.1473,
      "reward": 0.7237671315670013,
      "reward_std": 0.0859699547290802,
      "reward_std_all": 0.08371635153889656,
      "rewards-max/forensics_sinkhorn_reward": 0.9418161511421204,
      "rewards-max/forensics_strict_edge_reward": 0.4626513756811619,
      "rewards-max/forensics_tiou_reward": 0.8836004436016083,
      "rewards/forensics_sinkhorn_reward": 0.6355989873409271,
      "rewards/forensics_strict_edge_reward": 0.02417264529503882,
      "rewards/forensics_tiou_reward": 0.415902316570282,
      "step": 295
    },
    {
      "completion_length": 25.5,
      "epoch": 0.9093701996927803,
      "grad_norm": 4.428289413452148,
      "kl": 0.2900390625,
      "learning_rate": 7.730769230769231e-07,
      "loss": 0.1949,
      "reward": 0.6990401148796082,
      "reward_std": 0.09815074503421783,
      "reward_std_all": 0.10085657984018326,
      "rewards-max/forensics_sinkhorn_reward": 0.9590698480606079,
      "rewards-max/forensics_strict_edge_reward": 0.04371035424992442,
      "rewards-max/forensics_tiou_reward": 0.9187713265419006,
      "rewards/forensics_sinkhorn_reward": 0.5836626589298248,
      "rewards/forensics_strict_edge_reward": 0.0039052016800269485,
      "rewards/forensics_tiou_reward": 0.3731176555156708,
      "step": 296
    },
    {
      "completion_length": 24.84375,
      "epoch": 0.9124423963133641,
      "grad_norm": 4.516807556152344,
      "kl": 0.271484375,
      "learning_rate": 7.723076923076923e-07,
      "loss": 0.0449,
      "reward": 0.6684067249298096,
      "reward_std": 0.068304143846035,
      "reward_std_all": 0.10006116330623627,
      "rewards-max/forensics_sinkhorn_reward": 0.9391745626926422,
      "rewards-max/forensics_strict_edge_reward": 0.1603681594133377,
      "rewards-max/forensics_tiou_reward": 0.8755998611450195,
      "rewards/forensics_sinkhorn_reward": 0.6084852814674377,
      "rewards/forensics_strict_edge_reward": 0.01122452737763524,
      "rewards/forensics_tiou_reward": 0.3746866285800934,
      "step": 297
    },
    {
      "completion_length": 25.25,
      "epoch": 0.9155145929339478,
      "grad_norm": 4.1301751136779785,
      "kl": 0.3212890625,
      "learning_rate": 7.715384615384615e-07,
      "loss": 0.1003,
      "reward": 0.7491476237773895,
      "reward_std": 0.05193051137030125,
      "reward_std_all": 0.05768792703747749,
      "rewards-max/forensics_sinkhorn_reward": 0.9317513704299927,
      "rewards-max/forensics_strict_edge_reward": 0.21969910897314548,
      "rewards-max/forensics_tiou_reward": 0.8692281544208527,
      "rewards/forensics_sinkhorn_reward": 0.6244244575500488,
      "rewards/forensics_strict_edge_reward": 0.009370067156851292,
      "rewards/forensics_tiou_reward": 0.39235980808734894,
      "step": 298
    },
    {
      "completion_length": 24.71875,
      "epoch": 0.9185867895545314,
      "grad_norm": 5.248650550842285,
      "kl": 0.28515625,
      "learning_rate": 7.707692307692307e-07,
      "loss": 0.0929,
      "reward": 0.7641482949256897,
      "reward_std": 0.057937076315283775,
      "reward_std_all": 0.05430866591632366,
      "rewards-max/forensics_sinkhorn_reward": 0.97834312915802,
      "rewards-max/forensics_strict_edge_reward": 0.7094893157482147,
      "rewards-max/forensics_tiou_reward": 0.9737511277198792,
      "rewards/forensics_sinkhorn_reward": 0.6541356146335602,
      "rewards/forensics_strict_edge_reward": 0.03973455261439085,
      "rewards/forensics_tiou_reward": 0.4091818630695343,
      "step": 299
    },
    {
      "completion_length": 22.90625,
      "epoch": 0.9216589861751152,
      "grad_norm": 4.561899662017822,
      "kl": 0.3349609375,
      "learning_rate": 7.699999999999999e-07,
      "loss": 0.0762,
      "reward": 0.699769139289856,
      "reward_std": 0.08865214884281158,
      "reward_std_all": 0.09403994679450989,
      "rewards-max/forensics_sinkhorn_reward": 0.8761423230171204,
      "rewards-max/forensics_strict_edge_reward": 0.1633713562041521,
      "rewards-max/forensics_tiou_reward": 0.7328893840312958,
      "rewards/forensics_sinkhorn_reward": 0.572772890329361,
      "rewards/forensics_strict_edge_reward": 0.008868860895745456,
      "rewards/forensics_tiou_reward": 0.28295662999153137,
      "step": 300
    },
    {
      "completion_length": 27.21875,
      "epoch": 0.9247311827956989,
      "grad_norm": 5.984442234039307,
      "kl": 0.3056640625,
      "learning_rate": 7.692307692307693e-07,
      "loss": 0.1322,
      "reward": 0.7078374624252319,
      "reward_std": 0.07505258545279503,
      "reward_std_all": 0.0888281762599945,
      "rewards-max/forensics_sinkhorn_reward": 0.9654982686042786,
      "rewards-max/forensics_strict_edge_reward": 0.02928034239448607,
      "rewards-max/forensics_tiou_reward": 0.9339741766452789,
      "rewards/forensics_sinkhorn_reward": 0.5905752778053284,
      "rewards/forensics_strict_edge_reward": 0.0027536442794371396,
      "rewards/forensics_tiou_reward": 0.3873390704393387,
      "step": 301
    },
    {
      "completion_length": 26.390625,
      "epoch": 0.9278033794162827,
      "grad_norm": 4.165913105010986,
      "kl": 0.34375,
      "learning_rate": 7.684615384615384e-07,
      "loss": 0.0792,
      "reward": 0.7437675595283508,
      "reward_std": 0.035327247343957424,
      "reward_std_all": 0.050568124279379845,
      "rewards-max/forensics_sinkhorn_reward": 0.9769213795661926,
      "rewards-max/forensics_strict_edge_reward": 0.37680648267269135,
      "rewards-max/forensics_tiou_reward": 0.9647789001464844,
      "rewards/forensics_sinkhorn_reward": 0.6332288980484009,
      "rewards/forensics_strict_edge_reward": 0.02483657654374838,
      "rewards/forensics_tiou_reward": 0.4251600652933121,
      "step": 302
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.9308755760368663,
      "grad_norm": 6.3576130867004395,
      "kl": 0.4931640625,
      "learning_rate": 7.676923076923077e-07,
      "loss": 0.1399,
      "reward": 0.6676984429359436,
      "reward_std": 0.12409999594092369,
      "reward_std_all": 0.13369430229067802,
      "rewards-max/forensics_sinkhorn_reward": 0.9214170873165131,
      "rewards-max/forensics_strict_edge_reward": 0.1345666623674333,
      "rewards-max/forensics_tiou_reward": 0.8233234286308289,
      "rewards/forensics_sinkhorn_reward": 0.5664394497871399,
      "rewards/forensics_strict_edge_reward": 0.006971884839003906,
      "rewards/forensics_tiou_reward": 0.30799075961112976,
      "step": 303
    },
    {
      "completion_length": 26.953125,
      "epoch": 0.9339477726574501,
      "grad_norm": 4.459205627441406,
      "kl": 0.3564453125,
      "learning_rate": 7.669230769230769e-07,
      "loss": 0.0828,
      "reward": 0.6727229356765747,
      "reward_std": 0.10204028710722923,
      "reward_std_all": 0.1112598106265068,
      "rewards-max/forensics_sinkhorn_reward": 0.967591255903244,
      "rewards-max/forensics_strict_edge_reward": 0.42603883147239685,
      "rewards-max/forensics_tiou_reward": 0.945504754781723,
      "rewards/forensics_sinkhorn_reward": 0.5035796463489532,
      "rewards/forensics_strict_edge_reward": 0.034931283444166183,
      "rewards/forensics_tiou_reward": 0.35796183347702026,
      "step": 304
    },
    {
      "completion_length": 26.734375,
      "epoch": 0.9370199692780338,
      "grad_norm": 4.53789758682251,
      "kl": 0.2822265625,
      "learning_rate": 7.661538461538461e-07,
      "loss": 0.1944,
      "reward": 0.7299318611621857,
      "reward_std": 0.0928104780614376,
      "reward_std_all": 0.08734187111258507,
      "rewards-max/forensics_sinkhorn_reward": 0.9780260324478149,
      "rewards-max/forensics_strict_edge_reward": 0.10070898942649364,
      "rewards-max/forensics_tiou_reward": 0.9669848084449768,
      "rewards/forensics_sinkhorn_reward": 0.6240084767341614,
      "rewards/forensics_strict_edge_reward": 0.010453511495143175,
      "rewards/forensics_tiou_reward": 0.41636328399181366,
      "step": 305
    },
    {
      "completion_length": 26.890625,
      "epoch": 0.9400921658986175,
      "grad_norm": 5.962343215942383,
      "kl": 0.302734375,
      "learning_rate": 7.653846153846153e-07,
      "loss": 0.0745,
      "reward": 0.7240256369113922,
      "reward_std": 0.09504618495702744,
      "reward_std_all": 0.09036466479301453,
      "rewards-max/forensics_sinkhorn_reward": 0.9410138726234436,
      "rewards-max/forensics_strict_edge_reward": 0.3205677345395088,
      "rewards-max/forensics_tiou_reward": 0.8814308941364288,
      "rewards/forensics_sinkhorn_reward": 0.6415120959281921,
      "rewards/forensics_strict_edge_reward": 0.02194668957963586,
      "rewards/forensics_tiou_reward": 0.4167020171880722,
      "step": 306
    },
    {
      "completion_length": 24.59375,
      "epoch": 0.9431643625192012,
      "grad_norm": 6.36277961730957,
      "kl": 0.505859375,
      "learning_rate": 7.646153846153845e-07,
      "loss": 0.1414,
      "reward": 0.7291260063648224,
      "reward_std": 0.06925836578011513,
      "reward_std_all": 0.0707474946975708,
      "rewards-max/forensics_sinkhorn_reward": 0.9229342043399811,
      "rewards-max/forensics_strict_edge_reward": 0.3020684075600002,
      "rewards-max/forensics_tiou_reward": 0.8218704164028168,
      "rewards/forensics_sinkhorn_reward": 0.619948148727417,
      "rewards/forensics_strict_edge_reward": 0.02295227309150505,
      "rewards/forensics_tiou_reward": 0.35505685210227966,
      "step": 307
    },
    {
      "completion_length": 24.953125,
      "epoch": 0.946236559139785,
      "grad_norm": 4.538080215454102,
      "kl": 0.2666015625,
      "learning_rate": 7.638461538461539e-07,
      "loss": 0.0321,
      "reward": 0.6892088353633881,
      "reward_std": 0.11952335759997368,
      "reward_std_all": 0.1168345995247364,
      "rewards-max/forensics_sinkhorn_reward": 0.9541870355606079,
      "rewards-max/forensics_strict_edge_reward": 0.3444378301501274,
      "rewards-max/forensics_tiou_reward": 0.9162771105766296,
      "rewards/forensics_sinkhorn_reward": 0.5811231136322021,
      "rewards/forensics_strict_edge_reward": 0.022781209088861942,
      "rewards/forensics_tiou_reward": 0.2949971854686737,
      "step": 308
    },
    {
      "completion_length": 26.234375,
      "epoch": 0.9493087557603687,
      "grad_norm": 4.3277692794799805,
      "kl": 0.3193359375,
      "learning_rate": 7.630769230769231e-07,
      "loss": 0.1163,
      "reward": 0.7175950407981873,
      "reward_std": 0.06151561066508293,
      "reward_std_all": 0.07581937871873379,
      "rewards-max/forensics_sinkhorn_reward": 0.971974641084671,
      "rewards-max/forensics_strict_edge_reward": 0.45442917943000793,
      "rewards-max/forensics_tiou_reward": 0.9527762234210968,
      "rewards/forensics_sinkhorn_reward": 0.6547692716121674,
      "rewards/forensics_strict_edge_reward": 0.045373572036623955,
      "rewards/forensics_tiou_reward": 0.4365759640932083,
      "step": 309
    },
    {
      "completion_length": 25.640625,
      "epoch": 0.9523809523809523,
      "grad_norm": 4.717088222503662,
      "kl": 0.3642578125,
      "learning_rate": 7.623076923076923e-07,
      "loss": 0.0861,
      "reward": 0.6965654194355011,
      "reward_std": 0.08287546783685684,
      "reward_std_all": 0.09541179053485394,
      "rewards-max/forensics_sinkhorn_reward": 0.9622521102428436,
      "rewards-max/forensics_strict_edge_reward": 0.15972213703207672,
      "rewards-max/forensics_tiou_reward": 0.9278564453125,
      "rewards/forensics_sinkhorn_reward": 0.5862343311309814,
      "rewards/forensics_strict_edge_reward": 0.0067446535103954375,
      "rewards/forensics_tiou_reward": 0.3718530386686325,
      "step": 310
    },
    {
      "completion_length": 26.03125,
      "epoch": 0.9554531490015361,
      "grad_norm": 4.450074672698975,
      "kl": 0.3330078125,
      "learning_rate": 7.615384615384615e-07,
      "loss": 0.1154,
      "reward": 0.7427191138267517,
      "reward_std": 0.06667456030845642,
      "reward_std_all": 0.06563736125826836,
      "rewards-max/forensics_sinkhorn_reward": 0.9677404165267944,
      "rewards-max/forensics_strict_edge_reward": 0.2884034365415573,
      "rewards-max/forensics_tiou_reward": 0.9428229928016663,
      "rewards/forensics_sinkhorn_reward": 0.6401242017745972,
      "rewards/forensics_strict_edge_reward": 0.014998025260865688,
      "rewards/forensics_tiou_reward": 0.4016861915588379,
      "step": 311
    },
    {
      "completion_length": 27.90625,
      "epoch": 0.9585253456221198,
      "grad_norm": 3.905634880065918,
      "kl": 0.27587890625,
      "learning_rate": 7.607692307692307e-07,
      "loss": -0.0245,
      "reward": 0.7784863114356995,
      "reward_std": 0.03142197243869305,
      "reward_std_all": 0.03161345422267914,
      "rewards-max/forensics_sinkhorn_reward": 0.953275591135025,
      "rewards-max/forensics_strict_edge_reward": 0.21228813380002975,
      "rewards-max/forensics_tiou_reward": 0.9102238118648529,
      "rewards/forensics_sinkhorn_reward": 0.6876585781574249,
      "rewards/forensics_strict_edge_reward": 0.014734788332134485,
      "rewards/forensics_tiou_reward": 0.48231247067451477,
      "step": 312
    },
    {
      "completion_length": 24.59375,
      "epoch": 0.9615975422427036,
      "grad_norm": 4.565439701080322,
      "kl": 0.2880859375,
      "learning_rate": 7.599999999999999e-07,
      "loss": 0.0154,
      "reward": 0.7544332444667816,
      "reward_std": 0.03999071568250656,
      "reward_std_all": 0.04647022671997547,
      "rewards-max/forensics_sinkhorn_reward": 0.9766354858875275,
      "rewards-max/forensics_strict_edge_reward": 0.2554169371724129,
      "rewards-max/forensics_tiou_reward": 0.962184876203537,
      "rewards/forensics_sinkhorn_reward": 0.6559754312038422,
      "rewards/forensics_strict_edge_reward": 0.01266572019085288,
      "rewards/forensics_tiou_reward": 0.43548719584941864,
      "step": 313
    },
    {
      "completion_length": 28.625,
      "epoch": 0.9646697388632872,
      "grad_norm": 4.616283893585205,
      "kl": 0.3203125,
      "learning_rate": 7.592307692307692e-07,
      "loss": 0.0697,
      "reward": 0.6729896366596222,
      "reward_std": 0.07028049416840076,
      "reward_std_all": 0.09923014603555202,
      "rewards-max/forensics_sinkhorn_reward": 0.9006368219852448,
      "rewards-max/forensics_strict_edge_reward": 0.1242650868371129,
      "rewards-max/forensics_tiou_reward": 0.8187947571277618,
      "rewards/forensics_sinkhorn_reward": 0.5414410531520844,
      "rewards/forensics_strict_edge_reward": 0.005588885163888335,
      "rewards/forensics_tiou_reward": 0.33800986409187317,
      "step": 314
    },
    {
      "completion_length": 23.15625,
      "epoch": 0.967741935483871,
      "grad_norm": 5.738778591156006,
      "kl": 0.28515625,
      "learning_rate": 7.584615384615384e-07,
      "loss": 0.0855,
      "reward": 0.7429161071777344,
      "reward_std": 0.07141944393515587,
      "reward_std_all": 0.06886360608041286,
      "rewards-max/forensics_sinkhorn_reward": 0.9702902436256409,
      "rewards-max/forensics_strict_edge_reward": 0.627477690577507,
      "rewards-max/forensics_tiou_reward": 0.9537985324859619,
      "rewards/forensics_sinkhorn_reward": 0.66853067278862,
      "rewards/forensics_strict_edge_reward": 0.032979543786495924,
      "rewards/forensics_tiou_reward": 0.46993783116340637,
      "step": 315
    },
    {
      "completion_length": 24.75,
      "epoch": 0.9708141321044547,
      "grad_norm": 6.098962783813477,
      "kl": 0.310546875,
      "learning_rate": 7.576923076923077e-07,
      "loss": 0.1718,
      "reward": 0.7275183200836182,
      "reward_std": 0.09960155934095383,
      "reward_std_all": 0.09286190569400787,
      "rewards-max/forensics_sinkhorn_reward": 0.907855898141861,
      "rewards-max/forensics_strict_edge_reward": 0.18995207082480192,
      "rewards-max/forensics_tiou_reward": 0.8083353340625763,
      "rewards/forensics_sinkhorn_reward": 0.5987396836280823,
      "rewards/forensics_strict_edge_reward": 0.012633270700462162,
      "rewards/forensics_tiou_reward": 0.41542962193489075,
      "step": 316
    },
    {
      "completion_length": 22.296875,
      "epoch": 0.9738863287250384,
      "grad_norm": 4.6264214515686035,
      "kl": 0.3056640625,
      "learning_rate": 7.569230769230769e-07,
      "loss": 0.1346,
      "reward": 0.7239348292350769,
      "reward_std": 0.1033632829785347,
      "reward_std_all": 0.09628850966691971,
      "rewards-max/forensics_sinkhorn_reward": 0.962471216917038,
      "rewards-max/forensics_strict_edge_reward": 0.29105816036462784,
      "rewards-max/forensics_tiou_reward": 0.9349110722541809,
      "rewards/forensics_sinkhorn_reward": 0.6151956021785736,
      "rewards/forensics_strict_edge_reward": 0.023381546139717102,
      "rewards/forensics_tiou_reward": 0.3964979499578476,
      "step": 317
    },
    {
      "completion_length": 23.609375,
      "epoch": 0.9769585253456221,
      "grad_norm": 5.274281978607178,
      "kl": 0.326171875,
      "learning_rate": 7.561538461538461e-07,
      "loss": 0.0757,
      "reward": 0.7262996733188629,
      "reward_std": 0.0710979700088501,
      "reward_std_all": 0.07727339677512646,
      "rewards-max/forensics_sinkhorn_reward": 0.9553490579128265,
      "rewards-max/forensics_strict_edge_reward": 0.2954631745815277,
      "rewards-max/forensics_tiou_reward": 0.8992795050144196,
      "rewards/forensics_sinkhorn_reward": 0.6369271278381348,
      "rewards/forensics_strict_edge_reward": 0.017689512111246586,
      "rewards/forensics_tiou_reward": 0.40029312670230865,
      "step": 318
    },
    {
      "completion_length": 23.078125,
      "epoch": 0.9800307219662059,
      "grad_norm": 5.675666332244873,
      "kl": 0.23974609375,
      "learning_rate": 7.553846153846154e-07,
      "loss": 0.2499,
      "reward": 0.7401043772697449,
      "reward_std": 0.08478302136063576,
      "reward_std_all": 0.07872291654348373,
      "rewards-max/forensics_sinkhorn_reward": 0.9677499234676361,
      "rewards-max/forensics_strict_edge_reward": 0.16922210901975632,
      "rewards-max/forensics_tiou_reward": 0.9445432126522064,
      "rewards/forensics_sinkhorn_reward": 0.6619195938110352,
      "rewards/forensics_strict_edge_reward": 0.01417362317442894,
      "rewards/forensics_tiou_reward": 0.4708969295024872,
      "step": 319
    },
    {
      "completion_length": 23.5625,
      "epoch": 0.9831029185867896,
      "grad_norm": 4.555226802825928,
      "kl": 0.27197265625,
      "learning_rate": 7.546153846153845e-07,
      "loss": 0.1198,
      "reward": 0.7131784856319427,
      "reward_std": 0.05602079909294844,
      "reward_std_all": 0.07030285894870758,
      "rewards-max/forensics_sinkhorn_reward": 0.9328892230987549,
      "rewards-max/forensics_strict_edge_reward": 0.13943806290626526,
      "rewards-max/forensics_tiou_reward": 0.8644803166389465,
      "rewards/forensics_sinkhorn_reward": 0.5988501906394958,
      "rewards/forensics_strict_edge_reward": 0.006375822820700705,
      "rewards/forensics_tiou_reward": 0.349368155002594,
      "step": 320
    },
    {
      "completion_length": 22.171875,
      "epoch": 0.9861751152073732,
      "grad_norm": 6.212404251098633,
      "kl": 0.5009765625,
      "learning_rate": 7.538461538461538e-07,
      "loss": 0.034,
      "reward": 0.7720372080802917,
      "reward_std": 0.039276767522096634,
      "reward_std_all": 0.038674263283610344,
      "rewards-max/forensics_sinkhorn_reward": 0.9705677628517151,
      "rewards-max/forensics_strict_edge_reward": 0.3020239621400833,
      "rewards-max/forensics_tiou_reward": 0.9495319724082947,
      "rewards/forensics_sinkhorn_reward": 0.725098192691803,
      "rewards/forensics_strict_edge_reward": 0.02235479559749365,
      "rewards/forensics_tiou_reward": 0.4681696593761444,
      "step": 321
    },
    {
      "completion_length": 22.0,
      "epoch": 0.989247311827957,
      "grad_norm": 5.419009685516357,
      "kl": 0.30859375,
      "learning_rate": 7.530769230769231e-07,
      "loss": 0.0102,
      "reward": 0.7287951111793518,
      "reward_std": 0.07146528735756874,
      "reward_std_all": 0.07436562888324261,
      "rewards-max/forensics_sinkhorn_reward": 0.9553743302822113,
      "rewards-max/forensics_strict_edge_reward": 0.3972083032131195,
      "rewards-max/forensics_tiou_reward": 0.9228329360485077,
      "rewards/forensics_sinkhorn_reward": 0.6315914988517761,
      "rewards/forensics_strict_edge_reward": 0.019809349440038204,
      "rewards/forensics_tiou_reward": 0.3868337422609329,
      "step": 322
    },
    {
      "completion_length": 23.859375,
      "epoch": 0.9923195084485407,
      "grad_norm": 5.461663722991943,
      "kl": 0.443359375,
      "learning_rate": 7.523076923076923e-07,
      "loss": 0.046,
      "reward": 0.7460785508155823,
      "reward_std": 0.0521665383130312,
      "reward_std_all": 0.058971259742975235,
      "rewards-max/forensics_sinkhorn_reward": 0.9671130478382111,
      "rewards-max/forensics_strict_edge_reward": 0.5049435347318649,
      "rewards-max/forensics_tiou_reward": 0.9453136622905731,
      "rewards/forensics_sinkhorn_reward": 0.6608531475067139,
      "rewards/forensics_strict_edge_reward": 0.0416463240981102,
      "rewards/forensics_tiou_reward": 0.41939760744571686,
      "step": 323
    },
    {
      "completion_length": 26.375,
      "epoch": 0.9953917050691244,
      "grad_norm": 6.382907390594482,
      "kl": 0.3515625,
      "learning_rate": 7.515384615384615e-07,
      "loss": 0.0793,
      "reward": 0.7459478378295898,
      "reward_std": 0.07201658189296722,
      "reward_std_all": 0.06823875568807125,
      "rewards-max/forensics_sinkhorn_reward": 0.9698248207569122,
      "rewards-max/forensics_strict_edge_reward": 0.30878981202840805,
      "rewards-max/forensics_tiou_reward": 0.9513213634490967,
      "rewards/forensics_sinkhorn_reward": 0.6564161777496338,
      "rewards/forensics_strict_edge_reward": 0.015952121000736952,
      "rewards/forensics_tiou_reward": 0.39976367354393005,
      "step": 324
    },
    {
      "completion_length": 22.765625,
      "epoch": 0.9984639016897081,
      "grad_norm": 5.036190509796143,
      "kl": 0.3603515625,
      "learning_rate": 7.507692307692307e-07,
      "loss": 0.0624,
      "reward": 0.7821961343288422,
      "reward_std": 0.03084599319845438,
      "reward_std_all": 0.030136998742818832,
      "rewards-max/forensics_sinkhorn_reward": 0.9544697999954224,
      "rewards-max/forensics_strict_edge_reward": 0.19615400582551956,
      "rewards-max/forensics_tiou_reward": 0.9179760217666626,
      "rewards/forensics_sinkhorn_reward": 0.7174319326877594,
      "rewards/forensics_strict_edge_reward": 0.014415203593671322,
      "rewards/forensics_tiou_reward": 0.5121301561594009,
      "step": 325
    },
    {
      "completion_length": 25.0,
      "epoch": 1.0,
      "grad_norm": 4.069174766540527,
      "kl": 0.275390625,
      "learning_rate": 7.5e-07,
      "loss": -0.0174,
      "reward": 0.44018441438674927,
      "reward_std": 0.051772601902484894,
      "reward_std_all": 0.22223301231861115,
      "rewards-max/forensics_sinkhorn_reward": 0.9244433045387268,
      "rewards-max/forensics_strict_edge_reward": 0.056888237595558167,
      "rewards-max/forensics_tiou_reward": 0.8361986875534058,
      "rewards/forensics_sinkhorn_reward": 0.4116998314857483,
      "rewards/forensics_strict_edge_reward": 0.011699064634740353,
      "rewards/forensics_tiou_reward": 0.3060425817966461,
      "step": 326
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.0030721966205838,
      "grad_norm": 6.5232253074646,
      "kl": 0.3828125,
      "learning_rate": 7.492307692307692e-07,
      "loss": 0.0556,
      "reward": 0.7712801098823547,
      "reward_std": 0.030752409133128822,
      "reward_std_all": 0.03291882295161486,
      "rewards-max/forensics_sinkhorn_reward": 0.9838407039642334,
      "rewards-max/forensics_strict_edge_reward": 0.48733407258987427,
      "rewards-max/forensics_tiou_reward": 0.983500748872757,
      "rewards/forensics_sinkhorn_reward": 0.6842014193534851,
      "rewards/forensics_strict_edge_reward": 0.026083704084157944,
      "rewards/forensics_tiou_reward": 0.45335978269577026,
      "step": 327
    },
    {
      "completion_length": 26.78125,
      "epoch": 1.0061443932411673,
      "grad_norm": 5.25710916519165,
      "kl": 0.322265625,
      "learning_rate": 7.484615384615384e-07,
      "loss": 0.1038,
      "reward": 0.7628179788589478,
      "reward_std": 0.05957613326609135,
      "reward_std_all": 0.056621573865413666,
      "rewards-max/forensics_sinkhorn_reward": 0.9502261877059937,
      "rewards-max/forensics_strict_edge_reward": 0.42908311635255814,
      "rewards-max/forensics_tiou_reward": 0.8954010903835297,
      "rewards/forensics_sinkhorn_reward": 0.6583039462566376,
      "rewards/forensics_strict_edge_reward": 0.03466776758432388,
      "rewards/forensics_tiou_reward": 0.4165460169315338,
      "step": 328
    },
    {
      "completion_length": 22.9375,
      "epoch": 1.0092165898617511,
      "grad_norm": 5.038679122924805,
      "kl": 0.3603515625,
      "learning_rate": 7.476923076923077e-07,
      "loss": 0.122,
      "reward": 0.7356495261192322,
      "reward_std": 0.07443144544959068,
      "reward_std_all": 0.07357639446854591,
      "rewards-max/forensics_sinkhorn_reward": 0.9762657284736633,
      "rewards-max/forensics_strict_edge_reward": 0.6109570264816284,
      "rewards-max/forensics_tiou_reward": 0.9629242420196533,
      "rewards/forensics_sinkhorn_reward": 0.6521379351615906,
      "rewards/forensics_strict_edge_reward": 0.03693399112671614,
      "rewards/forensics_tiou_reward": 0.4328727424144745,
      "step": 329
    },
    {
      "completion_length": 21.375,
      "epoch": 1.012288786482335,
      "grad_norm": 5.299869537353516,
      "kl": 0.2724609375,
      "learning_rate": 7.469230769230769e-07,
      "loss": 0.0757,
      "reward": 0.7693981528282166,
      "reward_std": 0.04655118752270937,
      "reward_std_all": 0.044595527462661266,
      "rewards-max/forensics_sinkhorn_reward": 0.9744019210338593,
      "rewards-max/forensics_strict_edge_reward": 0.5277008209377527,
      "rewards-max/forensics_tiou_reward": 0.9635161459445953,
      "rewards/forensics_sinkhorn_reward": 0.7558472156524658,
      "rewards/forensics_strict_edge_reward": 0.024635709589347243,
      "rewards/forensics_tiou_reward": 0.5182502567768097,
      "step": 330
    },
    {
      "completion_length": 25.765625,
      "epoch": 1.0153609831029187,
      "grad_norm": 5.484259128570557,
      "kl": 0.453125,
      "learning_rate": 7.461538461538461e-07,
      "loss": 0.082,
      "reward": 0.7536516487598419,
      "reward_std": 0.06267036125063896,
      "reward_std_all": 0.06006099842488766,
      "rewards-max/forensics_sinkhorn_reward": 0.9628370702266693,
      "rewards-max/forensics_strict_edge_reward": 0.3640780746936798,
      "rewards-max/forensics_tiou_reward": 0.9381213486194611,
      "rewards/forensics_sinkhorn_reward": 0.6722725629806519,
      "rewards/forensics_strict_edge_reward": 0.022430481389164925,
      "rewards/forensics_tiou_reward": 0.5052646696567535,
      "step": 331
    },
    {
      "completion_length": 25.8125,
      "epoch": 1.0184331797235022,
      "grad_norm": 4.708318710327148,
      "kl": 0.326171875,
      "learning_rate": 7.453846153846154e-07,
      "loss": 0.015,
      "reward": 0.7160808742046356,
      "reward_std": 0.04387286212295294,
      "reward_std_all": 0.06696613319218159,
      "rewards-max/forensics_sinkhorn_reward": 0.9577760994434357,
      "rewards-max/forensics_strict_edge_reward": 0.4099828153848648,
      "rewards-max/forensics_tiou_reward": 0.9090525209903717,
      "rewards/forensics_sinkhorn_reward": 0.6275877356529236,
      "rewards/forensics_strict_edge_reward": 0.03354858327656984,
      "rewards/forensics_tiou_reward": 0.4662575423717499,
      "step": 332
    },
    {
      "completion_length": 22.9375,
      "epoch": 1.021505376344086,
      "grad_norm": 5.579792499542236,
      "kl": 0.26123046875,
      "learning_rate": 7.446153846153845e-07,
      "loss": 0.0594,
      "reward": 0.7379709780216217,
      "reward_std": 0.066835792735219,
      "reward_std_all": 0.06720786169171333,
      "rewards-max/forensics_sinkhorn_reward": 0.97389155626297,
      "rewards-max/forensics_strict_edge_reward": 0.5190997421741486,
      "rewards-max/forensics_tiou_reward": 0.9622214436531067,
      "rewards/forensics_sinkhorn_reward": 0.623563677072525,
      "rewards/forensics_strict_edge_reward": 0.03136863652616739,
      "rewards/forensics_tiou_reward": 0.3779880553483963,
      "step": 333
    },
    {
      "completion_length": 25.796875,
      "epoch": 1.0245775729646698,
      "grad_norm": 5.789605140686035,
      "kl": 0.330078125,
      "learning_rate": 7.438461538461538e-07,
      "loss": 0.1087,
      "reward": 0.7010677456855774,
      "reward_std": 0.08983528986573219,
      "reward_std_all": 0.09354599192738533,
      "rewards-max/forensics_sinkhorn_reward": 0.8180283904075623,
      "rewards-max/forensics_strict_edge_reward": 0.4531010836362839,
      "rewards-max/forensics_tiou_reward": 0.711036205291748,
      "rewards/forensics_sinkhorn_reward": 0.5501611232757568,
      "rewards/forensics_strict_edge_reward": 0.023059600032866,
      "rewards/forensics_tiou_reward": 0.29221629351377487,
      "step": 334
    },
    {
      "completion_length": 25.375,
      "epoch": 1.0276497695852536,
      "grad_norm": 4.782975673675537,
      "kl": 0.2802734375,
      "learning_rate": 7.43076923076923e-07,
      "loss": 0.0777,
      "reward": 0.7404308021068573,
      "reward_std": 0.06400563940405846,
      "reward_std_all": 0.06400476396083832,
      "rewards-max/forensics_sinkhorn_reward": 0.9627240002155304,
      "rewards-max/forensics_strict_edge_reward": 0.2864861562848091,
      "rewards-max/forensics_tiou_reward": 0.9362058341503143,
      "rewards/forensics_sinkhorn_reward": 0.635664165019989,
      "rewards/forensics_strict_edge_reward": 0.01623094640672207,
      "rewards/forensics_tiou_reward": 0.5063251554965973,
      "step": 335
    },
    {
      "completion_length": 23.25,
      "epoch": 1.0307219662058371,
      "grad_norm": 5.685061931610107,
      "kl": 0.3203125,
      "learning_rate": 7.423076923076923e-07,
      "loss": 0.1195,
      "reward": 0.7284466326236725,
      "reward_std": 0.06436601281166077,
      "reward_std_all": 0.06888153962790966,
      "rewards-max/forensics_sinkhorn_reward": 0.9446777999401093,
      "rewards-max/forensics_strict_edge_reward": 0.1446031853556633,
      "rewards-max/forensics_tiou_reward": 0.8914779424667358,
      "rewards/forensics_sinkhorn_reward": 0.6165918409824371,
      "rewards/forensics_strict_edge_reward": 0.010336792096495628,
      "rewards/forensics_tiou_reward": 0.3138621598482132,
      "step": 336
    },
    {
      "completion_length": 26.59375,
      "epoch": 1.033794162826421,
      "grad_norm": 4.454103946685791,
      "kl": 0.2998046875,
      "learning_rate": 7.415384615384616e-07,
      "loss": -0.0384,
      "reward": 0.6933217346668243,
      "reward_std": 0.06666751392185688,
      "reward_std_all": 0.08589470200240612,
      "rewards-max/forensics_sinkhorn_reward": 0.9558619558811188,
      "rewards-max/forensics_strict_edge_reward": 0.3705078065395355,
      "rewards-max/forensics_tiou_reward": 0.9107231795787811,
      "rewards/forensics_sinkhorn_reward": 0.6231594383716583,
      "rewards/forensics_strict_edge_reward": 0.027164621278643608,
      "rewards/forensics_tiou_reward": 0.41683730483055115,
      "step": 337
    },
    {
      "completion_length": 24.234375,
      "epoch": 1.0368663594470047,
      "grad_norm": 5.521909713745117,
      "kl": 0.26904296875,
      "learning_rate": 7.407692307692307e-07,
      "loss": 0.0397,
      "reward": 0.7719112038612366,
      "reward_std": 0.02923220582306385,
      "reward_std_all": 0.03448585234582424,
      "rewards-max/forensics_sinkhorn_reward": 0.9668638408184052,
      "rewards-max/forensics_strict_edge_reward": 0.40730567276477814,
      "rewards-max/forensics_tiou_reward": 0.9463289082050323,
      "rewards/forensics_sinkhorn_reward": 0.6916154026985168,
      "rewards/forensics_strict_edge_reward": 0.028762402944266796,
      "rewards/forensics_tiou_reward": 0.4584682881832123,
      "step": 338
    },
    {
      "completion_length": 21.84375,
      "epoch": 1.0399385560675882,
      "grad_norm": 4.683046817779541,
      "kl": 0.2861328125,
      "learning_rate": 7.4e-07,
      "loss": -0.0209,
      "reward": 0.7633624374866486,
      "reward_std": 0.0333536472171545,
      "reward_std_all": 0.03740416653454304,
      "rewards-max/forensics_sinkhorn_reward": 0.9387861490249634,
      "rewards-max/forensics_strict_edge_reward": 0.21407660469412804,
      "rewards-max/forensics_tiou_reward": 0.8717910051345825,
      "rewards/forensics_sinkhorn_reward": 0.6821276843547821,
      "rewards/forensics_strict_edge_reward": 0.018129590898752213,
      "rewards/forensics_tiou_reward": 0.3962770998477936,
      "step": 339
    },
    {
      "completion_length": 21.375,
      "epoch": 1.043010752688172,
      "grad_norm": 5.464211463928223,
      "kl": 0.3310546875,
      "learning_rate": 7.392307692307692e-07,
      "loss": 0.096,
      "reward": 0.7111254334449768,
      "reward_std": 0.08049963414669037,
      "reward_std_all": 0.0897342823445797,
      "rewards-max/forensics_sinkhorn_reward": 0.9054564237594604,
      "rewards-max/forensics_strict_edge_reward": 0.22546802199212834,
      "rewards-max/forensics_tiou_reward": 0.8063730597496033,
      "rewards/forensics_sinkhorn_reward": 0.6194252073764801,
      "rewards/forensics_strict_edge_reward": 0.009734266313898843,
      "rewards/forensics_tiou_reward": 0.32043755054473877,
      "step": 340
    },
    {
      "completion_length": 23.53125,
      "epoch": 1.0460829493087558,
      "grad_norm": 4.409336090087891,
      "kl": 0.3486328125,
      "learning_rate": 7.384615384615384e-07,
      "loss": 0.0313,
      "reward": 0.7153233289718628,
      "reward_std": 0.0652081985026598,
      "reward_std_all": 0.07795758731663227,
      "rewards-max/forensics_sinkhorn_reward": 0.9618370532989502,
      "rewards-max/forensics_strict_edge_reward": 0.20259803161025047,
      "rewards-max/forensics_tiou_reward": 0.9291830360889435,
      "rewards/forensics_sinkhorn_reward": 0.6623166799545288,
      "rewards/forensics_strict_edge_reward": 0.014027952216565609,
      "rewards/forensics_tiou_reward": 0.4122554361820221,
      "step": 341
    },
    {
      "completion_length": 24.296875,
      "epoch": 1.0491551459293396,
      "grad_norm": 5.136780261993408,
      "kl": 0.26123046875,
      "learning_rate": 7.376923076923076e-07,
      "loss": 0.189,
      "reward": 0.7160176634788513,
      "reward_std": 0.09791762009263039,
      "reward_std_all": 0.09548680298030376,
      "rewards-max/forensics_sinkhorn_reward": 0.9787740111351013,
      "rewards-max/forensics_strict_edge_reward": 0.5244058072566986,
      "rewards-max/forensics_tiou_reward": 0.9707050323486328,
      "rewards/forensics_sinkhorn_reward": 0.6029734313488007,
      "rewards/forensics_strict_edge_reward": 0.04261682741343975,
      "rewards/forensics_tiou_reward": 0.3943108767271042,
      "step": 342
    },
    {
      "completion_length": 26.78125,
      "epoch": 1.0522273425499231,
      "grad_norm": 4.6874871253967285,
      "kl": 0.291015625,
      "learning_rate": 7.369230769230769e-07,
      "loss": 0.1562,
      "reward": 0.7412816882133484,
      "reward_std": 0.06380273960530758,
      "reward_std_all": 0.0645313523709774,
      "rewards-max/forensics_sinkhorn_reward": 0.9351920187473297,
      "rewards-max/forensics_strict_edge_reward": 0.3036012724041939,
      "rewards-max/forensics_tiou_reward": 0.8675690591335297,
      "rewards/forensics_sinkhorn_reward": 0.646982729434967,
      "rewards/forensics_strict_edge_reward": 0.024395876098424196,
      "rewards/forensics_tiou_reward": 0.4183529168367386,
      "step": 343
    },
    {
      "completion_length": 28.890625,
      "epoch": 1.055299539170507,
      "grad_norm": 3.7175822257995605,
      "kl": 0.501953125,
      "learning_rate": 7.361538461538461e-07,
      "loss": 0.1325,
      "reward": 0.7415932416915894,
      "reward_std": 0.06510379537940025,
      "reward_std_all": 0.06638904847204685,
      "rewards-max/forensics_sinkhorn_reward": 0.9612990021705627,
      "rewards-max/forensics_strict_edge_reward": 0.5436826646327972,
      "rewards-max/forensics_tiou_reward": 0.9303955435752869,
      "rewards/forensics_sinkhorn_reward": 0.6485432386398315,
      "rewards/forensics_strict_edge_reward": 0.03928843419998884,
      "rewards/forensics_tiou_reward": 0.4391590505838394,
      "step": 344
    },
    {
      "completion_length": 27.140625,
      "epoch": 1.0583717357910907,
      "grad_norm": 5.1129865646362305,
      "kl": 0.3310546875,
      "learning_rate": 7.353846153846154e-07,
      "loss": 0.1477,
      "reward": 0.7367787063121796,
      "reward_std": 0.06722080148756504,
      "reward_std_all": 0.06770892813801765,
      "rewards-max/forensics_sinkhorn_reward": 0.9586937129497528,
      "rewards-max/forensics_strict_edge_reward": 0.13954956084489822,
      "rewards-max/forensics_tiou_reward": 0.9171761274337769,
      "rewards/forensics_sinkhorn_reward": 0.6044915914535522,
      "rewards/forensics_strict_edge_reward": 0.008864706847816706,
      "rewards/forensics_tiou_reward": 0.39603403210639954,
      "step": 345
    },
    {
      "completion_length": 24.125,
      "epoch": 1.0614439324116745,
      "grad_norm": 4.952395915985107,
      "kl": 0.2880859375,
      "learning_rate": 7.346153846153846e-07,
      "loss": 0.1941,
      "reward": 0.7359858751296997,
      "reward_std": 0.054730867967009544,
      "reward_std_all": 0.06280291080474854,
      "rewards-max/forensics_sinkhorn_reward": 0.9370302259922028,
      "rewards-max/forensics_strict_edge_reward": 0.06868258118629456,
      "rewards-max/forensics_tiou_reward": 0.8544569909572601,
      "rewards/forensics_sinkhorn_reward": 0.6595478057861328,
      "rewards/forensics_strict_edge_reward": 0.0036994002293795347,
      "rewards/forensics_tiou_reward": 0.37962231040000916,
      "step": 346
    },
    {
      "completion_length": 28.015625,
      "epoch": 1.064516129032258,
      "grad_norm": 5.158090591430664,
      "kl": 0.28515625,
      "learning_rate": 7.338461538461538e-07,
      "loss": 0.1276,
      "reward": 0.6773715913295746,
      "reward_std": 0.07384410593658686,
      "reward_std_all": 0.10055441595613956,
      "rewards-max/forensics_sinkhorn_reward": 0.9232003688812256,
      "rewards-max/forensics_strict_edge_reward": 0.16181171499192715,
      "rewards-max/forensics_tiou_reward": 0.852808028459549,
      "rewards/forensics_sinkhorn_reward": 0.5861642807722092,
      "rewards/forensics_strict_edge_reward": 0.007685899268835783,
      "rewards/forensics_tiou_reward": 0.35463153570890427,
      "step": 347
    },
    {
      "completion_length": 26.765625,
      "epoch": 1.0675883256528418,
      "grad_norm": 3.882765054702759,
      "kl": 0.234375,
      "learning_rate": 7.33076923076923e-07,
      "loss": 0.1063,
      "reward": 0.7268039882183075,
      "reward_std": 0.05554939806461334,
      "reward_std_all": 0.06721842661499977,
      "rewards-max/forensics_sinkhorn_reward": 0.9721307158470154,
      "rewards-max/forensics_strict_edge_reward": 0.522010862827301,
      "rewards-max/forensics_tiou_reward": 0.9526381194591522,
      "rewards/forensics_sinkhorn_reward": 0.631500780582428,
      "rewards/forensics_strict_edge_reward": 0.04995128698647022,
      "rewards/forensics_tiou_reward": 0.4256672263145447,
      "step": 348
    },
    {
      "completion_length": 24.421875,
      "epoch": 1.0706605222734256,
      "grad_norm": 5.105747699737549,
      "kl": 0.3095703125,
      "learning_rate": 7.323076923076923e-07,
      "loss": 0.0365,
      "reward": 0.6963982284069061,
      "reward_std": 0.03197315987199545,
      "reward_std_all": 0.07267910428345203,
      "rewards-max/forensics_sinkhorn_reward": 0.9648174047470093,
      "rewards-max/forensics_strict_edge_reward": 0.1689353547990322,
      "rewards-max/forensics_tiou_reward": 0.9346980154514313,
      "rewards/forensics_sinkhorn_reward": 0.6361867189407349,
      "rewards/forensics_strict_edge_reward": 0.010626958217471838,
      "rewards/forensics_tiou_reward": 0.43073514103889465,
      "step": 349
    },
    {
      "completion_length": 25.0625,
      "epoch": 1.0737327188940091,
      "grad_norm": 4.905978202819824,
      "kl": 0.259765625,
      "learning_rate": 7.315384615384616e-07,
      "loss": 0.0436,
      "reward": 0.7065085470676422,
      "reward_std": 0.10346878692507744,
      "reward_std_all": 0.10097235813736916,
      "rewards-max/forensics_sinkhorn_reward": 0.9783202409744263,
      "rewards-max/forensics_strict_edge_reward": 0.09861540421843529,
      "rewards-max/forensics_tiou_reward": 0.9609129130840302,
      "rewards/forensics_sinkhorn_reward": 0.6104743778705597,
      "rewards/forensics_strict_edge_reward": 0.007709207944571972,
      "rewards/forensics_tiou_reward": 0.4324275702238083,
      "step": 350
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.076804915514593,
      "grad_norm": 4.433413028717041,
      "kl": 0.3115234375,
      "learning_rate": 7.307692307692307e-07,
      "loss": 0.1109,
      "reward": 0.7025966942310333,
      "reward_std": 0.06721393205225468,
      "reward_std_all": 0.08544540032744408,
      "rewards-max/forensics_sinkhorn_reward": 0.9646797776222229,
      "rewards-max/forensics_strict_edge_reward": 0.49555233120918274,
      "rewards-max/forensics_tiou_reward": 0.9421854913234711,
      "rewards/forensics_sinkhorn_reward": 0.6031790375709534,
      "rewards/forensics_strict_edge_reward": 0.03571083955466747,
      "rewards/forensics_tiou_reward": 0.39160768687725067,
      "step": 351
    },
    {
      "completion_length": 24.015625,
      "epoch": 1.0798771121351767,
      "grad_norm": 4.8089599609375,
      "kl": 0.333984375,
      "learning_rate": 7.3e-07,
      "loss": 0.0391,
      "reward": 0.7563628256320953,
      "reward_std": 0.06435124948620796,
      "reward_std_all": 0.06037097983062267,
      "rewards-max/forensics_sinkhorn_reward": 0.9489348828792572,
      "rewards-max/forensics_strict_edge_reward": 0.30516481399536133,
      "rewards-max/forensics_tiou_reward": 0.8985392153263092,
      "rewards/forensics_sinkhorn_reward": 0.6820250749588013,
      "rewards/forensics_strict_edge_reward": 0.04039867501705885,
      "rewards/forensics_tiou_reward": 0.47216424345970154,
      "step": 352
    },
    {
      "completion_length": 26.25,
      "epoch": 1.0829493087557605,
      "grad_norm": 4.0860981941223145,
      "kl": 0.2646484375,
      "learning_rate": 7.292307692307692e-07,
      "loss": 0.1263,
      "reward": 0.7166531085968018,
      "reward_std": 0.06677178107202053,
      "reward_std_all": 0.07805014215409756,
      "rewards-max/forensics_sinkhorn_reward": 0.9618272483348846,
      "rewards-max/forensics_strict_edge_reward": 0.3998595178127289,
      "rewards-max/forensics_tiou_reward": 0.9320651888847351,
      "rewards/forensics_sinkhorn_reward": 0.5770215690135956,
      "rewards/forensics_strict_edge_reward": 0.036408218555152416,
      "rewards/forensics_tiou_reward": 0.4066161811351776,
      "step": 353
    },
    {
      "completion_length": 23.671875,
      "epoch": 1.086021505376344,
      "grad_norm": 4.869163990020752,
      "kl": 0.2626953125,
      "learning_rate": 7.284615384615384e-07,
      "loss": 0.012,
      "reward": 0.7123892307281494,
      "reward_std": 0.08798769488930702,
      "reward_std_all": 0.089122474193573,
      "rewards-max/forensics_sinkhorn_reward": 0.9612850248813629,
      "rewards-max/forensics_strict_edge_reward": 0.31747178733348846,
      "rewards-max/forensics_tiou_reward": 0.931983232498169,
      "rewards/forensics_sinkhorn_reward": 0.6150238215923309,
      "rewards/forensics_strict_edge_reward": 0.02275478048250079,
      "rewards/forensics_tiou_reward": 0.35163021087646484,
      "step": 354
    },
    {
      "completion_length": 27.75,
      "epoch": 1.0890937019969278,
      "grad_norm": 4.148290157318115,
      "kl": 0.3134765625,
      "learning_rate": 7.276923076923076e-07,
      "loss": -0.0526,
      "reward": 0.7519903182983398,
      "reward_std": 0.037276883609592915,
      "reward_std_all": 0.045222291722893715,
      "rewards-max/forensics_sinkhorn_reward": 0.9738377630710602,
      "rewards-max/forensics_strict_edge_reward": 0.3468516580760479,
      "rewards-max/forensics_tiou_reward": 0.9552290439605713,
      "rewards/forensics_sinkhorn_reward": 0.6206185817718506,
      "rewards/forensics_strict_edge_reward": 0.01517688948661089,
      "rewards/forensics_tiou_reward": 0.44169439375400543,
      "step": 355
    },
    {
      "completion_length": 23.671875,
      "epoch": 1.0921658986175116,
      "grad_norm": 5.239220142364502,
      "kl": 0.302734375,
      "learning_rate": 7.269230769230769e-07,
      "loss": 0.1235,
      "reward": 0.7407229244709015,
      "reward_std": 0.0855274423956871,
      "reward_std_all": 0.07923310995101929,
      "rewards-max/forensics_sinkhorn_reward": 0.968146413564682,
      "rewards-max/forensics_strict_edge_reward": 0.5789424777030945,
      "rewards-max/forensics_tiou_reward": 0.9509422779083252,
      "rewards/forensics_sinkhorn_reward": 0.6651403605937958,
      "rewards/forensics_strict_edge_reward": 0.048447477631270885,
      "rewards/forensics_tiou_reward": 0.47757217288017273,
      "step": 356
    },
    {
      "completion_length": 29.9375,
      "epoch": 1.0952380952380953,
      "grad_norm": 4.3488335609436035,
      "kl": 0.3701171875,
      "learning_rate": 7.261538461538461e-07,
      "loss": 0.1431,
      "reward": 0.6894685924053192,
      "reward_std": 0.10006356798112392,
      "reward_std_all": 0.10732712224125862,
      "rewards-max/forensics_sinkhorn_reward": 0.9838669896125793,
      "rewards-max/forensics_strict_edge_reward": 0.39325864613056183,
      "rewards-max/forensics_tiou_reward": 0.9853071868419647,
      "rewards/forensics_sinkhorn_reward": 0.5608572661876678,
      "rewards/forensics_strict_edge_reward": 0.01979593187570572,
      "rewards/forensics_tiou_reward": 0.3417194336652756,
      "step": 357
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.098310291858679,
      "grad_norm": 6.490664482116699,
      "kl": 0.5751953125,
      "learning_rate": 7.253846153846154e-07,
      "loss": 0.1711,
      "reward": 0.7393988966941833,
      "reward_std": 0.07757123932242393,
      "reward_std_all": 0.07401640340685844,
      "rewards-max/forensics_sinkhorn_reward": 0.9645118713378906,
      "rewards-max/forensics_strict_edge_reward": 0.2745785266160965,
      "rewards-max/forensics_tiou_reward": 0.9420419335365295,
      "rewards/forensics_sinkhorn_reward": 0.6599747836589813,
      "rewards/forensics_strict_edge_reward": 0.021996992640197277,
      "rewards/forensics_tiou_reward": 0.44993801414966583,
      "step": 358
    },
    {
      "completion_length": 24.171875,
      "epoch": 1.1013824884792627,
      "grad_norm": 5.226527214050293,
      "kl": 0.275390625,
      "learning_rate": 7.246153846153846e-07,
      "loss": 0.1368,
      "reward": 0.6995422840118408,
      "reward_std": 0.09943197667598724,
      "reward_std_all": 0.10062237456440926,
      "rewards-max/forensics_sinkhorn_reward": 0.9528496861457825,
      "rewards-max/forensics_strict_edge_reward": 0.3751096725463867,
      "rewards-max/forensics_tiou_reward": 0.8929374814033508,
      "rewards/forensics_sinkhorn_reward": 0.6273287236690521,
      "rewards/forensics_strict_edge_reward": 0.024141822010278702,
      "rewards/forensics_tiou_reward": 0.41951027512550354,
      "step": 359
    },
    {
      "completion_length": 26.4375,
      "epoch": 1.1044546850998465,
      "grad_norm": 5.545285224914551,
      "kl": 0.3076171875,
      "learning_rate": 7.238461538461538e-07,
      "loss": 0.0917,
      "reward": 0.70267054438591,
      "reward_std": 0.08528693299740553,
      "reward_std_all": 0.09686724096536636,
      "rewards-max/forensics_sinkhorn_reward": 0.9706766605377197,
      "rewards-max/forensics_strict_edge_reward": 0.2183932140469551,
      "rewards-max/forensics_tiou_reward": 0.9479299187660217,
      "rewards/forensics_sinkhorn_reward": 0.6001707315444946,
      "rewards/forensics_strict_edge_reward": 0.015928265172988176,
      "rewards/forensics_tiou_reward": 0.42993272840976715,
      "step": 360
    },
    {
      "completion_length": 24.484375,
      "epoch": 1.10752688172043,
      "grad_norm": 5.093259334564209,
      "kl": 0.330078125,
      "learning_rate": 7.23076923076923e-07,
      "loss": 0.1387,
      "reward": 0.7729305922985077,
      "reward_std": 0.03346532303839922,
      "reward_std_all": 0.03393362648785114,
      "rewards-max/forensics_sinkhorn_reward": 0.9459232687950134,
      "rewards-max/forensics_strict_edge_reward": 0.13805102556943893,
      "rewards-max/forensics_tiou_reward": 0.8974866569042206,
      "rewards/forensics_sinkhorn_reward": 0.6844998598098755,
      "rewards/forensics_strict_edge_reward": 0.012213307432830334,
      "rewards/forensics_tiou_reward": 0.46188871562480927,
      "step": 361
    },
    {
      "completion_length": 23.890625,
      "epoch": 1.1105990783410138,
      "grad_norm": 4.057621955871582,
      "kl": 0.27001953125,
      "learning_rate": 7.223076923076922e-07,
      "loss": 0.2034,
      "reward": 0.7219972610473633,
      "reward_std": 0.07456196751445532,
      "reward_std_all": 0.08081941306591034,
      "rewards-max/forensics_sinkhorn_reward": 0.9775674045085907,
      "rewards-max/forensics_strict_edge_reward": 0.31592947617173195,
      "rewards-max/forensics_tiou_reward": 0.9687309861183167,
      "rewards/forensics_sinkhorn_reward": 0.6130233705043793,
      "rewards/forensics_strict_edge_reward": 0.02094413246959448,
      "rewards/forensics_tiou_reward": 0.39373330771923065,
      "step": 362
    },
    {
      "completion_length": 23.546875,
      "epoch": 1.1136712749615976,
      "grad_norm": 5.088776588439941,
      "kl": 0.2861328125,
      "learning_rate": 7.215384615384616e-07,
      "loss": 0.1121,
      "reward": 0.7287051677703857,
      "reward_std": 0.06449736282229424,
      "reward_std_all": 0.07367340847849846,
      "rewards-max/forensics_sinkhorn_reward": 0.970709353685379,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9565964341163635,
      "rewards/forensics_sinkhorn_reward": 0.6283666789531708,
      "rewards/forensics_strict_edge_reward": 0.04537720326334238,
      "rewards/forensics_tiou_reward": 0.4139246940612793,
      "step": 363
    },
    {
      "completion_length": 23.328125,
      "epoch": 1.1167434715821813,
      "grad_norm": 4.99839973449707,
      "kl": 0.3251953125,
      "learning_rate": 7.207692307692307e-07,
      "loss": 0.2121,
      "reward": 0.695686548948288,
      "reward_std": 0.05527090094983578,
      "reward_std_all": 0.08354153111577034,
      "rewards-max/forensics_sinkhorn_reward": 0.9202092587947845,
      "rewards-max/forensics_strict_edge_reward": 0.11389087233692408,
      "rewards-max/forensics_tiou_reward": 0.8450584411621094,
      "rewards/forensics_sinkhorn_reward": 0.5909688174724579,
      "rewards/forensics_strict_edge_reward": 0.008740717050386593,
      "rewards/forensics_tiou_reward": 0.3582640737295151,
      "step": 364
    },
    {
      "completion_length": 26.09375,
      "epoch": 1.119815668202765,
      "grad_norm": 5.129161834716797,
      "kl": 0.36328125,
      "learning_rate": 7.2e-07,
      "loss": 0.1639,
      "reward": 0.7037399411201477,
      "reward_std": 0.08035746961832047,
      "reward_std_all": 0.09247727319598198,
      "rewards-max/forensics_sinkhorn_reward": 0.9668709933757782,
      "rewards-max/forensics_strict_edge_reward": 0.31303603015840054,
      "rewards-max/forensics_tiou_reward": 0.925593376159668,
      "rewards/forensics_sinkhorn_reward": 0.6309108436107635,
      "rewards/forensics_strict_edge_reward": 0.021254288381896913,
      "rewards/forensics_tiou_reward": 0.41040678322315216,
      "step": 365
    },
    {
      "completion_length": 25.71875,
      "epoch": 1.1228878648233487,
      "grad_norm": 4.941572189331055,
      "kl": 0.4013671875,
      "learning_rate": 7.192307692307692e-07,
      "loss": -0.0623,
      "reward": 0.7582085728645325,
      "reward_std": 0.04573220573365688,
      "reward_std_all": 0.047686523757874966,
      "rewards-max/forensics_sinkhorn_reward": 0.9604351818561554,
      "rewards-max/forensics_strict_edge_reward": 0.3572814464569092,
      "rewards-max/forensics_tiou_reward": 0.9327906966209412,
      "rewards/forensics_sinkhorn_reward": 0.7052366137504578,
      "rewards/forensics_strict_edge_reward": 0.029929418116807938,
      "rewards/forensics_tiou_reward": 0.4658564627170563,
      "step": 366
    },
    {
      "completion_length": 20.96875,
      "epoch": 1.1259600614439325,
      "grad_norm": 5.082736968994141,
      "kl": 0.3828125,
      "learning_rate": 7.184615384615384e-07,
      "loss": 0.0987,
      "reward": 0.7329655885696411,
      "reward_std": 0.09003087505698204,
      "reward_std_all": 0.0848781168460846,
      "rewards-max/forensics_sinkhorn_reward": 0.9765724241733551,
      "rewards-max/forensics_strict_edge_reward": 0.38159448746591806,
      "rewards-max/forensics_tiou_reward": 0.9690037369728088,
      "rewards/forensics_sinkhorn_reward": 0.696137398481369,
      "rewards/forensics_strict_edge_reward": 0.019367964705452323,
      "rewards/forensics_tiou_reward": 0.46274295449256897,
      "step": 367
    },
    {
      "completion_length": 23.328125,
      "epoch": 1.129032258064516,
      "grad_norm": 6.1726603507995605,
      "kl": 0.3427734375,
      "learning_rate": 7.176923076923077e-07,
      "loss": 0.139,
      "reward": 0.7219104766845703,
      "reward_std": 0.09573372080922127,
      "reward_std_all": 0.09162085503339767,
      "rewards-max/forensics_sinkhorn_reward": 0.9494635760784149,
      "rewards-max/forensics_strict_edge_reward": 0.019192379899322987,
      "rewards-max/forensics_tiou_reward": 0.8979628682136536,
      "rewards/forensics_sinkhorn_reward": 0.6316624283790588,
      "rewards/forensics_strict_edge_reward": 0.0013652974157594144,
      "rewards/forensics_tiou_reward": 0.41045308113098145,
      "step": 368
    },
    {
      "completion_length": 27.9375,
      "epoch": 1.1321044546850998,
      "grad_norm": 5.365482330322266,
      "kl": 0.384765625,
      "learning_rate": 7.169230769230768e-07,
      "loss": 0.0856,
      "reward": 0.7752025425434113,
      "reward_std": 0.03727804124355316,
      "reward_std_all": 0.03660541772842407,
      "rewards-max/forensics_sinkhorn_reward": 0.947285920381546,
      "rewards-max/forensics_strict_edge_reward": 0.2694893293082714,
      "rewards-max/forensics_tiou_reward": 0.9010272026062012,
      "rewards/forensics_sinkhorn_reward": 0.6520871520042419,
      "rewards/forensics_strict_edge_reward": 0.01461270754225552,
      "rewards/forensics_tiou_reward": 0.43992146849632263,
      "step": 369
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.1351766513056836,
      "grad_norm": 5.9626994132995605,
      "kl": 0.416015625,
      "learning_rate": 7.161538461538461e-07,
      "loss": 0.0062,
      "reward": 0.7761496305465698,
      "reward_std": 0.04417762532830238,
      "reward_std_all": 0.041319320909678936,
      "rewards-max/forensics_sinkhorn_reward": 0.9539452791213989,
      "rewards-max/forensics_strict_edge_reward": 0.04582523554563522,
      "rewards-max/forensics_tiou_reward": 0.8863555490970612,
      "rewards/forensics_sinkhorn_reward": 0.6940065920352936,
      "rewards/forensics_strict_edge_reward": 0.003664318355731666,
      "rewards/forensics_tiou_reward": 0.4404832720756531,
      "step": 370
    },
    {
      "completion_length": 23.421875,
      "epoch": 1.1382488479262673,
      "grad_norm": 3.860517978668213,
      "kl": 0.28125,
      "learning_rate": 7.153846153846154e-07,
      "loss": 0.0046,
      "reward": 0.7535005211830139,
      "reward_std": 0.0449691116809845,
      "reward_std_all": 0.04851413518190384,
      "rewards-max/forensics_sinkhorn_reward": 0.9742431044578552,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9636968374252319,
      "rewards/forensics_sinkhorn_reward": 0.7060754299163818,
      "rewards/forensics_strict_edge_reward": 0.05455482751131058,
      "rewards/forensics_tiou_reward": 0.4943193793296814,
      "step": 371
    },
    {
      "completion_length": 26.78125,
      "epoch": 1.141321044546851,
      "grad_norm": 4.017478942871094,
      "kl": 0.2880859375,
      "learning_rate": 7.146153846153846e-07,
      "loss": 0.1192,
      "reward": 0.7533818185329437,
      "reward_std": 0.03007879899814725,
      "reward_std_all": 0.04355363501235843,
      "rewards-max/forensics_sinkhorn_reward": 0.9174283444881439,
      "rewards-max/forensics_strict_edge_reward": 0.2025199718773365,
      "rewards-max/forensics_tiou_reward": 0.8257502019405365,
      "rewards/forensics_sinkhorn_reward": 0.688584566116333,
      "rewards/forensics_strict_edge_reward": 0.011061883182264864,
      "rewards/forensics_tiou_reward": 0.428619921207428,
      "step": 372
    },
    {
      "completion_length": 24.140625,
      "epoch": 1.1443932411674347,
      "grad_norm": 5.919499397277832,
      "kl": 0.490234375,
      "learning_rate": 7.138461538461538e-07,
      "loss": 0.1552,
      "reward": 0.705357700586319,
      "reward_std": 0.07295587100088596,
      "reward_std_all": 0.08584576845169067,
      "rewards-max/forensics_sinkhorn_reward": 0.9370514750480652,
      "rewards-max/forensics_strict_edge_reward": 0.3040425323997624,
      "rewards-max/forensics_tiou_reward": 0.8559199273586273,
      "rewards/forensics_sinkhorn_reward": 0.5944815874099731,
      "rewards/forensics_strict_edge_reward": 0.01305701513774693,
      "rewards/forensics_tiou_reward": 0.3312510699033737,
      "step": 373
    },
    {
      "completion_length": 26.0,
      "epoch": 1.1474654377880185,
      "grad_norm": 3.932982921600342,
      "kl": 0.33203125,
      "learning_rate": 7.13076923076923e-07,
      "loss": 0.1487,
      "reward": 0.7783376276493073,
      "reward_std": 0.03525251895189285,
      "reward_std_all": 0.034027816727757454,
      "rewards-max/forensics_sinkhorn_reward": 0.9853339493274689,
      "rewards-max/forensics_strict_edge_reward": 0.75,
      "rewards-max/forensics_tiou_reward": 0.98786261677742,
      "rewards/forensics_sinkhorn_reward": 0.7302046418190002,
      "rewards/forensics_strict_edge_reward": 0.048993706703186035,
      "rewards/forensics_tiou_reward": 0.5458841025829315,
      "step": 374
    },
    {
      "completion_length": 25.265625,
      "epoch": 1.1505376344086022,
      "grad_norm": 5.023838996887207,
      "kl": 0.4384765625,
      "learning_rate": 7.123076923076922e-07,
      "loss": 0.1977,
      "reward": 0.6907939612865448,
      "reward_std": 0.06290065869688988,
      "reward_std_all": 0.08600348234176636,
      "rewards-max/forensics_sinkhorn_reward": 0.9285116791725159,
      "rewards-max/forensics_strict_edge_reward": 0.23452549171634018,
      "rewards-max/forensics_tiou_reward": 0.8611279428005219,
      "rewards/forensics_sinkhorn_reward": 0.5922646522521973,
      "rewards/forensics_strict_edge_reward": 0.010475275194039568,
      "rewards/forensics_tiou_reward": 0.3605249226093292,
      "step": 375
    },
    {
      "completion_length": 23.328125,
      "epoch": 1.1536098310291858,
      "grad_norm": 6.372043132781982,
      "kl": 0.369140625,
      "learning_rate": 7.115384615384616e-07,
      "loss": 0.1205,
      "reward": 0.7204229831695557,
      "reward_std": 0.0615121740847826,
      "reward_std_all": 0.07238712348043919,
      "rewards-max/forensics_sinkhorn_reward": 0.9725786745548248,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9560487866401672,
      "rewards/forensics_sinkhorn_reward": 0.6419077217578888,
      "rewards/forensics_strict_edge_reward": 0.029723948799073696,
      "rewards/forensics_tiou_reward": 0.3914182484149933,
      "step": 376
    },
    {
      "completion_length": 24.09375,
      "epoch": 1.1566820276497696,
      "grad_norm": 4.701287269592285,
      "kl": 0.3310546875,
      "learning_rate": 7.107692307692308e-07,
      "loss": 0.1255,
      "reward": 0.7552669644355774,
      "reward_std": 0.061202822253108025,
      "reward_std_all": 0.05869750306010246,
      "rewards-max/forensics_sinkhorn_reward": 0.9713727235794067,
      "rewards-max/forensics_strict_edge_reward": 0.38722455501556396,
      "rewards-max/forensics_tiou_reward": 0.9534231424331665,
      "rewards/forensics_sinkhorn_reward": 0.6514106690883636,
      "rewards/forensics_strict_edge_reward": 0.03485987242311239,
      "rewards/forensics_tiou_reward": 0.45600366592407227,
      "step": 377
    },
    {
      "completion_length": 24.75,
      "epoch": 1.1597542242703534,
      "grad_norm": 4.939821243286133,
      "kl": 0.3818359375,
      "learning_rate": 7.1e-07,
      "loss": 0.0817,
      "reward": 0.7673652768135071,
      "reward_std": 0.048149894922971725,
      "reward_std_all": 0.04598664492368698,
      "rewards-max/forensics_sinkhorn_reward": 0.9779700636863708,
      "rewards-max/forensics_strict_edge_reward": 0.5050012171268463,
      "rewards-max/forensics_tiou_reward": 0.969051867723465,
      "rewards/forensics_sinkhorn_reward": 0.7101665139198303,
      "rewards/forensics_strict_edge_reward": 0.04603123292326927,
      "rewards/forensics_tiou_reward": 0.5094492584466934,
      "step": 378
    },
    {
      "completion_length": 25.40625,
      "epoch": 1.1628264208909371,
      "grad_norm": 5.85036039352417,
      "kl": 0.365234375,
      "learning_rate": 7.092307692307692e-07,
      "loss": 0.092,
      "reward": 0.7663902640342712,
      "reward_std": 0.036326806992292404,
      "reward_std_all": 0.03822603262960911,
      "rewards-max/forensics_sinkhorn_reward": 0.9628797769546509,
      "rewards-max/forensics_strict_edge_reward": 0.06886708550155163,
      "rewards-max/forensics_tiou_reward": 0.9261695444583893,
      "rewards/forensics_sinkhorn_reward": 0.7007268667221069,
      "rewards/forensics_strict_edge_reward": 0.003919002832844853,
      "rewards/forensics_tiou_reward": 0.4784766137599945,
      "step": 379
    },
    {
      "completion_length": 22.421875,
      "epoch": 1.1658986175115207,
      "grad_norm": 5.556816577911377,
      "kl": 0.3134765625,
      "learning_rate": 7.084615384615384e-07,
      "loss": 0.035,
      "reward": 0.7150071263313293,
      "reward_std": 0.03239140287041664,
      "reward_std_all": 0.06048484705388546,
      "rewards-max/forensics_sinkhorn_reward": 0.9819843173027039,
      "rewards-max/forensics_strict_edge_reward": 0.22191549092531204,
      "rewards-max/forensics_tiou_reward": 0.9784660935401917,
      "rewards/forensics_sinkhorn_reward": 0.6192464530467987,
      "rewards/forensics_strict_edge_reward": 0.016065931878983974,
      "rewards/forensics_tiou_reward": 0.457990437746048,
      "step": 380
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.1689708141321045,
      "grad_norm": 5.523351669311523,
      "kl": 0.2685546875,
      "learning_rate": 7.076923076923077e-07,
      "loss": 0.0648,
      "reward": 0.7520072162151337,
      "reward_std": 0.06315061822533607,
      "reward_std_all": 0.06050817854702473,
      "rewards-max/forensics_sinkhorn_reward": 0.9643662869930267,
      "rewards-max/forensics_strict_edge_reward": 0.24938895553350449,
      "rewards-max/forensics_tiou_reward": 0.9320798814296722,
      "rewards/forensics_sinkhorn_reward": 0.7194646596908569,
      "rewards/forensics_strict_edge_reward": 0.01750128995627165,
      "rewards/forensics_tiou_reward": 0.5386775135993958,
      "step": 381
    },
    {
      "completion_length": 22.46875,
      "epoch": 1.1720430107526882,
      "grad_norm": 4.538328647613525,
      "kl": 0.328125,
      "learning_rate": 7.069230769230768e-07,
      "loss": -0.0623,
      "reward": 0.7562895715236664,
      "reward_std": 0.06022134516388178,
      "reward_std_all": 0.057988088577985764,
      "rewards-max/forensics_sinkhorn_reward": 0.9714804291725159,
      "rewards-max/forensics_strict_edge_reward": 0.2621621936559677,
      "rewards-max/forensics_tiou_reward": 0.9503844380378723,
      "rewards/forensics_sinkhorn_reward": 0.6637700498104095,
      "rewards/forensics_strict_edge_reward": 0.017959614284336567,
      "rewards/forensics_tiou_reward": 0.43948549032211304,
      "step": 382
    },
    {
      "completion_length": 26.265625,
      "epoch": 1.1751152073732718,
      "grad_norm": 4.784759044647217,
      "kl": 0.3486328125,
      "learning_rate": 7.061538461538462e-07,
      "loss": 0.0523,
      "reward": 0.7531720101833344,
      "reward_std": 0.03443547245115042,
      "reward_std_all": 0.045257849618792534,
      "rewards-max/forensics_sinkhorn_reward": 0.9499661922454834,
      "rewards-max/forensics_strict_edge_reward": 0.43089205026626587,
      "rewards-max/forensics_tiou_reward": 0.8994744718074799,
      "rewards/forensics_sinkhorn_reward": 0.6489244699478149,
      "rewards/forensics_strict_edge_reward": 0.0216276403516531,
      "rewards/forensics_tiou_reward": 0.4522920846939087,
      "step": 383
    },
    {
      "completion_length": 21.421875,
      "epoch": 1.1781874039938556,
      "grad_norm": 6.637493133544922,
      "kl": 0.5810546875,
      "learning_rate": 7.053846153846154e-07,
      "loss": 0.069,
      "reward": 0.7091512978076935,
      "reward_std": 0.03934021294116974,
      "reward_std_all": 0.06780919432640076,
      "rewards-max/forensics_sinkhorn_reward": 0.9696749150753021,
      "rewards-max/forensics_strict_edge_reward": 0.4676465690135956,
      "rewards-max/forensics_tiou_reward": 0.9507628977298737,
      "rewards/forensics_sinkhorn_reward": 0.6475930213928223,
      "rewards/forensics_strict_edge_reward": 0.028171148151159286,
      "rewards/forensics_tiou_reward": 0.4351295083761215,
      "step": 384
    },
    {
      "completion_length": 24.921875,
      "epoch": 1.1812596006144394,
      "grad_norm": 11.20231819152832,
      "kl": 0.744140625,
      "learning_rate": 7.046153846153846e-07,
      "loss": 0.0676,
      "reward": 0.7509405314922333,
      "reward_std": 0.05001312121748924,
      "reward_std_all": 0.051766835153102875,
      "rewards-max/forensics_sinkhorn_reward": 0.9711464047431946,
      "rewards-max/forensics_strict_edge_reward": 0.678628072142601,
      "rewards-max/forensics_tiou_reward": 0.950438916683197,
      "rewards/forensics_sinkhorn_reward": 0.6944758594036102,
      "rewards/forensics_strict_edge_reward": 0.05324988253414631,
      "rewards/forensics_tiou_reward": 0.4553184062242508,
      "step": 385
    },
    {
      "completion_length": 22.6875,
      "epoch": 1.1843317972350231,
      "grad_norm": 5.463379383087158,
      "kl": 0.494140625,
      "learning_rate": 7.038461538461539e-07,
      "loss": 0.1002,
      "reward": 0.7812439501285553,
      "reward_std": 0.035365186631679535,
      "reward_std_all": 0.03364601172506809,
      "rewards-max/forensics_sinkhorn_reward": 0.9730395972728729,
      "rewards-max/forensics_strict_edge_reward": 0.5003146827220917,
      "rewards-max/forensics_tiou_reward": 0.9557293355464935,
      "rewards/forensics_sinkhorn_reward": 0.7109757661819458,
      "rewards/forensics_strict_edge_reward": 0.04400166030973196,
      "rewards/forensics_tiou_reward": 0.4839387387037277,
      "step": 386
    },
    {
      "completion_length": 27.390625,
      "epoch": 1.1874039938556067,
      "grad_norm": 5.550500392913818,
      "kl": 0.4111328125,
      "learning_rate": 7.03076923076923e-07,
      "loss": 0.2079,
      "reward": 0.716901570558548,
      "reward_std": 0.06556586362421513,
      "reward_std_all": 0.07431020587682724,
      "rewards-max/forensics_sinkhorn_reward": 0.9609038531780243,
      "rewards-max/forensics_strict_edge_reward": 0.3157382905483246,
      "rewards-max/forensics_tiou_reward": 0.9357147514820099,
      "rewards/forensics_sinkhorn_reward": 0.5760548710823059,
      "rewards/forensics_strict_edge_reward": 0.03541812114417553,
      "rewards/forensics_tiou_reward": 0.413249209523201,
      "step": 387
    },
    {
      "completion_length": 24.28125,
      "epoch": 1.1904761904761905,
      "grad_norm": 5.215056896209717,
      "kl": 0.4677734375,
      "learning_rate": 7.023076923076922e-07,
      "loss": 0.2074,
      "reward": 0.703719973564148,
      "reward_std": 0.07468299753963947,
      "reward_std_all": 0.08637234754860401,
      "rewards-max/forensics_sinkhorn_reward": 0.970595121383667,
      "rewards-max/forensics_strict_edge_reward": 0.43434032052755356,
      "rewards-max/forensics_tiou_reward": 0.9527353644371033,
      "rewards/forensics_sinkhorn_reward": 0.6231361925601959,
      "rewards/forensics_strict_edge_reward": 0.02726274123415351,
      "rewards/forensics_tiou_reward": 0.3749818652868271,
      "step": 388
    },
    {
      "completion_length": 22.921875,
      "epoch": 1.1935483870967742,
      "grad_norm": 4.975463390350342,
      "kl": 0.423828125,
      "learning_rate": 7.015384615384615e-07,
      "loss": 0.0559,
      "reward": 0.7384157776832581,
      "reward_std": 0.05759632494300604,
      "reward_std_all": 0.0644400529563427,
      "rewards-max/forensics_sinkhorn_reward": 0.9611814320087433,
      "rewards-max/forensics_strict_edge_reward": 0.299142200499773,
      "rewards-max/forensics_tiou_reward": 0.9283141195774078,
      "rewards/forensics_sinkhorn_reward": 0.6468666195869446,
      "rewards/forensics_strict_edge_reward": 0.020544966217130423,
      "rewards/forensics_tiou_reward": 0.39290495216846466,
      "step": 389
    },
    {
      "completion_length": 24.265625,
      "epoch": 1.1966205837173578,
      "grad_norm": 7.106782913208008,
      "kl": 0.4951171875,
      "learning_rate": 7.007692307692308e-07,
      "loss": -0.0515,
      "reward": 0.736294150352478,
      "reward_std": 0.05435333214700222,
      "reward_std_all": 0.060308460146188736,
      "rewards-max/forensics_sinkhorn_reward": 0.9405150413513184,
      "rewards-max/forensics_strict_edge_reward": 0.5062990635633469,
      "rewards-max/forensics_tiou_reward": 0.8858324587345123,
      "rewards/forensics_sinkhorn_reward": 0.6249639391899109,
      "rewards/forensics_strict_edge_reward": 0.029889557510614395,
      "rewards/forensics_tiou_reward": 0.3744823634624481,
      "step": 390
    },
    {
      "completion_length": 24.84375,
      "epoch": 1.1996927803379416,
      "grad_norm": 5.523309707641602,
      "kl": 0.287109375,
      "learning_rate": 7e-07,
      "loss": -0.0776,
      "reward": 0.704420268535614,
      "reward_std": 0.05254830978810787,
      "reward_std_all": 0.07356040924787521,
      "rewards-max/forensics_sinkhorn_reward": 0.9628363251686096,
      "rewards-max/forensics_strict_edge_reward": 0.48901838064193726,
      "rewards-max/forensics_tiou_reward": 0.935145378112793,
      "rewards/forensics_sinkhorn_reward": 0.6342571079730988,
      "rewards/forensics_strict_edge_reward": 0.043482573702931404,
      "rewards/forensics_tiou_reward": 0.42010368406772614,
      "step": 391
    },
    {
      "completion_length": 23.859375,
      "epoch": 1.2027649769585254,
      "grad_norm": 6.066749095916748,
      "kl": 0.3740234375,
      "learning_rate": 6.992307692307692e-07,
      "loss": 0.0537,
      "reward": 0.7688207626342773,
      "reward_std": 0.026670812629163265,
      "reward_std_all": 0.03149584122002125,
      "rewards-max/forensics_sinkhorn_reward": 0.9754943251609802,
      "rewards-max/forensics_strict_edge_reward": 0.4813055247068405,
      "rewards-max/forensics_tiou_reward": 0.963564544916153,
      "rewards/forensics_sinkhorn_reward": 0.6665282845497131,
      "rewards/forensics_strict_edge_reward": 0.03635541349649429,
      "rewards/forensics_tiou_reward": 0.48163744807243347,
      "step": 392
    },
    {
      "completion_length": 22.609375,
      "epoch": 1.2058371735791091,
      "grad_norm": 4.888509273529053,
      "kl": 0.3095703125,
      "learning_rate": 6.984615384615384e-07,
      "loss": -0.0061,
      "reward": 0.7778407633304596,
      "reward_std": 0.04084134381264448,
      "reward_std_all": 0.03827979229390621,
      "rewards-max/forensics_sinkhorn_reward": 0.9451143145561218,
      "rewards-max/forensics_strict_edge_reward": 0.3607640452682972,
      "rewards-max/forensics_tiou_reward": 0.8915465474128723,
      "rewards/forensics_sinkhorn_reward": 0.7458031475543976,
      "rewards/forensics_strict_edge_reward": 0.04095881059765816,
      "rewards/forensics_tiou_reward": 0.5438811480998993,
      "step": 393
    },
    {
      "completion_length": 23.96875,
      "epoch": 1.2089093701996927,
      "grad_norm": 5.3575119972229,
      "kl": 0.3896484375,
      "learning_rate": 6.976923076923077e-07,
      "loss": 0.1771,
      "reward": 0.7496497631072998,
      "reward_std": 0.07394290715456009,
      "reward_std_all": 0.06881858594715595,
      "rewards-max/forensics_sinkhorn_reward": 0.9847067296504974,
      "rewards-max/forensics_strict_edge_reward": 0.7317707538604736,
      "rewards-max/forensics_tiou_reward": 0.9855985045433044,
      "rewards/forensics_sinkhorn_reward": 0.6718767583370209,
      "rewards/forensics_strict_edge_reward": 0.05525295436382294,
      "rewards/forensics_tiou_reward": 0.41968366503715515,
      "step": 394
    },
    {
      "completion_length": 26.21875,
      "epoch": 1.2119815668202765,
      "grad_norm": 5.300227165222168,
      "kl": 0.33984375,
      "learning_rate": 6.969230769230769e-07,
      "loss": 0.1489,
      "reward": 0.7527578771114349,
      "reward_std": 0.06059332750737667,
      "reward_std_all": 0.058348577469587326,
      "rewards-max/forensics_sinkhorn_reward": 0.9722802042961121,
      "rewards-max/forensics_strict_edge_reward": 0.3811161369085312,
      "rewards-max/forensics_tiou_reward": 0.9615190327167511,
      "rewards/forensics_sinkhorn_reward": 0.6648964881896973,
      "rewards/forensics_strict_edge_reward": 0.02523314766585827,
      "rewards/forensics_tiou_reward": 0.4375849813222885,
      "step": 395
    },
    {
      "completion_length": 22.34375,
      "epoch": 1.2150537634408602,
      "grad_norm": 6.180971145629883,
      "kl": 0.28125,
      "learning_rate": 6.961538461538461e-07,
      "loss": -0.0651,
      "reward": 0.7602989077568054,
      "reward_std": 0.035282169468700886,
      "reward_std_all": 0.04008125048130751,
      "rewards-max/forensics_sinkhorn_reward": 0.9843040108680725,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9835715889930725,
      "rewards/forensics_sinkhorn_reward": 0.7008366286754608,
      "rewards/forensics_strict_edge_reward": 0.032159700989723206,
      "rewards/forensics_tiou_reward": 0.47866207361221313,
      "step": 396
    },
    {
      "completion_length": 23.828125,
      "epoch": 1.218125960061444,
      "grad_norm": 6.453510284423828,
      "kl": 0.330078125,
      "learning_rate": 6.953846153846154e-07,
      "loss": 0.1041,
      "reward": 0.7631095349788666,
      "reward_std": 0.05868270806968212,
      "reward_std_all": 0.05491025373339653,
      "rewards-max/forensics_sinkhorn_reward": 0.8961812257766724,
      "rewards-max/forensics_strict_edge_reward": 0.24689825624227524,
      "rewards-max/forensics_tiou_reward": 0.7979996800422668,
      "rewards/forensics_sinkhorn_reward": 0.7054180204868317,
      "rewards/forensics_strict_edge_reward": 0.011307218810543418,
      "rewards/forensics_tiou_reward": 0.44562622904777527,
      "step": 397
    },
    {
      "completion_length": 23.734375,
      "epoch": 1.2211981566820276,
      "grad_norm": 5.939719200134277,
      "kl": 0.2763671875,
      "learning_rate": 6.946153846153846e-07,
      "loss": 0.0456,
      "reward": 0.7611210942268372,
      "reward_std": 0.03873504791408777,
      "reward_std_all": 0.042093995958566666,
      "rewards-max/forensics_sinkhorn_reward": 0.9423052072525024,
      "rewards-max/forensics_strict_edge_reward": 0.4006809741258621,
      "rewards-max/forensics_tiou_reward": 0.8892492949962616,
      "rewards/forensics_sinkhorn_reward": 0.6647128164768219,
      "rewards/forensics_strict_edge_reward": 0.02624394651502371,
      "rewards/forensics_tiou_reward": 0.40123072266578674,
      "step": 398
    },
    {
      "completion_length": 22.703125,
      "epoch": 1.2242703533026114,
      "grad_norm": 5.53760290145874,
      "kl": 0.41015625,
      "learning_rate": 6.938461538461539e-07,
      "loss": 0.1245,
      "reward": 0.7572589218616486,
      "reward_std": 0.06009731441736221,
      "reward_std_all": 0.056974858045578,
      "rewards-max/forensics_sinkhorn_reward": 0.9643122851848602,
      "rewards-max/forensics_strict_edge_reward": 0.09117192029953003,
      "rewards-max/forensics_tiou_reward": 0.929130882024765,
      "rewards/forensics_sinkhorn_reward": 0.6858052015304565,
      "rewards/forensics_strict_edge_reward": 0.0069029307924211025,
      "rewards/forensics_tiou_reward": 0.4471043199300766,
      "step": 399
    },
    {
      "completion_length": 22.828125,
      "epoch": 1.2273425499231951,
      "grad_norm": 4.785621643066406,
      "kl": 0.3681640625,
      "learning_rate": 6.93076923076923e-07,
      "loss": 0.0389,
      "reward": 0.7483276724815369,
      "reward_std": 0.02761043608188629,
      "reward_std_all": 0.044433388859033585,
      "rewards-max/forensics_sinkhorn_reward": 0.9282114505767822,
      "rewards-max/forensics_strict_edge_reward": 0.500002600481821,
      "rewards-max/forensics_tiou_reward": 0.8335831463336945,
      "rewards/forensics_sinkhorn_reward": 0.6657837331295013,
      "rewards/forensics_strict_edge_reward": 0.0325553690869782,
      "rewards/forensics_tiou_reward": 0.4052453935146332,
      "step": 400
    },
    {
      "completion_length": 23.078125,
      "epoch": 1.230414746543779,
      "grad_norm": 6.137651443481445,
      "kl": 0.2958984375,
      "learning_rate": 6.923076923076922e-07,
      "loss": 0.0602,
      "reward": 0.7636309564113617,
      "reward_std": 0.037921481765806675,
      "reward_std_all": 0.041101960465312004,
      "rewards-max/forensics_sinkhorn_reward": 0.9774194359779358,
      "rewards-max/forensics_strict_edge_reward": 0.33061711490154266,
      "rewards-max/forensics_tiou_reward": 0.9658393561840057,
      "rewards/forensics_sinkhorn_reward": 0.7125409543514252,
      "rewards/forensics_strict_edge_reward": 0.03484705276787281,
      "rewards/forensics_tiou_reward": 0.5020971596240997,
      "step": 401
    },
    {
      "completion_length": 24.6875,
      "epoch": 1.2334869431643625,
      "grad_norm": 5.091580390930176,
      "kl": 0.3095703125,
      "learning_rate": 6.915384615384615e-07,
      "loss": 0.1793,
      "reward": 0.7470583319664001,
      "reward_std": 0.06991340965032578,
      "reward_std_all": 0.06662004068493843,
      "rewards-max/forensics_sinkhorn_reward": 0.970445990562439,
      "rewards-max/forensics_strict_edge_reward": 0.29233212023973465,
      "rewards-max/forensics_tiou_reward": 0.9533126950263977,
      "rewards/forensics_sinkhorn_reward": 0.6274858117103577,
      "rewards/forensics_strict_edge_reward": 0.02356792939826846,
      "rewards/forensics_tiou_reward": 0.4208230823278427,
      "step": 402
    },
    {
      "completion_length": 24.34375,
      "epoch": 1.2365591397849462,
      "grad_norm": 5.328117370605469,
      "kl": 0.474609375,
      "learning_rate": 6.907692307692307e-07,
      "loss": 0.0433,
      "reward": 0.7650695443153381,
      "reward_std": 0.04774007387459278,
      "reward_std_all": 0.046236660331487656,
      "rewards-max/forensics_sinkhorn_reward": 0.9715365171432495,
      "rewards-max/forensics_strict_edge_reward": 0.49658530950546265,
      "rewards-max/forensics_tiou_reward": 0.955530971288681,
      "rewards/forensics_sinkhorn_reward": 0.6827006936073303,
      "rewards/forensics_strict_edge_reward": 0.038956498727202415,
      "rewards/forensics_tiou_reward": 0.4884827435016632,
      "step": 403
    },
    {
      "completion_length": 27.75,
      "epoch": 1.23963133640553,
      "grad_norm": 6.0358428955078125,
      "kl": 0.3037109375,
      "learning_rate": 6.9e-07,
      "loss": 0.0956,
      "reward": 0.7576004266738892,
      "reward_std": 0.04491964913904667,
      "reward_std_all": 0.04627121891826391,
      "rewards-max/forensics_sinkhorn_reward": 0.9518763720989227,
      "rewards-max/forensics_strict_edge_reward": 0.6714635640382767,
      "rewards-max/forensics_tiou_reward": 0.8971479535102844,
      "rewards/forensics_sinkhorn_reward": 0.67644202709198,
      "rewards/forensics_strict_edge_reward": 0.069273816421628,
      "rewards/forensics_tiou_reward": 0.4805741161108017,
      "step": 404
    },
    {
      "completion_length": 25.703125,
      "epoch": 1.2427035330261136,
      "grad_norm": 6.571341514587402,
      "kl": 0.30078125,
      "learning_rate": 6.892307692307692e-07,
      "loss": 0.0196,
      "reward": 0.7419091463088989,
      "reward_std": 0.05348331481218338,
      "reward_std_all": 0.05767982825636864,
      "rewards-max/forensics_sinkhorn_reward": 0.9629882872104645,
      "rewards-max/forensics_strict_edge_reward": 0.35705363750457764,
      "rewards-max/forensics_tiou_reward": 0.9322124719619751,
      "rewards/forensics_sinkhorn_reward": 0.6573598980903625,
      "rewards/forensics_strict_edge_reward": 0.025302057154476643,
      "rewards/forensics_tiou_reward": 0.4139716923236847,
      "step": 405
    },
    {
      "completion_length": 23.921875,
      "epoch": 1.2457757296466974,
      "grad_norm": 5.266321182250977,
      "kl": 0.361328125,
      "learning_rate": 6.884615384615384e-07,
      "loss": 0.0406,
      "reward": 0.7594382464885712,
      "reward_std": 0.055022455751895905,
      "reward_std_all": 0.05290442518889904,
      "rewards-max/forensics_sinkhorn_reward": 0.9494965970516205,
      "rewards-max/forensics_strict_edge_reward": 0.2578485757112503,
      "rewards-max/forensics_tiou_reward": 0.8999035954475403,
      "rewards/forensics_sinkhorn_reward": 0.6883430480957031,
      "rewards/forensics_strict_edge_reward": 0.02475346578285098,
      "rewards/forensics_tiou_reward": 0.46503427624702454,
      "step": 406
    },
    {
      "completion_length": 26.03125,
      "epoch": 1.2488479262672811,
      "grad_norm": 5.377367973327637,
      "kl": 0.427734375,
      "learning_rate": 6.876923076923077e-07,
      "loss": 0.0234,
      "reward": 0.7746008932590485,
      "reward_std": 0.04081245884299278,
      "reward_std_all": 0.03900137450546026,
      "rewards-max/forensics_sinkhorn_reward": 0.9739806950092316,
      "rewards-max/forensics_strict_edge_reward": 0.13309736549854279,
      "rewards-max/forensics_tiou_reward": 0.9620120227336884,
      "rewards/forensics_sinkhorn_reward": 0.7390490174293518,
      "rewards/forensics_strict_edge_reward": 0.01592549867928028,
      "rewards/forensics_tiou_reward": 0.5781500935554504,
      "step": 407
    },
    {
      "completion_length": 26.6875,
      "epoch": 1.251920122887865,
      "grad_norm": 5.019155025482178,
      "kl": 0.3525390625,
      "learning_rate": 6.869230769230769e-07,
      "loss": 0.0314,
      "reward": 0.7659505307674408,
      "reward_std": 0.0414460189640522,
      "reward_std_all": 0.04189744032919407,
      "rewards-max/forensics_sinkhorn_reward": 0.9649366140365601,
      "rewards-max/forensics_strict_edge_reward": 0.11701957089826465,
      "rewards-max/forensics_tiou_reward": 0.93968266248703,
      "rewards/forensics_sinkhorn_reward": 0.6944863796234131,
      "rewards/forensics_strict_edge_reward": 0.01072358645615168,
      "rewards/forensics_tiou_reward": 0.46311868727207184,
      "step": 408
    },
    {
      "completion_length": 26.265625,
      "epoch": 1.2549923195084485,
      "grad_norm": 4.468495845794678,
      "kl": 0.4130859375,
      "learning_rate": 6.861538461538461e-07,
      "loss": 0.0831,
      "reward": 0.7106419801712036,
      "reward_std": 0.08553005754947662,
      "reward_std_all": 0.08953767642378807,
      "rewards-max/forensics_sinkhorn_reward": 0.9452911913394928,
      "rewards-max/forensics_strict_edge_reward": 0.3209934961050749,
      "rewards-max/forensics_tiou_reward": 0.8912693560123444,
      "rewards/forensics_sinkhorn_reward": 0.6314242780208588,
      "rewards/forensics_strict_edge_reward": 0.039264365390408784,
      "rewards/forensics_tiou_reward": 0.4187232702970505,
      "step": 409
    },
    {
      "completion_length": 26.875,
      "epoch": 1.2580645161290323,
      "grad_norm": 4.755374908447266,
      "kl": 0.3046875,
      "learning_rate": 6.853846153846153e-07,
      "loss": 0.0887,
      "reward": 0.7340790629386902,
      "reward_std": 0.06611501984298229,
      "reward_std_all": 0.07460720837116241,
      "rewards-max/forensics_sinkhorn_reward": 0.9644893705844879,
      "rewards-max/forensics_strict_edge_reward": 0.15358472242951393,
      "rewards-max/forensics_tiou_reward": 0.9397370517253876,
      "rewards/forensics_sinkhorn_reward": 0.6359727382659912,
      "rewards/forensics_strict_edge_reward": 0.012984803761355579,
      "rewards/forensics_tiou_reward": 0.4827301651239395,
      "step": 410
    },
    {
      "completion_length": 24.46875,
      "epoch": 1.261136712749616,
      "grad_norm": 5.56267786026001,
      "kl": 0.3203125,
      "learning_rate": 6.846153846153846e-07,
      "loss": 0.0261,
      "reward": 0.6972094774246216,
      "reward_std": 0.1001950278878212,
      "reward_std_all": 0.10271400585770607,
      "rewards-max/forensics_sinkhorn_reward": 0.9394736886024475,
      "rewards-max/forensics_strict_edge_reward": 0.20135767571628094,
      "rewards-max/forensics_tiou_reward": 0.8693278729915619,
      "rewards/forensics_sinkhorn_reward": 0.6080149114131927,
      "rewards/forensics_strict_edge_reward": 0.01807851518969983,
      "rewards/forensics_tiou_reward": 0.3634907007217407,
      "step": 411
    },
    {
      "completion_length": 24.75,
      "epoch": 1.2642089093701996,
      "grad_norm": 5.06547212600708,
      "kl": 0.3427734375,
      "learning_rate": 6.838461538461539e-07,
      "loss": 0.0109,
      "reward": 0.7195131182670593,
      "reward_std": 0.0851756352931261,
      "reward_std_all": 0.08426081947982311,
      "rewards-max/forensics_sinkhorn_reward": 0.9794503152370453,
      "rewards-max/forensics_strict_edge_reward": 0.6799494326114655,
      "rewards-max/forensics_tiou_reward": 0.9760307967662811,
      "rewards/forensics_sinkhorn_reward": 0.6591773331165314,
      "rewards/forensics_strict_edge_reward": 0.04401226621121168,
      "rewards/forensics_tiou_reward": 0.45516109466552734,
      "step": 412
    },
    {
      "completion_length": 23.671875,
      "epoch": 1.2672811059907834,
      "grad_norm": 6.751143932342529,
      "kl": 0.3212890625,
      "learning_rate": 6.83076923076923e-07,
      "loss": 0.0662,
      "reward": 0.7367844581604004,
      "reward_std": 0.06490660645067692,
      "reward_std_all": 0.0657050684094429,
      "rewards-max/forensics_sinkhorn_reward": 0.9693970084190369,
      "rewards-max/forensics_strict_edge_reward": 0.4811747968196869,
      "rewards-max/forensics_tiou_reward": 0.9498176872730255,
      "rewards/forensics_sinkhorn_reward": 0.6464760601520538,
      "rewards/forensics_strict_edge_reward": 0.034121522679924965,
      "rewards/forensics_tiou_reward": 0.4741969406604767,
      "step": 413
    },
    {
      "completion_length": 23.796875,
      "epoch": 1.2703533026113671,
      "grad_norm": 4.345170021057129,
      "kl": 0.306640625,
      "learning_rate": 6.823076923076922e-07,
      "loss": 0.0018,
      "reward": 0.7940849959850311,
      "reward_std": 0.01721092965453863,
      "reward_std_all": 0.017329112626612186,
      "rewards-max/forensics_sinkhorn_reward": 0.9683248102664948,
      "rewards-max/forensics_strict_edge_reward": 0.7172991037368774,
      "rewards-max/forensics_tiou_reward": 0.9536813795566559,
      "rewards/forensics_sinkhorn_reward": 0.73919478058815,
      "rewards/forensics_strict_edge_reward": 0.05210913345217705,
      "rewards/forensics_tiou_reward": 0.5266871452331543,
      "step": 414
    },
    {
      "completion_length": 23.359375,
      "epoch": 1.273425499231951,
      "grad_norm": 5.274348258972168,
      "kl": 0.3857421875,
      "learning_rate": 6.815384615384615e-07,
      "loss": 0.1064,
      "reward": 0.762504905462265,
      "reward_std": 0.04066611733287573,
      "reward_std_all": 0.04332965146750212,
      "rewards-max/forensics_sinkhorn_reward": 0.9581749439239502,
      "rewards-max/forensics_strict_edge_reward": 0.1786501407623291,
      "rewards-max/forensics_tiou_reward": 0.9281459450721741,
      "rewards/forensics_sinkhorn_reward": 0.6546294689178467,
      "rewards/forensics_strict_edge_reward": 0.01984592853114009,
      "rewards/forensics_tiou_reward": 0.42399929463863373,
      "step": 415
    },
    {
      "completion_length": 25.8125,
      "epoch": 1.2764976958525347,
      "grad_norm": 5.343204975128174,
      "kl": 0.33984375,
      "learning_rate": 6.807692307692307e-07,
      "loss": 0.0938,
      "reward": 0.7396661043167114,
      "reward_std": 0.05958719551563263,
      "reward_std_all": 0.06633422710001469,
      "rewards-max/forensics_sinkhorn_reward": 0.9781955778598785,
      "rewards-max/forensics_strict_edge_reward": 0.6886506080627441,
      "rewards-max/forensics_tiou_reward": 0.9733940660953522,
      "rewards/forensics_sinkhorn_reward": 0.6411230266094208,
      "rewards/forensics_strict_edge_reward": 0.04456376191228628,
      "rewards/forensics_tiou_reward": 0.4425838738679886,
      "step": 416
    },
    {
      "completion_length": 22.296875,
      "epoch": 1.2795698924731183,
      "grad_norm": 5.178619861602783,
      "kl": 0.310546875,
      "learning_rate": 6.800000000000001e-07,
      "loss": 0.0409,
      "reward": 0.7460606098175049,
      "reward_std": 0.061285536736249924,
      "reward_std_all": 0.06567895784974098,
      "rewards-max/forensics_sinkhorn_reward": 0.9571309089660645,
      "rewards-max/forensics_strict_edge_reward": 0.41768769919872284,
      "rewards-max/forensics_tiou_reward": 0.9265190362930298,
      "rewards/forensics_sinkhorn_reward": 0.7135094702243805,
      "rewards/forensics_strict_edge_reward": 0.05267422553151846,
      "rewards/forensics_tiou_reward": 0.4726390987634659,
      "step": 417
    },
    {
      "completion_length": 24.21875,
      "epoch": 1.282642089093702,
      "grad_norm": 4.934727668762207,
      "kl": 0.3427734375,
      "learning_rate": 6.792307692307692e-07,
      "loss": 0.0687,
      "reward": 0.7713978886604309,
      "reward_std": 0.027306292206048965,
      "reward_std_all": 0.03274091612547636,
      "rewards-max/forensics_sinkhorn_reward": 0.9664061665534973,
      "rewards-max/forensics_strict_edge_reward": 0.6090865582227707,
      "rewards-max/forensics_tiou_reward": 0.9432389736175537,
      "rewards/forensics_sinkhorn_reward": 0.7237415611743927,
      "rewards/forensics_strict_edge_reward": 0.06698725745081902,
      "rewards/forensics_tiou_reward": 0.5089618265628815,
      "step": 418
    },
    {
      "completion_length": 22.265625,
      "epoch": 1.2857142857142856,
      "grad_norm": 5.588454246520996,
      "kl": 0.5595703125,
      "learning_rate": 6.784615384615384e-07,
      "loss": 0.0458,
      "reward": 0.7598841786384583,
      "reward_std": 0.061880555003881454,
      "reward_std_all": 0.05781625583767891,
      "rewards-max/forensics_sinkhorn_reward": 0.9851837754249573,
      "rewards-max/forensics_strict_edge_reward": 0.6637960225343704,
      "rewards-max/forensics_tiou_reward": 0.9896169900894165,
      "rewards/forensics_sinkhorn_reward": 0.6964998841285706,
      "rewards/forensics_strict_edge_reward": 0.040898711420595646,
      "rewards/forensics_tiou_reward": 0.485261008143425,
      "step": 419
    },
    {
      "completion_length": 26.8125,
      "epoch": 1.2887864823348694,
      "grad_norm": 4.038378715515137,
      "kl": 0.376953125,
      "learning_rate": 6.776923076923077e-07,
      "loss": 0.2096,
      "reward": 0.7567437887191772,
      "reward_std": 0.041648661717772484,
      "reward_std_all": 0.047642020508646965,
      "rewards-max/forensics_sinkhorn_reward": 0.9636580944061279,
      "rewards-max/forensics_strict_edge_reward": 0.14708662405610085,
      "rewards-max/forensics_tiou_reward": 0.9337850511074066,
      "rewards/forensics_sinkhorn_reward": 0.6484553515911102,
      "rewards/forensics_strict_edge_reward": 0.011313569732010365,
      "rewards/forensics_tiou_reward": 0.5042825937271118,
      "step": 420
    },
    {
      "completion_length": 24.328125,
      "epoch": 1.2918586789554531,
      "grad_norm": 5.197645664215088,
      "kl": 0.41015625,
      "learning_rate": 6.769230769230769e-07,
      "loss": 0.0245,
      "reward": 0.7779974937438965,
      "reward_std": 0.012578352354466915,
      "reward_std_all": 0.02305297739803791,
      "rewards-max/forensics_sinkhorn_reward": 0.9688867330551147,
      "rewards-max/forensics_strict_edge_reward": 0.5244058072566986,
      "rewards-max/forensics_tiou_reward": 0.9467858076095581,
      "rewards/forensics_sinkhorn_reward": 0.6960921287536621,
      "rewards/forensics_strict_edge_reward": 0.0322961937636137,
      "rewards/forensics_tiou_reward": 0.47571834921836853,
      "step": 421
    },
    {
      "completion_length": 27.953125,
      "epoch": 1.294930875576037,
      "grad_norm": 5.28088903427124,
      "kl": 0.3994140625,
      "learning_rate": 6.761538461538461e-07,
      "loss": -0.0014,
      "reward": 0.7062678635120392,
      "reward_std": 0.051832702942192554,
      "reward_std_all": 0.07229302637279034,
      "rewards-max/forensics_sinkhorn_reward": 0.8911093473434448,
      "rewards-max/forensics_strict_edge_reward": 0.00657187651086133,
      "rewards-max/forensics_tiou_reward": 0.7619866132736206,
      "rewards/forensics_sinkhorn_reward": 0.5944293737411499,
      "rewards/forensics_strict_edge_reward": 0.0004976907655418472,
      "rewards/forensics_tiou_reward": 0.32594606280326843,
      "step": 422
    },
    {
      "completion_length": 24.53125,
      "epoch": 1.2980030721966207,
      "grad_norm": 5.359442234039307,
      "kl": 0.3046875,
      "learning_rate": 6.753846153846153e-07,
      "loss": 0.0156,
      "reward": 0.7548668384552002,
      "reward_std": 0.023535209707915783,
      "reward_std_all": 0.0384617205709219,
      "rewards-max/forensics_sinkhorn_reward": 0.9612857699394226,
      "rewards-max/forensics_strict_edge_reward": 0.2783775255084038,
      "rewards-max/forensics_tiou_reward": 0.9222113192081451,
      "rewards/forensics_sinkhorn_reward": 0.678889274597168,
      "rewards/forensics_strict_edge_reward": 0.024141279980540276,
      "rewards/forensics_tiou_reward": 0.42775607109069824,
      "step": 423
    },
    {
      "completion_length": 25.359375,
      "epoch": 1.3010752688172043,
      "grad_norm": 5.728133678436279,
      "kl": 0.2734375,
      "learning_rate": 6.746153846153845e-07,
      "loss": 0.1072,
      "reward": 0.7597199976444244,
      "reward_std": 0.053406222723424435,
      "reward_std_all": 0.05163306184113026,
      "rewards-max/forensics_sinkhorn_reward": 0.9818386733531952,
      "rewards-max/forensics_strict_edge_reward": 0.6188668608665466,
      "rewards-max/forensics_tiou_reward": 0.9803265631198883,
      "rewards/forensics_sinkhorn_reward": 0.7099983394145966,
      "rewards/forensics_strict_edge_reward": 0.07357300072908401,
      "rewards/forensics_tiou_reward": 0.5426423400640488,
      "step": 424
    },
    {
      "completion_length": 25.734375,
      "epoch": 1.304147465437788,
      "grad_norm": 4.547812461853027,
      "kl": 0.361328125,
      "learning_rate": 6.738461538461539e-07,
      "loss": 0.0806,
      "reward": 0.7729334533214569,
      "reward_std": 0.043588994070887566,
      "reward_std_all": 0.04143994487822056,
      "rewards-max/forensics_sinkhorn_reward": 0.9654707610607147,
      "rewards-max/forensics_strict_edge_reward": 0.358170572668314,
      "rewards-max/forensics_tiou_reward": 0.9423960149288177,
      "rewards/forensics_sinkhorn_reward": 0.6871117949485779,
      "rewards/forensics_strict_edge_reward": 0.03551684133708477,
      "rewards/forensics_tiou_reward": 0.5567377507686615,
      "step": 425
    },
    {
      "completion_length": 23.265625,
      "epoch": 1.3072196620583718,
      "grad_norm": 5.449618816375732,
      "kl": 0.3701171875,
      "learning_rate": 6.730769230769231e-07,
      "loss": -0.0522,
      "reward": 0.7601140737533569,
      "reward_std": 0.030024817446246743,
      "reward_std_all": 0.03779458673670888,
      "rewards-max/forensics_sinkhorn_reward": 0.962809830904007,
      "rewards-max/forensics_strict_edge_reward": 0.30629871785640717,
      "rewards-max/forensics_tiou_reward": 0.9344659447669983,
      "rewards/forensics_sinkhorn_reward": 0.7048460245132446,
      "rewards/forensics_strict_edge_reward": 0.022307662293314934,
      "rewards/forensics_tiou_reward": 0.4587845504283905,
      "step": 426
    },
    {
      "completion_length": 26.78125,
      "epoch": 1.3102918586789554,
      "grad_norm": 4.9074296951293945,
      "kl": 0.3671875,
      "learning_rate": 6.723076923076922e-07,
      "loss": 0.0131,
      "reward": 0.7181048095226288,
      "reward_std": 0.035381271969527006,
      "reward_std_all": 0.06352386018261313,
      "rewards-max/forensics_sinkhorn_reward": 0.9857507646083832,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9919514358043671,
      "rewards/forensics_sinkhorn_reward": 0.6496648490428925,
      "rewards/forensics_strict_edge_reward": 0.04280165582895279,
      "rewards/forensics_tiou_reward": 0.4694490134716034,
      "step": 427
    },
    {
      "completion_length": 23.546875,
      "epoch": 1.3133640552995391,
      "grad_norm": 4.537160873413086,
      "kl": 0.3837890625,
      "learning_rate": 6.715384615384615e-07,
      "loss": 0.1145,
      "reward": 0.7564666271209717,
      "reward_std": 0.061020178720355034,
      "reward_std_all": 0.057956211268901825,
      "rewards-max/forensics_sinkhorn_reward": 0.9860623478889465,
      "rewards-max/forensics_strict_edge_reward": 0.31128623336553574,
      "rewards-max/forensics_tiou_reward": 0.9915926158428192,
      "rewards/forensics_sinkhorn_reward": 0.7192114591598511,
      "rewards/forensics_strict_edge_reward": 0.020420901477336884,
      "rewards/forensics_tiou_reward": 0.4896942973136902,
      "step": 428
    },
    {
      "completion_length": 26.78125,
      "epoch": 1.316436251920123,
      "grad_norm": 5.351689338684082,
      "kl": 0.5439453125,
      "learning_rate": 6.707692307692307e-07,
      "loss": 0.0492,
      "reward": 0.7533479332923889,
      "reward_std": 0.03519282117486,
      "reward_std_all": 0.0444186432287097,
      "rewards-max/forensics_sinkhorn_reward": 0.9693750739097595,
      "rewards-max/forensics_strict_edge_reward": 0.42263901233673096,
      "rewards-max/forensics_tiou_reward": 0.9533836841583252,
      "rewards/forensics_sinkhorn_reward": 0.6726695001125336,
      "rewards/forensics_strict_edge_reward": 0.04833747819066048,
      "rewards/forensics_tiou_reward": 0.5110324025154114,
      "step": 429
    },
    {
      "completion_length": 23.828125,
      "epoch": 1.3195084485407067,
      "grad_norm": 5.979557037353516,
      "kl": 0.318359375,
      "learning_rate": 6.7e-07,
      "loss": -0.0359,
      "reward": 0.7484568655490875,
      "reward_std": 0.039512354880571365,
      "reward_std_all": 0.0477580651640892,
      "rewards-max/forensics_sinkhorn_reward": 0.9686741828918457,
      "rewards-max/forensics_strict_edge_reward": 0.2741949260234833,
      "rewards-max/forensics_tiou_reward": 0.945987731218338,
      "rewards/forensics_sinkhorn_reward": 0.7089811861515045,
      "rewards/forensics_strict_edge_reward": 0.019164778292179108,
      "rewards/forensics_tiou_reward": 0.511741504073143,
      "step": 430
    },
    {
      "completion_length": 27.125,
      "epoch": 1.3225806451612903,
      "grad_norm": 5.882032871246338,
      "kl": 0.578125,
      "learning_rate": 6.692307692307692e-07,
      "loss": -0.0008,
      "reward": 0.7226289808750153,
      "reward_std": 0.09920601733028889,
      "reward_std_all": 0.09357111155986786,
      "rewards-max/forensics_sinkhorn_reward": 0.9531757235527039,
      "rewards-max/forensics_strict_edge_reward": 0.26160913705825806,
      "rewards-max/forensics_tiou_reward": 0.9134719371795654,
      "rewards/forensics_sinkhorn_reward": 0.6368927657604218,
      "rewards/forensics_strict_edge_reward": 0.014454232063144445,
      "rewards/forensics_tiou_reward": 0.43852056562900543,
      "step": 431
    },
    {
      "completion_length": 25.8125,
      "epoch": 1.325652841781874,
      "grad_norm": 5.267425060272217,
      "kl": 0.396484375,
      "learning_rate": 6.684615384615384e-07,
      "loss": 0.0737,
      "reward": 0.7251671552658081,
      "reward_std": 0.08306456729769707,
      "reward_std_all": 0.08167093619704247,
      "rewards-max/forensics_sinkhorn_reward": 0.9543765485286713,
      "rewards-max/forensics_strict_edge_reward": 0.1905384510755539,
      "rewards-max/forensics_tiou_reward": 0.9077609479427338,
      "rewards/forensics_sinkhorn_reward": 0.6435138881206512,
      "rewards/forensics_strict_edge_reward": 0.012316930806264281,
      "rewards/forensics_tiou_reward": 0.4745345115661621,
      "step": 432
    },
    {
      "completion_length": 26.09375,
      "epoch": 1.3287250384024578,
      "grad_norm": 6.055846691131592,
      "kl": 0.29296875,
      "learning_rate": 6.676923076923077e-07,
      "loss": 0.0633,
      "reward": 0.6977596879005432,
      "reward_std": 0.06767685897648335,
      "reward_std_all": 0.08715129643678665,
      "rewards-max/forensics_sinkhorn_reward": 0.9663968086242676,
      "rewards-max/forensics_strict_edge_reward": 0.4439111799001694,
      "rewards-max/forensics_tiou_reward": 0.9453791677951813,
      "rewards/forensics_sinkhorn_reward": 0.603346586227417,
      "rewards/forensics_strict_edge_reward": 0.030973616987466812,
      "rewards/forensics_tiou_reward": 0.41741982102394104,
      "step": 433
    },
    {
      "completion_length": 28.90625,
      "epoch": 1.3317972350230414,
      "grad_norm": 5.7707648277282715,
      "kl": 0.3330078125,
      "learning_rate": 6.669230769230769e-07,
      "loss": 0.0727,
      "reward": 0.7836198210716248,
      "reward_std": 0.026452787220478058,
      "reward_std_all": 0.026277659460902214,
      "rewards-max/forensics_sinkhorn_reward": 0.9571852087974548,
      "rewards-max/forensics_strict_edge_reward": 0.4211420267820358,
      "rewards-max/forensics_tiou_reward": 0.9024053812026978,
      "rewards/forensics_sinkhorn_reward": 0.7371154129505157,
      "rewards/forensics_strict_edge_reward": 0.032908057793974876,
      "rewards/forensics_tiou_reward": 0.5294168293476105,
      "step": 434
    },
    {
      "completion_length": 28.03125,
      "epoch": 1.3348694316436251,
      "grad_norm": 4.49787712097168,
      "kl": 0.26611328125,
      "learning_rate": 6.661538461538462e-07,
      "loss": 0.0861,
      "reward": 0.7287339568138123,
      "reward_std": 0.052314138039946556,
      "reward_std_all": 0.06612623110413551,
      "rewards-max/forensics_sinkhorn_reward": 0.9659319221973419,
      "rewards-max/forensics_strict_edge_reward": 0.28806766867637634,
      "rewards-max/forensics_tiou_reward": 0.9398342370986938,
      "rewards/forensics_sinkhorn_reward": 0.6264222860336304,
      "rewards/forensics_strict_edge_reward": 0.026289636036381125,
      "rewards/forensics_tiou_reward": 0.44011783599853516,
      "step": 435
    },
    {
      "completion_length": 23.625,
      "epoch": 1.337941628264209,
      "grad_norm": 5.372823715209961,
      "kl": 0.3701171875,
      "learning_rate": 6.653846153846153e-07,
      "loss": 0.1508,
      "reward": 0.6488839089870453,
      "reward_std": 0.11805910989642143,
      "reward_std_all": 0.13308453559875488,
      "rewards-max/forensics_sinkhorn_reward": 0.9541599154472351,
      "rewards-max/forensics_strict_edge_reward": 0.4029718041419983,
      "rewards-max/forensics_tiou_reward": 0.91758593916893,
      "rewards/forensics_sinkhorn_reward": 0.5739449858665466,
      "rewards/forensics_strict_edge_reward": 0.02708863979205489,
      "rewards/forensics_tiou_reward": 0.3500109314918518,
      "step": 436
    },
    {
      "completion_length": 25.375,
      "epoch": 1.3410138248847927,
      "grad_norm": 5.659058570861816,
      "kl": 0.3330078125,
      "learning_rate": 6.646153846153845e-07,
      "loss": -0.0093,
      "reward": 0.7569410800933838,
      "reward_std": 0.041844031773507595,
      "reward_std_all": 0.04885000362992287,
      "rewards-max/forensics_sinkhorn_reward": 0.9699166119098663,
      "rewards-max/forensics_strict_edge_reward": 0.4004678800702095,
      "rewards-max/forensics_tiou_reward": 0.9542185962200165,
      "rewards/forensics_sinkhorn_reward": 0.7016985714435577,
      "rewards/forensics_strict_edge_reward": 0.02882689516991377,
      "rewards/forensics_tiou_reward": 0.4988396465778351,
      "step": 437
    },
    {
      "completion_length": 23.828125,
      "epoch": 1.3440860215053765,
      "grad_norm": 5.828526973724365,
      "kl": 0.333984375,
      "learning_rate": 6.638461538461539e-07,
      "loss": 0.1871,
      "reward": 0.7391511797904968,
      "reward_std": 0.04891222342848778,
      "reward_std_all": 0.05897550284862518,
      "rewards-max/forensics_sinkhorn_reward": 0.95928755402565,
      "rewards-max/forensics_strict_edge_reward": 0.21820376813411713,
      "rewards-max/forensics_tiou_reward": 0.9240971207618713,
      "rewards/forensics_sinkhorn_reward": 0.6241203844547272,
      "rewards/forensics_strict_edge_reward": 0.023268242366611958,
      "rewards/forensics_tiou_reward": 0.4610573351383209,
      "step": 438
    },
    {
      "completion_length": 25.71875,
      "epoch": 1.34715821812596,
      "grad_norm": 4.699626445770264,
      "kl": 0.4287109375,
      "learning_rate": 6.630769230769231e-07,
      "loss": 0.1189,
      "reward": 0.6331384181976318,
      "reward_std": 0.0944109819829464,
      "reward_std_all": 0.13326172903180122,
      "rewards-max/forensics_sinkhorn_reward": 0.9809719026088715,
      "rewards-max/forensics_strict_edge_reward": 0.45074790716171265,
      "rewards-max/forensics_tiou_reward": 0.9778651595115662,
      "rewards/forensics_sinkhorn_reward": 0.5241567343473434,
      "rewards/forensics_strict_edge_reward": 0.039372945204377174,
      "rewards/forensics_tiou_reward": 0.3468393087387085,
      "step": 439
    },
    {
      "completion_length": 23.3125,
      "epoch": 1.3502304147465438,
      "grad_norm": 5.922904968261719,
      "kl": 0.3134765625,
      "learning_rate": 6.623076923076923e-07,
      "loss": -0.0362,
      "reward": 0.7521412372589111,
      "reward_std": 0.054512448608875275,
      "reward_std_all": 0.05478407070040703,
      "rewards-max/forensics_sinkhorn_reward": 0.9605553150177002,
      "rewards-max/forensics_strict_edge_reward": 0.5915155783295631,
      "rewards-max/forensics_tiou_reward": 0.9340323209762573,
      "rewards/forensics_sinkhorn_reward": 0.6994121372699738,
      "rewards/forensics_strict_edge_reward": 0.03324454417452216,
      "rewards/forensics_tiou_reward": 0.5168862342834473,
      "step": 440
    },
    {
      "completion_length": 22.671875,
      "epoch": 1.3533026113671274,
      "grad_norm": 4.978237152099609,
      "kl": 0.2958984375,
      "learning_rate": 6.615384615384615e-07,
      "loss": 0.1609,
      "reward": 0.7464624643325806,
      "reward_std": 0.04178869165480137,
      "reward_std_all": 0.05199125036597252,
      "rewards-max/forensics_sinkhorn_reward": 0.9646816253662109,
      "rewards-max/forensics_strict_edge_reward": 0.18768735975027084,
      "rewards-max/forensics_tiou_reward": 0.927423894405365,
      "rewards/forensics_sinkhorn_reward": 0.6507531702518463,
      "rewards/forensics_strict_edge_reward": 0.01972717186436057,
      "rewards/forensics_tiou_reward": 0.482804074883461,
      "step": 441
    },
    {
      "completion_length": 25.703125,
      "epoch": 1.3563748079877112,
      "grad_norm": 5.549525260925293,
      "kl": 0.4541015625,
      "learning_rate": 6.607692307692307e-07,
      "loss": 0.1355,
      "reward": 0.7318307161331177,
      "reward_std": 0.07037589326500893,
      "reward_std_all": 0.07241618447005749,
      "rewards-max/forensics_sinkhorn_reward": 0.9587547183036804,
      "rewards-max/forensics_strict_edge_reward": 0.16993590258061886,
      "rewards-max/forensics_tiou_reward": 0.9275647401809692,
      "rewards/forensics_sinkhorn_reward": 0.6794244050979614,
      "rewards/forensics_strict_edge_reward": 0.009899363853037357,
      "rewards/forensics_tiou_reward": 0.48947662115097046,
      "step": 442
    },
    {
      "completion_length": 22.984375,
      "epoch": 1.359447004608295,
      "grad_norm": 6.128550052642822,
      "kl": 0.330078125,
      "learning_rate": 6.6e-07,
      "loss": 0.0464,
      "reward": 0.7389201521873474,
      "reward_std": 0.07659479230642319,
      "reward_std_all": 0.07332246750593185,
      "rewards-max/forensics_sinkhorn_reward": 0.968924343585968,
      "rewards-max/forensics_strict_edge_reward": 0.4599933922290802,
      "rewards-max/forensics_tiou_reward": 0.9476617276668549,
      "rewards/forensics_sinkhorn_reward": 0.6923450231552124,
      "rewards/forensics_strict_edge_reward": 0.02765718661248684,
      "rewards/forensics_tiou_reward": 0.5172357857227325,
      "step": 443
    },
    {
      "completion_length": 24.15625,
      "epoch": 1.3625192012288787,
      "grad_norm": 4.528996467590332,
      "kl": 0.2978515625,
      "learning_rate": 6.592307692307691e-07,
      "loss": 0.1549,
      "reward": 0.765867292881012,
      "reward_std": 0.0463245939463377,
      "reward_std_all": 0.04524038918316364,
      "rewards-max/forensics_sinkhorn_reward": 0.9775657653808594,
      "rewards-max/forensics_strict_edge_reward": 0.3375049978494644,
      "rewards-max/forensics_tiou_reward": 0.9640528559684753,
      "rewards/forensics_sinkhorn_reward": 0.6869164407253265,
      "rewards/forensics_strict_edge_reward": 0.02280052937567234,
      "rewards/forensics_tiou_reward": 0.4808439314365387,
      "step": 444
    },
    {
      "completion_length": 24.390625,
      "epoch": 1.3655913978494625,
      "grad_norm": 6.072962284088135,
      "kl": 0.3857421875,
      "learning_rate": 6.584615384615384e-07,
      "loss": 0.0843,
      "reward": 0.7891407310962677,
      "reward_std": 0.025773673318326473,
      "reward_std_all": 0.024895218200981617,
      "rewards-max/forensics_sinkhorn_reward": 0.95866858959198,
      "rewards-max/forensics_strict_edge_reward": 0.2783775255084038,
      "rewards-max/forensics_tiou_reward": 0.9190888404846191,
      "rewards/forensics_sinkhorn_reward": 0.6930271983146667,
      "rewards/forensics_strict_edge_reward": 0.030854259617626667,
      "rewards/forensics_tiou_reward": 0.5526127815246582,
      "step": 445
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.368663594470046,
      "grad_norm": 4.332334518432617,
      "kl": 0.30322265625,
      "learning_rate": 6.576923076923077e-07,
      "loss": 0.02,
      "reward": 0.7175230383872986,
      "reward_std": 0.028894992545247078,
      "reward_std_all": 0.06227435544133186,
      "rewards-max/forensics_sinkhorn_reward": 0.9590060412883759,
      "rewards-max/forensics_strict_edge_reward": 0.16209223633632064,
      "rewards-max/forensics_tiou_reward": 0.9273146986961365,
      "rewards/forensics_sinkhorn_reward": 0.6337037980556488,
      "rewards/forensics_strict_edge_reward": 0.013556835590861738,
      "rewards/forensics_tiou_reward": 0.38630980253219604,
      "step": 446
    },
    {
      "completion_length": 22.5,
      "epoch": 1.3717357910906298,
      "grad_norm": 6.79365348815918,
      "kl": 0.32958984375,
      "learning_rate": 6.569230769230769e-07,
      "loss": 0.0001,
      "reward": 0.7790404856204987,
      "reward_std": 0.03607247490435839,
      "reward_std_all": 0.03493919596076012,
      "rewards-max/forensics_sinkhorn_reward": 0.9634105265140533,
      "rewards-max/forensics_strict_edge_reward": 0.5034982860088348,
      "rewards-max/forensics_tiou_reward": 0.9397345185279846,
      "rewards/forensics_sinkhorn_reward": 0.7089015245437622,
      "rewards/forensics_strict_edge_reward": 0.05580070801079273,
      "rewards/forensics_tiou_reward": 0.5239086151123047,
      "step": 447
    },
    {
      "completion_length": 24.03125,
      "epoch": 1.3748079877112136,
      "grad_norm": 5.223072052001953,
      "kl": 0.390625,
      "learning_rate": 6.561538461538462e-07,
      "loss": 0.0505,
      "reward": 0.7766388654708862,
      "reward_std": 0.04642333276569843,
      "reward_std_all": 0.042505893856287,
      "rewards-max/forensics_sinkhorn_reward": 0.9704755544662476,
      "rewards-max/forensics_strict_edge_reward": 0.5016714781522751,
      "rewards-max/forensics_tiou_reward": 0.9570778608322144,
      "rewards/forensics_sinkhorn_reward": 0.7621843218803406,
      "rewards/forensics_strict_edge_reward": 0.05360681749880314,
      "rewards/forensics_tiou_reward": 0.5613920986652374,
      "step": 448
    },
    {
      "completion_length": 22.546875,
      "epoch": 1.3778801843317972,
      "grad_norm": 5.66400671005249,
      "kl": 0.353515625,
      "learning_rate": 6.553846153846153e-07,
      "loss": 0.0439,
      "reward": 0.7187929153442383,
      "reward_std": 0.03975044749677181,
      "reward_std_all": 0.062056176364421844,
      "rewards-max/forensics_sinkhorn_reward": 0.9267403483390808,
      "rewards-max/forensics_strict_edge_reward": 0.29640552774071693,
      "rewards-max/forensics_tiou_reward": 0.8591626286506653,
      "rewards/forensics_sinkhorn_reward": 0.6328714787960052,
      "rewards/forensics_strict_edge_reward": 0.019901643274351954,
      "rewards/forensics_tiou_reward": 0.38288700580596924,
      "step": 449
    },
    {
      "completion_length": 27.4375,
      "epoch": 1.380952380952381,
      "grad_norm": 5.522749900817871,
      "kl": 0.396484375,
      "learning_rate": 6.546153846153845e-07,
      "loss": 0.2289,
      "reward": 0.6871872544288635,
      "reward_std": 0.09369947761297226,
      "reward_std_all": 0.1021750383079052,
      "rewards-max/forensics_sinkhorn_reward": 0.9302526414394379,
      "rewards-max/forensics_strict_edge_reward": 0.18138347193598747,
      "rewards-max/forensics_tiou_reward": 0.854897528886795,
      "rewards/forensics_sinkhorn_reward": 0.553729772567749,
      "rewards/forensics_strict_edge_reward": 0.0175608245190233,
      "rewards/forensics_tiou_reward": 0.41413864493370056,
      "step": 450
    },
    {
      "completion_length": 24.4375,
      "epoch": 1.3840245775729647,
      "grad_norm": 4.693016529083252,
      "kl": 0.310546875,
      "learning_rate": 6.538461538461538e-07,
      "loss": 0.0455,
      "reward": 0.7332986891269684,
      "reward_std": 0.03634391166269779,
      "reward_std_all": 0.05504657607525587,
      "rewards-max/forensics_sinkhorn_reward": 0.9834811389446259,
      "rewards-max/forensics_strict_edge_reward": 0.37667712569236755,
      "rewards-max/forensics_tiou_reward": 0.9863311648368835,
      "rewards/forensics_sinkhorn_reward": 0.630521297454834,
      "rewards/forensics_strict_edge_reward": 0.02406847197562456,
      "rewards/forensics_tiou_reward": 0.4441870003938675,
      "step": 451
    },
    {
      "completion_length": 23.09375,
      "epoch": 1.3870967741935485,
      "grad_norm": 4.674777030944824,
      "kl": 0.4150390625,
      "learning_rate": 6.530769230769231e-07,
      "loss": 0.1324,
      "reward": 0.7002504765987396,
      "reward_std": 0.06842145323753357,
      "reward_std_all": 0.08414102718234062,
      "rewards-max/forensics_sinkhorn_reward": 0.9385740160942078,
      "rewards-max/forensics_strict_edge_reward": 0.12166334595531225,
      "rewards-max/forensics_tiou_reward": 0.8681321740150452,
      "rewards/forensics_sinkhorn_reward": 0.616990715265274,
      "rewards/forensics_strict_edge_reward": 0.010698401252739131,
      "rewards/forensics_tiou_reward": 0.3701307475566864,
      "step": 452
    },
    {
      "completion_length": 22.78125,
      "epoch": 1.390168970814132,
      "grad_norm": 5.010928630828857,
      "kl": 0.33203125,
      "learning_rate": 6.523076923076923e-07,
      "loss": 0.0631,
      "reward": 0.7597419321537018,
      "reward_std": 0.037166199181228876,
      "reward_std_all": 0.04178598802536726,
      "rewards-max/forensics_sinkhorn_reward": 0.959449827671051,
      "rewards-max/forensics_strict_edge_reward": 0.32904115319252014,
      "rewards-max/forensics_tiou_reward": 0.9208783209323883,
      "rewards/forensics_sinkhorn_reward": 0.6769888997077942,
      "rewards/forensics_strict_edge_reward": 0.02313975617289543,
      "rewards/forensics_tiou_reward": 0.4377858340740204,
      "step": 453
    },
    {
      "completion_length": 23.453125,
      "epoch": 1.3932411674347158,
      "grad_norm": 6.017242908477783,
      "kl": 0.25634765625,
      "learning_rate": 6.515384615384615e-07,
      "loss": 0.0978,
      "reward": 0.726673036813736,
      "reward_std": 0.06714279390871525,
      "reward_std_all": 0.07275083474814892,
      "rewards-max/forensics_sinkhorn_reward": 0.9829369485378265,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9816741645336151,
      "rewards/forensics_sinkhorn_reward": 0.6378666758537292,
      "rewards/forensics_strict_edge_reward": 0.04055118188261986,
      "rewards/forensics_tiou_reward": 0.3984917849302292,
      "step": 454
    },
    {
      "completion_length": 25.96875,
      "epoch": 1.3963133640552996,
      "grad_norm": 5.888276100158691,
      "kl": 0.322265625,
      "learning_rate": 6.507692307692307e-07,
      "loss": 0.043,
      "reward": 0.6983966827392578,
      "reward_std": 0.07743780687451363,
      "reward_std_all": 0.08934340253472328,
      "rewards-max/forensics_sinkhorn_reward": 0.9294695854187012,
      "rewards-max/forensics_strict_edge_reward": 0.16722321510314941,
      "rewards-max/forensics_tiou_reward": 0.8597559928894043,
      "rewards/forensics_sinkhorn_reward": 0.5772124826908112,
      "rewards/forensics_strict_edge_reward": 0.012084903195500374,
      "rewards/forensics_tiou_reward": 0.33277058601379395,
      "step": 455
    },
    {
      "completion_length": 22.09375,
      "epoch": 1.3993855606758832,
      "grad_norm": 5.335614204406738,
      "kl": 0.419921875,
      "learning_rate": 6.5e-07,
      "loss": 0.0595,
      "reward": 0.7607870101928711,
      "reward_std": 0.04566926043480635,
      "reward_std_all": 0.045639023184776306,
      "rewards-max/forensics_sinkhorn_reward": 0.9806822836399078,
      "rewards-max/forensics_strict_edge_reward": 0.6912624835968018,
      "rewards-max/forensics_tiou_reward": 0.9812442064285278,
      "rewards/forensics_sinkhorn_reward": 0.7069494724273682,
      "rewards/forensics_strict_edge_reward": 0.05546308867633343,
      "rewards/forensics_tiou_reward": 0.48115260899066925,
      "step": 456
    },
    {
      "completion_length": 24.375,
      "epoch": 1.402457757296467,
      "grad_norm": 7.182456016540527,
      "kl": 0.419921875,
      "learning_rate": 6.492307692307692e-07,
      "loss": 0.007,
      "reward": 0.7024045884609222,
      "reward_std": 0.04788309335708618,
      "reward_std_all": 0.07003890722990036,
      "rewards-max/forensics_sinkhorn_reward": 0.9657320976257324,
      "rewards-max/forensics_strict_edge_reward": 0.3828766196966171,
      "rewards-max/forensics_tiou_reward": 0.9402818083763123,
      "rewards/forensics_sinkhorn_reward": 0.6575144231319427,
      "rewards/forensics_strict_edge_reward": 0.03148449305444956,
      "rewards/forensics_tiou_reward": 0.4575274884700775,
      "step": 457
    },
    {
      "completion_length": 23.515625,
      "epoch": 1.4055299539170507,
      "grad_norm": 4.560762405395508,
      "kl": 0.38671875,
      "learning_rate": 6.484615384615383e-07,
      "loss": -0.0321,
      "reward": 0.7336248755455017,
      "reward_std": 0.05313788913190365,
      "reward_std_all": 0.06052808091044426,
      "rewards-max/forensics_sinkhorn_reward": 0.9667028188705444,
      "rewards-max/forensics_strict_edge_reward": 0.4012388288974762,
      "rewards-max/forensics_tiou_reward": 0.9455710053443909,
      "rewards/forensics_sinkhorn_reward": 0.6503848433494568,
      "rewards/forensics_strict_edge_reward": 0.02431434765458107,
      "rewards/forensics_tiou_reward": 0.4403650760650635,
      "step": 458
    },
    {
      "completion_length": 25.84375,
      "epoch": 1.4086021505376345,
      "grad_norm": 5.38216495513916,
      "kl": 0.3154296875,
      "learning_rate": 6.476923076923077e-07,
      "loss": 0.0942,
      "reward": 0.7311930656433105,
      "reward_std": 0.10096985101699829,
      "reward_std_all": 0.09265010803937912,
      "rewards-max/forensics_sinkhorn_reward": 0.9300140738487244,
      "rewards-max/forensics_strict_edge_reward": 0.1679408811032772,
      "rewards-max/forensics_tiou_reward": 0.8674562573432922,
      "rewards/forensics_sinkhorn_reward": 0.6803743541240692,
      "rewards/forensics_strict_edge_reward": 0.01925945235416293,
      "rewards/forensics_tiou_reward": 0.4475371837615967,
      "step": 459
    },
    {
      "completion_length": 25.828125,
      "epoch": 1.411674347158218,
      "grad_norm": 5.640000343322754,
      "kl": 0.4130859375,
      "learning_rate": 6.469230769230769e-07,
      "loss": 0.0764,
      "reward": 0.7732660174369812,
      "reward_std": 0.04447941086255014,
      "reward_std_all": 0.04216527449898422,
      "rewards-max/forensics_sinkhorn_reward": 0.9527307450771332,
      "rewards-max/forensics_strict_edge_reward": 0.2289443500339985,
      "rewards-max/forensics_tiou_reward": 0.9099697470664978,
      "rewards/forensics_sinkhorn_reward": 0.6910483241081238,
      "rewards/forensics_strict_edge_reward": 0.014517067931592464,
      "rewards/forensics_tiou_reward": 0.5177605450153351,
      "step": 460
    },
    {
      "completion_length": 21.8125,
      "epoch": 1.4147465437788018,
      "grad_norm": 6.174407482147217,
      "kl": 0.310546875,
      "learning_rate": 6.461538461538462e-07,
      "loss": 0.094,
      "reward": 0.7477553188800812,
      "reward_std": 0.08326489478349686,
      "reward_std_all": 0.07591811195015907,
      "rewards-max/forensics_sinkhorn_reward": 0.9878095984458923,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9963068664073944,
      "rewards/forensics_sinkhorn_reward": 0.7219759821891785,
      "rewards/forensics_strict_edge_reward": 0.08281925693154335,
      "rewards/forensics_tiou_reward": 0.5429399311542511,
      "step": 461
    },
    {
      "completion_length": 25.65625,
      "epoch": 1.4178187403993856,
      "grad_norm": 5.956150531768799,
      "kl": 0.2255859375,
      "learning_rate": 6.453846153846153e-07,
      "loss": -0.0462,
      "reward": 0.7375693917274475,
      "reward_std": 0.03431160282343626,
      "reward_std_all": 0.0506239328533411,
      "rewards-max/forensics_sinkhorn_reward": 0.934626430273056,
      "rewards-max/forensics_strict_edge_reward": 0.38278166949748993,
      "rewards-max/forensics_tiou_reward": 0.88123819231987,
      "rewards/forensics_sinkhorn_reward": 0.6689645051956177,
      "rewards/forensics_strict_edge_reward": 0.03706908971071243,
      "rewards/forensics_tiou_reward": 0.41223061084747314,
      "step": 462
    },
    {
      "completion_length": 25.671875,
      "epoch": 1.4208909370199692,
      "grad_norm": 5.4345784187316895,
      "kl": 0.3837890625,
      "learning_rate": 6.446153846153845e-07,
      "loss": 0.0903,
      "reward": 0.764348030090332,
      "reward_std": 0.04587617143988609,
      "reward_std_all": 0.04530830681324005,
      "rewards-max/forensics_sinkhorn_reward": 0.976733386516571,
      "rewards-max/forensics_strict_edge_reward": 0.24969295412302017,
      "rewards-max/forensics_tiou_reward": 0.9659955203533173,
      "rewards/forensics_sinkhorn_reward": 0.6624616086483002,
      "rewards/forensics_strict_edge_reward": 0.023584842681884766,
      "rewards/forensics_tiou_reward": 0.46741485595703125,
      "step": 463
    },
    {
      "completion_length": 27.6875,
      "epoch": 1.423963133640553,
      "grad_norm": 5.146247863769531,
      "kl": 0.296875,
      "learning_rate": 6.438461538461538e-07,
      "loss": 0.0424,
      "reward": 0.7333255112171173,
      "reward_std": 0.07630973495543003,
      "reward_std_all": 0.07449195347726345,
      "rewards-max/forensics_sinkhorn_reward": 0.9341091513633728,
      "rewards-max/forensics_strict_edge_reward": 0.19487615209072828,
      "rewards-max/forensics_tiou_reward": 0.8631338775157928,
      "rewards/forensics_sinkhorn_reward": 0.6104160845279694,
      "rewards/forensics_strict_edge_reward": 0.013017836259678006,
      "rewards/forensics_tiou_reward": 0.4153808653354645,
      "step": 464
    },
    {
      "completion_length": 24.734375,
      "epoch": 1.4270353302611367,
      "grad_norm": 5.760993003845215,
      "kl": 0.40625,
      "learning_rate": 6.43076923076923e-07,
      "loss": 0.0826,
      "reward": 0.724178820848465,
      "reward_std": 0.09374177362769842,
      "reward_std_all": 0.08935843780636787,
      "rewards-max/forensics_sinkhorn_reward": 0.9621523916721344,
      "rewards-max/forensics_strict_edge_reward": 0.4374866336584091,
      "rewards-max/forensics_tiou_reward": 0.9354381263256073,
      "rewards/forensics_sinkhorn_reward": 0.6340017020702362,
      "rewards/forensics_strict_edge_reward": 0.02662817481905222,
      "rewards/forensics_tiou_reward": 0.43288780748844147,
      "step": 465
    },
    {
      "completion_length": 28.109375,
      "epoch": 1.4301075268817205,
      "grad_norm": 6.887668609619141,
      "kl": 0.681640625,
      "learning_rate": 6.423076923076924e-07,
      "loss": 0.1322,
      "reward": 0.76303631067276,
      "reward_std": 0.04845791682600975,
      "reward_std_all": 0.0470899511128664,
      "rewards-max/forensics_sinkhorn_reward": 0.9645674526691437,
      "rewards-max/forensics_strict_edge_reward": 0.4737703502178192,
      "rewards-max/forensics_tiou_reward": 0.9404818713665009,
      "rewards/forensics_sinkhorn_reward": 0.7100059390068054,
      "rewards/forensics_strict_edge_reward": 0.0519431009888649,
      "rewards/forensics_tiou_reward": 0.5316962152719498,
      "step": 466
    },
    {
      "completion_length": 24.375,
      "epoch": 1.4331797235023043,
      "grad_norm": 4.920709609985352,
      "kl": 0.39453125,
      "learning_rate": 6.415384615384615e-07,
      "loss": 0.0626,
      "reward": 0.7741928100585938,
      "reward_std": 0.054323095828294754,
      "reward_std_all": 0.04895623028278351,
      "rewards-max/forensics_sinkhorn_reward": 0.9424555897712708,
      "rewards-max/forensics_strict_edge_reward": 0.2740888148546219,
      "rewards-max/forensics_tiou_reward": 0.8895005881786346,
      "rewards/forensics_sinkhorn_reward": 0.7365219295024872,
      "rewards/forensics_strict_edge_reward": 0.029462678357958794,
      "rewards/forensics_tiou_reward": 0.5644653737545013,
      "step": 467
    },
    {
      "completion_length": 24.796875,
      "epoch": 1.4362519201228878,
      "grad_norm": 5.020821571350098,
      "kl": 0.2822265625,
      "learning_rate": 6.407692307692307e-07,
      "loss": -0.0638,
      "reward": 0.8009403049945831,
      "reward_std": 0.007091271341778338,
      "reward_std_all": 0.008651531767100096,
      "rewards-max/forensics_sinkhorn_reward": 0.9686740338802338,
      "rewards-max/forensics_strict_edge_reward": 0.418794646859169,
      "rewards-max/forensics_tiou_reward": 0.9514094293117523,
      "rewards/forensics_sinkhorn_reward": 0.7541273534297943,
      "rewards/forensics_strict_edge_reward": 0.03638934995979071,
      "rewards/forensics_tiou_reward": 0.5560484230518341,
      "step": 468
    },
    {
      "completion_length": 25.421875,
      "epoch": 1.4393241167434716,
      "grad_norm": 6.203269004821777,
      "kl": 0.533203125,
      "learning_rate": 6.4e-07,
      "loss": 0.1944,
      "reward": 0.7222371697425842,
      "reward_std": 0.09109843894839287,
      "reward_std_all": 0.08804887905716896,
      "rewards-max/forensics_sinkhorn_reward": 0.9554590582847595,
      "rewards-max/forensics_strict_edge_reward": 0.267449701204896,
      "rewards-max/forensics_tiou_reward": 0.9134840667247772,
      "rewards/forensics_sinkhorn_reward": 0.6097921133041382,
      "rewards/forensics_strict_edge_reward": 0.01821011188440025,
      "rewards/forensics_tiou_reward": 0.4495835602283478,
      "step": 469
    },
    {
      "completion_length": 26.375,
      "epoch": 1.4423963133640554,
      "grad_norm": 4.841129302978516,
      "kl": 0.4248046875,
      "learning_rate": 6.392307692307692e-07,
      "loss": 0.095,
      "reward": 0.7171462178230286,
      "reward_std": 0.06058058328926563,
      "reward_std_all": 0.07313131913542747,
      "rewards-max/forensics_sinkhorn_reward": 0.9465544521808624,
      "rewards-max/forensics_strict_edge_reward": 0.21773628145456314,
      "rewards-max/forensics_tiou_reward": 0.8940295577049255,
      "rewards/forensics_sinkhorn_reward": 0.6157144010066986,
      "rewards/forensics_strict_edge_reward": 0.010852386243641376,
      "rewards/forensics_tiou_reward": 0.40505242347717285,
      "step": 470
    },
    {
      "completion_length": 26.3125,
      "epoch": 1.445468509984639,
      "grad_norm": 5.226351737976074,
      "kl": 0.2880859375,
      "learning_rate": 6.384615384615383e-07,
      "loss": 0.0889,
      "reward": 0.7330471575260162,
      "reward_std": 0.06620271503925323,
      "reward_std_all": 0.06829585507512093,
      "rewards-max/forensics_sinkhorn_reward": 0.9409716129302979,
      "rewards-max/forensics_strict_edge_reward": 0.2649586796760559,
      "rewards-max/forensics_tiou_reward": 0.889224499464035,
      "rewards/forensics_sinkhorn_reward": 0.6189144849777222,
      "rewards/forensics_strict_edge_reward": 0.022657049354165792,
      "rewards/forensics_tiou_reward": 0.41977033019065857,
      "step": 471
    },
    {
      "completion_length": 25.296875,
      "epoch": 1.4485407066052227,
      "grad_norm": 6.145089149475098,
      "kl": 0.32421875,
      "learning_rate": 6.376923076923077e-07,
      "loss": 0.0234,
      "reward": 0.7617684304714203,
      "reward_std": 0.02753036515787244,
      "reward_std_all": 0.03659810312092304,
      "rewards-max/forensics_sinkhorn_reward": 0.9548565149307251,
      "rewards-max/forensics_strict_edge_reward": 0.25112341716885567,
      "rewards-max/forensics_tiou_reward": 0.9102204740047455,
      "rewards/forensics_sinkhorn_reward": 0.709251195192337,
      "rewards/forensics_strict_edge_reward": 0.020678729517385364,
      "rewards/forensics_tiou_reward": 0.48906660079956055,
      "step": 472
    },
    {
      "completion_length": 23.328125,
      "epoch": 1.4516129032258065,
      "grad_norm": 4.973062992095947,
      "kl": 0.39453125,
      "learning_rate": 6.369230769230769e-07,
      "loss": 0.1242,
      "reward": 0.7365746200084686,
      "reward_std": 0.08127206936478615,
      "reward_std_all": 0.07745597511529922,
      "rewards-max/forensics_sinkhorn_reward": 0.9632484912872314,
      "rewards-max/forensics_strict_edge_reward": 0.35583512485027313,
      "rewards-max/forensics_tiou_reward": 0.9355964064598083,
      "rewards/forensics_sinkhorn_reward": 0.6980227530002594,
      "rewards/forensics_strict_edge_reward": 0.0367148332297802,
      "rewards/forensics_tiou_reward": 0.4894813150167465,
      "step": 473
    },
    {
      "completion_length": 27.875,
      "epoch": 1.4546850998463903,
      "grad_norm": 5.262224197387695,
      "kl": 0.3837890625,
      "learning_rate": 6.361538461538462e-07,
      "loss": 0.1036,
      "reward": 0.7296586632728577,
      "reward_std": 0.06451407168060541,
      "reward_std_all": 0.07343127578496933,
      "rewards-max/forensics_sinkhorn_reward": 0.9621127545833588,
      "rewards-max/forensics_strict_edge_reward": 0.3608009144663811,
      "rewards-max/forensics_tiou_reward": 0.93337282538414,
      "rewards/forensics_sinkhorn_reward": 0.5973003804683685,
      "rewards/forensics_strict_edge_reward": 0.030656080227345228,
      "rewards/forensics_tiou_reward": 0.4802952706813812,
      "step": 474
    },
    {
      "completion_length": 27.171875,
      "epoch": 1.4577572964669738,
      "grad_norm": 4.463325023651123,
      "kl": 0.318359375,
      "learning_rate": 6.353846153846154e-07,
      "loss": 0.0768,
      "reward": 0.7824205458164215,
      "reward_std": 0.026888360269367695,
      "reward_std_all": 0.02752157673239708,
      "rewards-max/forensics_sinkhorn_reward": 0.9190816581249237,
      "rewards-max/forensics_strict_edge_reward": 0.22396432980895042,
      "rewards-max/forensics_tiou_reward": 0.8444273471832275,
      "rewards/forensics_sinkhorn_reward": 0.7010295689105988,
      "rewards/forensics_strict_edge_reward": 0.024970742408186197,
      "rewards/forensics_tiou_reward": 0.49957306683063507,
      "step": 475
    },
    {
      "completion_length": 26.140625,
      "epoch": 1.4608294930875576,
      "grad_norm": 5.421820163726807,
      "kl": 0.31103515625,
      "learning_rate": 6.346153846153845e-07,
      "loss": 0.0643,
      "reward": 0.7679864764213562,
      "reward_std": 0.03954896703362465,
      "reward_std_all": 0.039718858897686005,
      "rewards-max/forensics_sinkhorn_reward": 0.9731675982475281,
      "rewards-max/forensics_strict_edge_reward": 0.48720505833625793,
      "rewards-max/forensics_tiou_reward": 0.9570183157920837,
      "rewards/forensics_sinkhorn_reward": 0.7255136668682098,
      "rewards/forensics_strict_edge_reward": 0.04150855354964733,
      "rewards/forensics_tiou_reward": 0.5572136044502258,
      "step": 476
    },
    {
      "completion_length": 25.90625,
      "epoch": 1.4639016897081414,
      "grad_norm": 5.942483425140381,
      "kl": 0.376953125,
      "learning_rate": 6.338461538461538e-07,
      "loss": 0.1592,
      "reward": 0.7381334006786346,
      "reward_std": 0.06999626010656357,
      "reward_std_all": 0.06841491349041462,
      "rewards-max/forensics_sinkhorn_reward": 0.9784746170043945,
      "rewards-max/forensics_strict_edge_reward": 0.4240923821926117,
      "rewards-max/forensics_tiou_reward": 0.969471663236618,
      "rewards/forensics_sinkhorn_reward": 0.6530072093009949,
      "rewards/forensics_strict_edge_reward": 0.026672497391700745,
      "rewards/forensics_tiou_reward": 0.4638911932706833,
      "step": 477
    },
    {
      "completion_length": 25.796875,
      "epoch": 1.466973886328725,
      "grad_norm": 5.857192516326904,
      "kl": 0.2822265625,
      "learning_rate": 6.33076923076923e-07,
      "loss": 0.2088,
      "reward": 0.7314821481704712,
      "reward_std": 0.07509421929717064,
      "reward_std_all": 0.07702062837779522,
      "rewards-max/forensics_sinkhorn_reward": 0.9703757166862488,
      "rewards-max/forensics_strict_edge_reward": 0.29383689165115356,
      "rewards-max/forensics_tiou_reward": 0.9520095884799957,
      "rewards/forensics_sinkhorn_reward": 0.639812171459198,
      "rewards/forensics_strict_edge_reward": 0.01409427635371685,
      "rewards/forensics_tiou_reward": 0.4593263119459152,
      "step": 478
    },
    {
      "completion_length": 26.078125,
      "epoch": 1.4700460829493087,
      "grad_norm": 5.307417869567871,
      "kl": 0.3486328125,
      "learning_rate": 6.323076923076924e-07,
      "loss": 0.154,
      "reward": 0.7738787829875946,
      "reward_std": 0.033215804025530815,
      "reward_std_all": 0.03387333545833826,
      "rewards-max/forensics_sinkhorn_reward": 0.9707845449447632,
      "rewards-max/forensics_strict_edge_reward": 0.47388121485710144,
      "rewards-max/forensics_tiou_reward": 0.9530546069145203,
      "rewards/forensics_sinkhorn_reward": 0.7036081850528717,
      "rewards/forensics_strict_edge_reward": 0.035526021383702755,
      "rewards/forensics_tiou_reward": 0.4888201951980591,
      "step": 479
    },
    {
      "completion_length": 25.09375,
      "epoch": 1.4731182795698925,
      "grad_norm": 4.470398426055908,
      "kl": 0.27978515625,
      "learning_rate": 6.315384615384615e-07,
      "loss": -0.0321,
      "reward": 0.7908511757850647,
      "reward_std": 0.020818700082600117,
      "reward_std_all": 0.021023834124207497,
      "rewards-max/forensics_sinkhorn_reward": 0.9621452391147614,
      "rewards-max/forensics_strict_edge_reward": 0.22456561401486397,
      "rewards-max/forensics_tiou_reward": 0.9343957304954529,
      "rewards/forensics_sinkhorn_reward": 0.7406664788722992,
      "rewards/forensics_strict_edge_reward": 0.016196837415918708,
      "rewards/forensics_tiou_reward": 0.49448980391025543,
      "step": 480
    },
    {
      "completion_length": 23.375,
      "epoch": 1.4761904761904763,
      "grad_norm": 4.271923065185547,
      "kl": 0.3291015625,
      "learning_rate": 6.307692307692307e-07,
      "loss": 0.0998,
      "reward": 0.7611181139945984,
      "reward_std": 0.05633043684065342,
      "reward_std_all": 0.0532715767621994,
      "rewards-max/forensics_sinkhorn_reward": 0.9885495007038116,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9988354444503784,
      "rewards/forensics_sinkhorn_reward": 0.6981660425662994,
      "rewards/forensics_strict_edge_reward": 0.08914429321885109,
      "rewards/forensics_tiou_reward": 0.5534124374389648,
      "step": 481
    },
    {
      "completion_length": 26.125,
      "epoch": 1.4792626728110598,
      "grad_norm": 4.985058307647705,
      "kl": 0.3662109375,
      "learning_rate": 6.3e-07,
      "loss": 0.1216,
      "reward": 0.7592186331748962,
      "reward_std": 0.03767818957567215,
      "reward_std_all": 0.04584990441799164,
      "rewards-max/forensics_sinkhorn_reward": 0.9672231078147888,
      "rewards-max/forensics_strict_edge_reward": 0.5995520651340485,
      "rewards-max/forensics_tiou_reward": 0.9475995600223541,
      "rewards/forensics_sinkhorn_reward": 0.6987148821353912,
      "rewards/forensics_strict_edge_reward": 0.04090213868767023,
      "rewards/forensics_tiou_reward": 0.4882786273956299,
      "step": 482
    },
    {
      "completion_length": 23.234375,
      "epoch": 1.4823348694316436,
      "grad_norm": 5.491433620452881,
      "kl": 0.357421875,
      "learning_rate": 6.292307692307692e-07,
      "loss": -0.0283,
      "reward": 0.7559386491775513,
      "reward_std": 0.052951088175177574,
      "reward_std_all": 0.052231963723897934,
      "rewards-max/forensics_sinkhorn_reward": 0.9738048911094666,
      "rewards-max/forensics_strict_edge_reward": 0.5081118941307068,
      "rewards-max/forensics_tiou_reward": 0.9566754996776581,
      "rewards/forensics_sinkhorn_reward": 0.7207598090171814,
      "rewards/forensics_strict_edge_reward": 0.03066817671060562,
      "rewards/forensics_tiou_reward": 0.4910038709640503,
      "step": 483
    },
    {
      "completion_length": 25.0,
      "epoch": 1.4854070660522274,
      "grad_norm": 5.440832614898682,
      "kl": 0.3623046875,
      "learning_rate": 6.284615384615385e-07,
      "loss": 0.1106,
      "reward": 0.6941671371459961,
      "reward_std": 0.04399886354804039,
      "reward_std_all": 0.07573081739246845,
      "rewards-max/forensics_sinkhorn_reward": 0.9623701572418213,
      "rewards-max/forensics_strict_edge_reward": 0.26513541489839554,
      "rewards-max/forensics_tiou_reward": 0.9138050973415375,
      "rewards/forensics_sinkhorn_reward": 0.5970367789268494,
      "rewards/forensics_strict_edge_reward": 0.019615347031503916,
      "rewards/forensics_tiou_reward": 0.4181775152683258,
      "step": 484
    },
    {
      "completion_length": 24.375,
      "epoch": 1.488479262672811,
      "grad_norm": 4.78824520111084,
      "kl": 0.310546875,
      "learning_rate": 6.276923076923076e-07,
      "loss": 0.061,
      "reward": 0.765929639339447,
      "reward_std": 0.04787936061620712,
      "reward_std_all": 0.0461056362837553,
      "rewards-max/forensics_sinkhorn_reward": 0.9580455124378204,
      "rewards-max/forensics_strict_edge_reward": 0.22297906875610352,
      "rewards-max/forensics_tiou_reward": 0.9250324368476868,
      "rewards/forensics_sinkhorn_reward": 0.6599295139312744,
      "rewards/forensics_strict_edge_reward": 0.02647464256733656,
      "rewards/forensics_tiou_reward": 0.5133508741855621,
      "step": 485
    },
    {
      "completion_length": 24.671875,
      "epoch": 1.4915514592933947,
      "grad_norm": 4.7570319175720215,
      "kl": 0.36328125,
      "learning_rate": 6.269230769230769e-07,
      "loss": 0.0144,
      "reward": 0.6856322288513184,
      "reward_std": 0.08548449724912643,
      "reward_std_all": 0.10145896673202515,
      "rewards-max/forensics_sinkhorn_reward": 0.9657348394393921,
      "rewards-max/forensics_strict_edge_reward": 0.4016733840107918,
      "rewards-max/forensics_tiou_reward": 0.9251219034194946,
      "rewards/forensics_sinkhorn_reward": 0.5909885168075562,
      "rewards/forensics_strict_edge_reward": 0.018948952667415142,
      "rewards/forensics_tiou_reward": 0.36055949330329895,
      "step": 486
    },
    {
      "completion_length": 23.328125,
      "epoch": 1.4946236559139785,
      "grad_norm": 5.129023551940918,
      "kl": 0.3876953125,
      "learning_rate": 6.261538461538462e-07,
      "loss": -0.0039,
      "reward": 0.7026250958442688,
      "reward_std": 0.04937359876930714,
      "reward_std_all": 0.0736211147159338,
      "rewards-max/forensics_sinkhorn_reward": 0.9729009866714478,
      "rewards-max/forensics_strict_edge_reward": 0.4900742173194885,
      "rewards-max/forensics_tiou_reward": 0.9560143947601318,
      "rewards/forensics_sinkhorn_reward": 0.6539756655693054,
      "rewards/forensics_strict_edge_reward": 0.027506080456078053,
      "rewards/forensics_tiou_reward": 0.4325917214155197,
      "step": 487
    },
    {
      "completion_length": 25.921875,
      "epoch": 1.4976958525345623,
      "grad_norm": 5.172447681427002,
      "kl": 0.3642578125,
      "learning_rate": 6.253846153846154e-07,
      "loss": -0.0787,
      "reward": 0.774590790271759,
      "reward_std": 0.023398300167173147,
      "reward_std_all": 0.027180170640349388,
      "rewards-max/forensics_sinkhorn_reward": 0.9786276519298553,
      "rewards-max/forensics_strict_edge_reward": 0.3489447683095932,
      "rewards-max/forensics_tiou_reward": 0.9729095101356506,
      "rewards/forensics_sinkhorn_reward": 0.7353460192680359,
      "rewards/forensics_strict_edge_reward": 0.032579309307038784,
      "rewards/forensics_tiou_reward": 0.5340449064970016,
      "step": 488
    },
    {
      "completion_length": 27.09375,
      "epoch": 1.500768049155146,
      "grad_norm": 5.137139320373535,
      "kl": 0.3046875,
      "learning_rate": 6.246153846153845e-07,
      "loss": 0.0255,
      "reward": 0.7238756120204926,
      "reward_std": 0.04331779107451439,
      "reward_std_all": 0.06239942088723183,
      "rewards-max/forensics_sinkhorn_reward": 0.9678837060928345,
      "rewards-max/forensics_strict_edge_reward": 0.3099369704723358,
      "rewards-max/forensics_tiou_reward": 0.9441044926643372,
      "rewards/forensics_sinkhorn_reward": 0.6772441267967224,
      "rewards/forensics_strict_edge_reward": 0.020503249019384384,
      "rewards/forensics_tiou_reward": 0.5114477723836899,
      "step": 489
    },
    {
      "completion_length": 23.3125,
      "epoch": 1.5038402457757296,
      "grad_norm": 5.722896575927734,
      "kl": 0.318359375,
      "learning_rate": 6.238461538461538e-07,
      "loss": -0.0095,
      "reward": 0.7687678337097168,
      "reward_std": 0.03487966815009713,
      "reward_std_all": 0.03776748478412628,
      "rewards-max/forensics_sinkhorn_reward": 0.969984382390976,
      "rewards-max/forensics_strict_edge_reward": 0.4086042046546936,
      "rewards-max/forensics_tiou_reward": 0.951095700263977,
      "rewards/forensics_sinkhorn_reward": 0.7310632169246674,
      "rewards/forensics_strict_edge_reward": 0.030988745391368866,
      "rewards/forensics_tiou_reward": 0.5113185048103333,
      "step": 490
    },
    {
      "completion_length": 22.40625,
      "epoch": 1.5069124423963134,
      "grad_norm": 5.438779830932617,
      "kl": 0.322265625,
      "learning_rate": 6.23076923076923e-07,
      "loss": 0.0367,
      "reward": 0.7823883891105652,
      "reward_std": 0.03174632787704468,
      "reward_std_all": 0.03072855994105339,
      "rewards-max/forensics_sinkhorn_reward": 0.9729318022727966,
      "rewards-max/forensics_strict_edge_reward": 0.5713747441768646,
      "rewards-max/forensics_tiou_reward": 0.9601384699344635,
      "rewards/forensics_sinkhorn_reward": 0.728596955537796,
      "rewards/forensics_strict_edge_reward": 0.034238116815686226,
      "rewards/forensics_tiou_reward": 0.5175522416830063,
      "step": 491
    },
    {
      "completion_length": 25.828125,
      "epoch": 1.509984639016897,
      "grad_norm": 6.306197643280029,
      "kl": 0.333984375,
      "learning_rate": 6.223076923076924e-07,
      "loss": -0.0682,
      "reward": 0.7177316248416901,
      "reward_std": 0.09819769859313965,
      "reward_std_all": 0.09498817101120949,
      "rewards-max/forensics_sinkhorn_reward": 0.9228087663650513,
      "rewards-max/forensics_strict_edge_reward": 0.28554875031113625,
      "rewards-max/forensics_tiou_reward": 0.8470984399318695,
      "rewards/forensics_sinkhorn_reward": 0.5805217325687408,
      "rewards/forensics_strict_edge_reward": 0.02283794421236962,
      "rewards/forensics_tiou_reward": 0.4033891260623932,
      "step": 492
    },
    {
      "completion_length": 25.28125,
      "epoch": 1.5130568356374807,
      "grad_norm": 5.329137325286865,
      "kl": 0.3681640625,
      "learning_rate": 6.215384615384615e-07,
      "loss": -0.0157,
      "reward": 0.6893908083438873,
      "reward_std": 0.0666851494461298,
      "reward_std_all": 0.08951896242797375,
      "rewards-max/forensics_sinkhorn_reward": 0.9759440720081329,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9713954031467438,
      "rewards/forensics_sinkhorn_reward": 0.5937162637710571,
      "rewards/forensics_strict_edge_reward": 0.04904411546885967,
      "rewards/forensics_tiou_reward": 0.4133895933628082,
      "step": 493
    },
    {
      "completion_length": 25.171875,
      "epoch": 1.5161290322580645,
      "grad_norm": 4.926624298095703,
      "kl": 0.271484375,
      "learning_rate": 6.207692307692307e-07,
      "loss": 0.0588,
      "reward": 0.742820531129837,
      "reward_std": 0.03841011039912701,
      "reward_std_all": 0.0520232729613781,
      "rewards-max/forensics_sinkhorn_reward": 0.9806202948093414,
      "rewards-max/forensics_strict_edge_reward": 0.21962502971291542,
      "rewards-max/forensics_tiou_reward": 0.977338045835495,
      "rewards/forensics_sinkhorn_reward": 0.6538054645061493,
      "rewards/forensics_strict_edge_reward": 0.028759022243320942,
      "rewards/forensics_tiou_reward": 0.4866178035736084,
      "step": 494
    },
    {
      "completion_length": 22.953125,
      "epoch": 1.5192012288786483,
      "grad_norm": 5.213372707366943,
      "kl": 0.35400390625,
      "learning_rate": 6.2e-07,
      "loss": 0.1423,
      "reward": 0.7449446618556976,
      "reward_std": 0.09950021654367447,
      "reward_std_all": 0.08866464905440807,
      "rewards-max/forensics_sinkhorn_reward": 0.9828963577747345,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9850572943687439,
      "rewards/forensics_sinkhorn_reward": 0.6796382367610931,
      "rewards/forensics_strict_edge_reward": 0.04832281172275543,
      "rewards/forensics_tiou_reward": 0.45383986830711365,
      "step": 495
    },
    {
      "completion_length": 25.046875,
      "epoch": 1.522273425499232,
      "grad_norm": 5.07951021194458,
      "kl": 0.3671875,
      "learning_rate": 6.192307692307692e-07,
      "loss": 0.1235,
      "reward": 0.7411483526229858,
      "reward_std": 0.027387085370719433,
      "reward_std_all": 0.048005806282162666,
      "rewards-max/forensics_sinkhorn_reward": 0.9728634357452393,
      "rewards-max/forensics_strict_edge_reward": 0.3978807330131531,
      "rewards-max/forensics_tiou_reward": 0.9613547623157501,
      "rewards/forensics_sinkhorn_reward": 0.7206229865550995,
      "rewards/forensics_strict_edge_reward": 0.03423301503062248,
      "rewards/forensics_tiou_reward": 0.5159850120544434,
      "step": 496
    },
    {
      "completion_length": 26.0625,
      "epoch": 1.5253456221198156,
      "grad_norm": 6.279761791229248,
      "kl": 0.41015625,
      "learning_rate": 6.184615384615385e-07,
      "loss": 0.0261,
      "reward": 0.7026030421257019,
      "reward_std": 0.08534626290202141,
      "reward_std_all": 0.09433533251285553,
      "rewards-max/forensics_sinkhorn_reward": 0.9695540964603424,
      "rewards-max/forensics_strict_edge_reward": 0.4555388540029526,
      "rewards-max/forensics_tiou_reward": 0.9438124597072601,
      "rewards/forensics_sinkhorn_reward": 0.6352017521858215,
      "rewards/forensics_strict_edge_reward": 0.02545281406491995,
      "rewards/forensics_tiou_reward": 0.42457468807697296,
      "step": 497
    },
    {
      "completion_length": 24.765625,
      "epoch": 1.5284178187403994,
      "grad_norm": 8.026131629943848,
      "kl": 0.59765625,
      "learning_rate": 6.176923076923076e-07,
      "loss": 0.1525,
      "reward": 0.7713207006454468,
      "reward_std": 0.04315897263586521,
      "reward_std_all": 0.04148993641138077,
      "rewards-max/forensics_sinkhorn_reward": 0.9759218096733093,
      "rewards-max/forensics_strict_edge_reward": 0.5283573269844055,
      "rewards-max/forensics_tiou_reward": 0.9677911698818207,
      "rewards/forensics_sinkhorn_reward": 0.7307253777980804,
      "rewards/forensics_strict_edge_reward": 0.040108487010002136,
      "rewards/forensics_tiou_reward": 0.5422569513320923,
      "step": 498
    },
    {
      "completion_length": 23.0625,
      "epoch": 1.531490015360983,
      "grad_norm": 75.16828155517578,
      "kl": 2.8095703125,
      "learning_rate": 6.169230769230768e-07,
      "loss": 0.1072,
      "reward": 0.7733217179775238,
      "reward_std": 0.05612978618592024,
      "reward_std_all": 0.05050097778439522,
      "rewards-max/forensics_sinkhorn_reward": 0.9826915860176086,
      "rewards-max/forensics_strict_edge_reward": 0.8106843829154968,
      "rewards-max/forensics_tiou_reward": 0.9842797517776489,
      "rewards/forensics_sinkhorn_reward": 0.7543104588985443,
      "rewards/forensics_strict_edge_reward": 0.07222249917685986,
      "rewards/forensics_tiou_reward": 0.5405448377132416,
      "step": 499
    },
    {
      "completion_length": 26.78125,
      "epoch": 1.5345622119815667,
      "grad_norm": 4.617488384246826,
      "kl": 0.3134765625,
      "learning_rate": 6.161538461538462e-07,
      "loss": 0.0009,
      "reward": 0.7646663784980774,
      "reward_std": 0.03804252203553915,
      "reward_std_all": 0.0398858143016696,
      "rewards-max/forensics_sinkhorn_reward": 0.979942113161087,
      "rewards-max/forensics_strict_edge_reward": 0.7633048295974731,
      "rewards-max/forensics_tiou_reward": 0.9746342599391937,
      "rewards/forensics_sinkhorn_reward": 0.711653858423233,
      "rewards/forensics_strict_edge_reward": 0.07823873683810234,
      "rewards/forensics_tiou_reward": 0.5937654674053192,
      "step": 500
    },
    {
      "completion_length": 23.96875,
      "epoch": 1.5376344086021505,
      "grad_norm": 6.20023775100708,
      "kl": 0.359375,
      "learning_rate": 6.153846153846154e-07,
      "loss": 0.0207,
      "reward": 0.7188615202903748,
      "reward_std": 0.06437059119343758,
      "reward_std_all": 0.07454442046582699,
      "rewards-max/forensics_sinkhorn_reward": 0.9630140662193298,
      "rewards-max/forensics_strict_edge_reward": 0.7426987290382385,
      "rewards-max/forensics_tiou_reward": 0.9319506585597992,
      "rewards/forensics_sinkhorn_reward": 0.65704345703125,
      "rewards/forensics_strict_edge_reward": 0.05336066894233227,
      "rewards/forensics_tiou_reward": 0.4758291244506836,
      "step": 501
    },
    {
      "completion_length": 25.53125,
      "epoch": 1.5407066052227343,
      "grad_norm": 5.979954242706299,
      "kl": 0.3408203125,
      "learning_rate": 6.146153846153845e-07,
      "loss": 0.002,
      "reward": 0.7685954868793488,
      "reward_std": 0.030059026554226875,
      "reward_std_all": 0.0361180417239666,
      "rewards-max/forensics_sinkhorn_reward": 0.9716709554195404,
      "rewards-max/forensics_strict_edge_reward": 0.5025001764297485,
      "rewards-max/forensics_tiou_reward": 0.9534036815166473,
      "rewards/forensics_sinkhorn_reward": 0.7186790108680725,
      "rewards/forensics_strict_edge_reward": 0.04363691993057728,
      "rewards/forensics_tiou_reward": 0.5143340229988098,
      "step": 502
    },
    {
      "completion_length": 25.453125,
      "epoch": 1.543778801843318,
      "grad_norm": 4.193984508514404,
      "kl": 0.314453125,
      "learning_rate": 6.138461538461538e-07,
      "loss": -0.0249,
      "reward": 0.749576061964035,
      "reward_std": 0.01972943777218461,
      "reward_std_all": 0.04143771482631564,
      "rewards-max/forensics_sinkhorn_reward": 0.9518750905990601,
      "rewards-max/forensics_strict_edge_reward": 0.4389758426696062,
      "rewards-max/forensics_tiou_reward": 0.8933911621570587,
      "rewards/forensics_sinkhorn_reward": 0.7249266505241394,
      "rewards/forensics_strict_edge_reward": 0.0432321063708514,
      "rewards/forensics_tiou_reward": 0.5248942524194717,
      "step": 503
    },
    {
      "completion_length": 23.015625,
      "epoch": 1.5468509984639018,
      "grad_norm": 6.88003396987915,
      "kl": 0.3857421875,
      "learning_rate": 6.13076923076923e-07,
      "loss": 0.0568,
      "reward": 0.7009950876235962,
      "reward_std": 0.08027425780892372,
      "reward_std_all": 0.08809517323970795,
      "rewards-max/forensics_sinkhorn_reward": 0.9673703014850616,
      "rewards-max/forensics_strict_edge_reward": 0.18004789389669895,
      "rewards-max/forensics_tiou_reward": 0.9417638182640076,
      "rewards/forensics_sinkhorn_reward": 0.5836058259010315,
      "rewards/forensics_strict_edge_reward": 0.014159700775053352,
      "rewards/forensics_tiou_reward": 0.4034224897623062,
      "step": 504
    },
    {
      "completion_length": 23.4375,
      "epoch": 1.5499231950844854,
      "grad_norm": 5.148420810699463,
      "kl": 0.322265625,
      "learning_rate": 6.123076923076923e-07,
      "loss": 0.0607,
      "reward": 0.7883286774158478,
      "reward_std": 0.029755299910902977,
      "reward_std_all": 0.02761128917336464,
      "rewards-max/forensics_sinkhorn_reward": 0.9700647592544556,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9536586999893188,
      "rewards/forensics_sinkhorn_reward": 0.7591460049152374,
      "rewards/forensics_strict_edge_reward": 0.050175029784440994,
      "rewards/forensics_tiou_reward": 0.5380484461784363,
      "step": 505
    },
    {
      "completion_length": 20.796875,
      "epoch": 1.5529953917050692,
      "grad_norm": 5.446613311767578,
      "kl": 0.328125,
      "learning_rate": 6.115384615384616e-07,
      "loss": 0.0505,
      "reward": 0.7051462531089783,
      "reward_std": 0.04785987176001072,
      "reward_std_all": 0.07383213099092245,
      "rewards-max/forensics_sinkhorn_reward": 0.9653002917766571,
      "rewards-max/forensics_strict_edge_reward": 0.29823410511016846,
      "rewards-max/forensics_tiou_reward": 0.9391639530658722,
      "rewards/forensics_sinkhorn_reward": 0.6332238614559174,
      "rewards/forensics_strict_edge_reward": 0.018334973137825727,
      "rewards/forensics_tiou_reward": 0.3846055716276169,
      "step": 506
    },
    {
      "completion_length": 27.46875,
      "epoch": 1.5560675883256527,
      "grad_norm": 3.7569711208343506,
      "kl": 0.3232421875,
      "learning_rate": 6.107692307692307e-07,
      "loss": 0.0517,
      "reward": 0.7575528919696808,
      "reward_std": 0.060191817581653595,
      "reward_std_all": 0.05715783312916756,
      "rewards-max/forensics_sinkhorn_reward": 0.9510732293128967,
      "rewards-max/forensics_strict_edge_reward": 0.37365320324897766,
      "rewards-max/forensics_tiou_reward": 0.9033877849578857,
      "rewards/forensics_sinkhorn_reward": 0.6747860312461853,
      "rewards/forensics_strict_edge_reward": 0.038638120517134666,
      "rewards/forensics_tiou_reward": 0.4730554670095444,
      "step": 507
    },
    {
      "completion_length": 28.953125,
      "epoch": 1.5591397849462365,
      "grad_norm": 5.173722267150879,
      "kl": 0.3701171875,
      "learning_rate": 6.1e-07,
      "loss": 0.0203,
      "reward": 0.7432098388671875,
      "reward_std": 0.04942832142114639,
      "reward_std_all": 0.05404222570359707,
      "rewards-max/forensics_sinkhorn_reward": 0.9487581551074982,
      "rewards-max/forensics_strict_edge_reward": 0.23115165531635284,
      "rewards-max/forensics_tiou_reward": 0.9071009159088135,
      "rewards/forensics_sinkhorn_reward": 0.6398332715034485,
      "rewards/forensics_strict_edge_reward": 0.012951777316629887,
      "rewards/forensics_tiou_reward": 0.4284488707780838,
      "step": 508
    },
    {
      "completion_length": 23.78125,
      "epoch": 1.5622119815668203,
      "grad_norm": 6.431527137756348,
      "kl": 0.388671875,
      "learning_rate": 6.092307692307692e-07,
      "loss": 0.036,
      "reward": 0.7582888007164001,
      "reward_std": 0.0410149022936821,
      "reward_std_all": 0.044835932552814484,
      "rewards-max/forensics_sinkhorn_reward": 0.9743995666503906,
      "rewards-max/forensics_strict_edge_reward": 0.47726282477378845,
      "rewards-max/forensics_tiou_reward": 0.9663092494010925,
      "rewards/forensics_sinkhorn_reward": 0.7138303518295288,
      "rewards/forensics_strict_edge_reward": 0.034117404371500015,
      "rewards/forensics_tiou_reward": 0.5016772747039795,
      "step": 509
    },
    {
      "completion_length": 24.546875,
      "epoch": 1.565284178187404,
      "grad_norm": 6.130740642547607,
      "kl": 0.3291015625,
      "learning_rate": 6.084615384615385e-07,
      "loss": 0.0312,
      "reward": 0.7761818766593933,
      "reward_std": 0.026847453322261572,
      "reward_std_all": 0.031245369464159012,
      "rewards-max/forensics_sinkhorn_reward": 0.9831710457801819,
      "rewards-max/forensics_strict_edge_reward": 0.4592044800519943,
      "rewards-max/forensics_tiou_reward": 0.9811296463012695,
      "rewards/forensics_sinkhorn_reward": 0.707394540309906,
      "rewards/forensics_strict_edge_reward": 0.03449343144893646,
      "rewards/forensics_tiou_reward": 0.5396591424942017,
      "step": 510
    },
    {
      "completion_length": 24.84375,
      "epoch": 1.5683563748079878,
      "grad_norm": 6.676623344421387,
      "kl": 0.392578125,
      "learning_rate": 6.076923076923076e-07,
      "loss": -0.0517,
      "reward": 0.7834282517433167,
      "reward_std": 0.02993514807894826,
      "reward_std_all": 0.02898330893367529,
      "rewards-max/forensics_sinkhorn_reward": 0.9675458669662476,
      "rewards-max/forensics_strict_edge_reward": 0.33960409462451935,
      "rewards-max/forensics_tiou_reward": 0.9466287195682526,
      "rewards/forensics_sinkhorn_reward": 0.7388473749160767,
      "rewards/forensics_strict_edge_reward": 0.038966722786426544,
      "rewards/forensics_tiou_reward": 0.5553527921438217,
      "step": 511
    },
    {
      "completion_length": 23.234375,
      "epoch": 1.5714285714285714,
      "grad_norm": 7.1581950187683105,
      "kl": 0.326171875,
      "learning_rate": 6.069230769230768e-07,
      "loss": -0.0164,
      "reward": 0.7152305245399475,
      "reward_std": 0.0660330094397068,
      "reward_std_all": 0.08055930584669113,
      "rewards-max/forensics_sinkhorn_reward": 0.9699397385120392,
      "rewards-max/forensics_strict_edge_reward": 0.43953123688697815,
      "rewards-max/forensics_tiou_reward": 0.9540151655673981,
      "rewards/forensics_sinkhorn_reward": 0.6562506556510925,
      "rewards/forensics_strict_edge_reward": 0.03892887756228447,
      "rewards/forensics_tiou_reward": 0.4869697093963623,
      "step": 512
    },
    {
      "completion_length": 22.84375,
      "epoch": 1.5745007680491552,
      "grad_norm": 7.502630710601807,
      "kl": 0.306640625,
      "learning_rate": 6.061538461538462e-07,
      "loss": 0.1036,
      "reward": 0.6793486773967743,
      "reward_std": 0.08347989618778229,
      "reward_std_all": 0.10419880971312523,
      "rewards-max/forensics_sinkhorn_reward": 0.9603855311870575,
      "rewards-max/forensics_strict_edge_reward": 0.43993423879146576,
      "rewards-max/forensics_tiou_reward": 0.9258522987365723,
      "rewards/forensics_sinkhorn_reward": 0.6016665101051331,
      "rewards/forensics_strict_edge_reward": 0.02960457280278206,
      "rewards/forensics_tiou_reward": 0.3812360167503357,
      "step": 513
    },
    {
      "completion_length": 24.828125,
      "epoch": 1.5775729646697387,
      "grad_norm": 4.640111446380615,
      "kl": 0.3896484375,
      "learning_rate": 6.053846153846154e-07,
      "loss": 0.0805,
      "reward": 0.7652702331542969,
      "reward_std": 0.02464998885989189,
      "reward_std_all": 0.034605441614985466,
      "rewards-max/forensics_sinkhorn_reward": 0.9781739711761475,
      "rewards-max/forensics_strict_edge_reward": 0.6082656979560852,
      "rewards-max/forensics_tiou_reward": 0.9731194078922272,
      "rewards/forensics_sinkhorn_reward": 0.7346929907798767,
      "rewards/forensics_strict_edge_reward": 0.05284910649061203,
      "rewards/forensics_tiou_reward": 0.5177166610956192,
      "step": 514
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.5806451612903225,
      "grad_norm": 5.511927127838135,
      "kl": 0.48828125,
      "learning_rate": 6.046153846153847e-07,
      "loss": 0.0014,
      "reward": 0.7093787491321564,
      "reward_std": 0.03073799330741167,
      "reward_std_all": 0.06413943320512772,
      "rewards-max/forensics_sinkhorn_reward": 0.9323551952838898,
      "rewards-max/forensics_strict_edge_reward": 0.38447941094636917,
      "rewards-max/forensics_tiou_reward": 0.8300945460796356,
      "rewards/forensics_sinkhorn_reward": 0.6237218379974365,
      "rewards/forensics_strict_edge_reward": 0.02016397751867771,
      "rewards/forensics_tiou_reward": 0.38030025362968445,
      "step": 515
    },
    {
      "completion_length": 23.421875,
      "epoch": 1.5837173579109063,
      "grad_norm": 4.693962574005127,
      "kl": 0.3154296875,
      "learning_rate": 6.038461538461538e-07,
      "loss": -0.0025,
      "reward": 0.7745187282562256,
      "reward_std": 0.04023334523662925,
      "reward_std_all": 0.038788420613855124,
      "rewards-max/forensics_sinkhorn_reward": 0.981503039598465,
      "rewards-max/forensics_strict_edge_reward": 0.5471633523702621,
      "rewards-max/forensics_tiou_reward": 0.9785449206829071,
      "rewards/forensics_sinkhorn_reward": 0.7362679541110992,
      "rewards/forensics_strict_edge_reward": 0.0480409637093544,
      "rewards/forensics_tiou_reward": 0.5588582754135132,
      "step": 516
    },
    {
      "completion_length": 22.40625,
      "epoch": 1.58678955453149,
      "grad_norm": 4.63372802734375,
      "kl": 0.4150390625,
      "learning_rate": 6.03076923076923e-07,
      "loss": 0.021,
      "reward": 0.757668524980545,
      "reward_std": 0.04364372417330742,
      "reward_std_all": 0.05021290294826031,
      "rewards-max/forensics_sinkhorn_reward": 0.981042206287384,
      "rewards-max/forensics_strict_edge_reward": 0.44672510027885437,
      "rewards-max/forensics_tiou_reward": 0.9788085520267487,
      "rewards/forensics_sinkhorn_reward": 0.7237284481525421,
      "rewards/forensics_strict_edge_reward": 0.04382279282435775,
      "rewards/forensics_tiou_reward": 0.5076935589313507,
      "step": 517
    },
    {
      "completion_length": 23.953125,
      "epoch": 1.5898617511520738,
      "grad_norm": 4.919848918914795,
      "kl": 0.3818359375,
      "learning_rate": 6.023076923076923e-07,
      "loss": -0.0726,
      "reward": 0.7491653263568878,
      "reward_std": 0.057542994152754545,
      "reward_std_all": 0.0574630624614656,
      "rewards-max/forensics_sinkhorn_reward": 0.9773454070091248,
      "rewards-max/forensics_strict_edge_reward": 0.43496479094028473,
      "rewards-max/forensics_tiou_reward": 0.9683101177215576,
      "rewards/forensics_sinkhorn_reward": 0.6807710528373718,
      "rewards/forensics_strict_edge_reward": 0.03845532797276974,
      "rewards/forensics_tiou_reward": 0.5354616791009903,
      "step": 518
    },
    {
      "completion_length": 22.25,
      "epoch": 1.5929339477726574,
      "grad_norm": 5.097001075744629,
      "kl": 0.3291015625,
      "learning_rate": 6.015384615384615e-07,
      "loss": -0.0645,
      "reward": 0.7710080742835999,
      "reward_std": 0.01694125309586525,
      "reward_std_all": 0.0250987708568573,
      "rewards-max/forensics_sinkhorn_reward": 0.9817072153091431,
      "rewards-max/forensics_strict_edge_reward": 0.2975686192512512,
      "rewards-max/forensics_tiou_reward": 0.978899210691452,
      "rewards/forensics_sinkhorn_reward": 0.7319058775901794,
      "rewards/forensics_strict_edge_reward": 0.025417760014533997,
      "rewards/forensics_tiou_reward": 0.5139302462339401,
      "step": 519
    },
    {
      "completion_length": 24.625,
      "epoch": 1.5960061443932412,
      "grad_norm": 6.048834323883057,
      "kl": 0.4140625,
      "learning_rate": 6.007692307692307e-07,
      "loss": 0.0896,
      "reward": 0.7353867888450623,
      "reward_std": 0.03989720158278942,
      "reward_std_all": 0.05381004698574543,
      "rewards-max/forensics_sinkhorn_reward": 0.9770475924015045,
      "rewards-max/forensics_strict_edge_reward": 0.40624599158763885,
      "rewards-max/forensics_tiou_reward": 0.9546754658222198,
      "rewards/forensics_sinkhorn_reward": 0.6595996916294098,
      "rewards/forensics_strict_edge_reward": 0.029646183364093304,
      "rewards/forensics_tiou_reward": 0.41596077382564545,
      "step": 520
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.5990783410138247,
      "grad_norm": 6.994251728057861,
      "kl": 0.3466796875,
      "learning_rate": 6e-07,
      "loss": 0.0154,
      "reward": 0.7455363273620605,
      "reward_std": 0.06825098022818565,
      "reward_std_all": 0.066363625228405,
      "rewards-max/forensics_sinkhorn_reward": 0.9376793801784515,
      "rewards-max/forensics_strict_edge_reward": 0.2499820590019226,
      "rewards-max/forensics_tiou_reward": 0.8798147737979889,
      "rewards/forensics_sinkhorn_reward": 0.6625109910964966,
      "rewards/forensics_strict_edge_reward": 0.02228048793040216,
      "rewards/forensics_tiou_reward": 0.4183476120233536,
      "step": 521
    },
    {
      "completion_length": 22.4375,
      "epoch": 1.6021505376344085,
      "grad_norm": 7.10900354385376,
      "kl": 0.369140625,
      "learning_rate": 5.992307692307692e-07,
      "loss": 0.0236,
      "reward": 0.7520388662815094,
      "reward_std": 0.06717575713992119,
      "reward_std_all": 0.06361652165651321,
      "rewards-max/forensics_sinkhorn_reward": 0.9523348212242126,
      "rewards-max/forensics_strict_edge_reward": 0.13945549726486206,
      "rewards-max/forensics_tiou_reward": 0.910289466381073,
      "rewards/forensics_sinkhorn_reward": 0.6918788552284241,
      "rewards/forensics_strict_edge_reward": 0.011484441347420216,
      "rewards/forensics_tiou_reward": 0.46542705595493317,
      "step": 522
    },
    {
      "completion_length": 25.65625,
      "epoch": 1.6052227342549923,
      "grad_norm": 6.70128870010376,
      "kl": 0.3271484375,
      "learning_rate": 5.984615384615385e-07,
      "loss": -0.0093,
      "reward": 0.7511451840400696,
      "reward_std": 0.03728553652763367,
      "reward_std_all": 0.045778023079037666,
      "rewards-max/forensics_sinkhorn_reward": 0.9725368916988373,
      "rewards-max/forensics_strict_edge_reward": 0.47595588862895966,
      "rewards-max/forensics_tiou_reward": 0.9534901976585388,
      "rewards/forensics_sinkhorn_reward": 0.7113662362098694,
      "rewards/forensics_strict_edge_reward": 0.052929021418094635,
      "rewards/forensics_tiou_reward": 0.5049398839473724,
      "step": 523
    },
    {
      "completion_length": 27.328125,
      "epoch": 1.608294930875576,
      "grad_norm": 5.852371692657471,
      "kl": 0.326171875,
      "learning_rate": 5.976923076923077e-07,
      "loss": -0.0039,
      "reward": 0.7395848035812378,
      "reward_std": 0.05173342116177082,
      "reward_std_all": 0.05758237838745117,
      "rewards-max/forensics_sinkhorn_reward": 0.9696585834026337,
      "rewards-max/forensics_strict_edge_reward": 0.4937075823545456,
      "rewards-max/forensics_tiou_reward": 0.9471540451049805,
      "rewards/forensics_sinkhorn_reward": 0.6480863690376282,
      "rewards/forensics_strict_edge_reward": 0.04178990237414837,
      "rewards/forensics_tiou_reward": 0.4576157033443451,
      "step": 524
    },
    {
      "completion_length": 25.328125,
      "epoch": 1.6113671274961598,
      "grad_norm": 5.773329734802246,
      "kl": 0.4072265625,
      "learning_rate": 5.969230769230768e-07,
      "loss": 0.0657,
      "reward": 0.6746137738227844,
      "reward_std": 0.12466683983802795,
      "reward_std_all": 0.12462234124541283,
      "rewards-max/forensics_sinkhorn_reward": 0.8930002748966217,
      "rewards-max/forensics_strict_edge_reward": 0.2963643968105316,
      "rewards-max/forensics_tiou_reward": 0.8315384089946747,
      "rewards/forensics_sinkhorn_reward": 0.5379731357097626,
      "rewards/forensics_strict_edge_reward": 0.01621901523321867,
      "rewards/forensics_tiou_reward": 0.3456367701292038,
      "step": 525
    },
    {
      "completion_length": 26.71875,
      "epoch": 1.6144393241167436,
      "grad_norm": 5.054827690124512,
      "kl": 0.46875,
      "learning_rate": 5.961538461538461e-07,
      "loss": 0.1112,
      "reward": 0.7482341825962067,
      "reward_std": 0.06474053859710693,
      "reward_std_all": 0.06229913420975208,
      "rewards-max/forensics_sinkhorn_reward": 0.9640743732452393,
      "rewards-max/forensics_strict_edge_reward": 0.38989683985710144,
      "rewards-max/forensics_tiou_reward": 0.9413377344608307,
      "rewards/forensics_sinkhorn_reward": 0.6929445266723633,
      "rewards/forensics_strict_edge_reward": 0.044007010757923126,
      "rewards/forensics_tiou_reward": 0.5067233890295029,
      "step": 526
    },
    {
      "completion_length": 25.25,
      "epoch": 1.6175115207373272,
      "grad_norm": 5.035441875457764,
      "kl": 0.3515625,
      "learning_rate": 5.953846153846154e-07,
      "loss": 0.0274,
      "reward": 0.756700187921524,
      "reward_std": 0.039169959258288145,
      "reward_std_all": 0.04302080534398556,
      "rewards-max/forensics_sinkhorn_reward": 0.9805004894733429,
      "rewards-max/forensics_strict_edge_reward": 0.3054930008947849,
      "rewards-max/forensics_tiou_reward": 0.9721026122570038,
      "rewards/forensics_sinkhorn_reward": 0.7159624993801117,
      "rewards/forensics_strict_edge_reward": 0.02121185278519988,
      "rewards/forensics_tiou_reward": 0.532558411359787,
      "step": 527
    },
    {
      "completion_length": 21.328125,
      "epoch": 1.620583717357911,
      "grad_norm": 6.60427713394165,
      "kl": 0.2998046875,
      "learning_rate": 5.946153846153847e-07,
      "loss": 0.04,
      "reward": 0.7513524293899536,
      "reward_std": 0.027814363595098257,
      "reward_std_all": 0.042764214565977454,
      "rewards-max/forensics_sinkhorn_reward": 0.9731715619564056,
      "rewards-max/forensics_strict_edge_reward": 0.45684996247291565,
      "rewards-max/forensics_tiou_reward": 0.9518038630485535,
      "rewards/forensics_sinkhorn_reward": 0.685475766658783,
      "rewards/forensics_strict_edge_reward": 0.027382273226976395,
      "rewards/forensics_tiou_reward": 0.4736974537372589,
      "step": 528
    },
    {
      "completion_length": 24.375,
      "epoch": 1.6236559139784945,
      "grad_norm": 6.856746196746826,
      "kl": 0.380859375,
      "learning_rate": 5.938461538461538e-07,
      "loss": 0.0621,
      "reward": 0.7080224752426147,
      "reward_std": 0.08918561413884163,
      "reward_std_all": 0.09104837849736214,
      "rewards-max/forensics_sinkhorn_reward": 0.9544336497783661,
      "rewards-max/forensics_strict_edge_reward": 0.3715772796422243,
      "rewards-max/forensics_tiou_reward": 0.8955654501914978,
      "rewards/forensics_sinkhorn_reward": 0.6201943159103394,
      "rewards/forensics_strict_edge_reward": 0.02223236206918955,
      "rewards/forensics_tiou_reward": 0.40656544268131256,
      "step": 529
    },
    {
      "completion_length": 24.25,
      "epoch": 1.6267281105990783,
      "grad_norm": 6.474728584289551,
      "kl": 0.345703125,
      "learning_rate": 5.93076923076923e-07,
      "loss": 0.0882,
      "reward": 0.7467065155506134,
      "reward_std": 0.061938995495438576,
      "reward_std_all": 0.06120816431939602,
      "rewards-max/forensics_sinkhorn_reward": 0.9473898112773895,
      "rewards-max/forensics_strict_edge_reward": 0.3757830262184143,
      "rewards-max/forensics_tiou_reward": 0.9098368585109711,
      "rewards/forensics_sinkhorn_reward": 0.7024423778057098,
      "rewards/forensics_strict_edge_reward": 0.032835966907441616,
      "rewards/forensics_tiou_reward": 0.4846417307853699,
      "step": 530
    },
    {
      "completion_length": 23.234375,
      "epoch": 1.629800307219662,
      "grad_norm": 6.5390238761901855,
      "kl": 0.326171875,
      "learning_rate": 5.923076923076923e-07,
      "loss": 0.1789,
      "reward": 0.6714667677879333,
      "reward_std": 0.07537243515253067,
      "reward_std_all": 0.10159574076533318,
      "rewards-max/forensics_sinkhorn_reward": 0.9680726528167725,
      "rewards-max/forensics_strict_edge_reward": 0.36827364563941956,
      "rewards-max/forensics_tiou_reward": 0.9463714957237244,
      "rewards/forensics_sinkhorn_reward": 0.5854615271091461,
      "rewards/forensics_strict_edge_reward": 0.027744916267693043,
      "rewards/forensics_tiou_reward": 0.43479104340076447,
      "step": 531
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.6328725038402458,
      "grad_norm": 6.531248092651367,
      "kl": 0.298828125,
      "learning_rate": 5.915384615384615e-07,
      "loss": 0.1049,
      "reward": 0.7441021502017975,
      "reward_std": 0.05471925064921379,
      "reward_std_all": 0.058554306626319885,
      "rewards-max/forensics_sinkhorn_reward": 0.9288320541381836,
      "rewards-max/forensics_strict_edge_reward": 0.31528612424153835,
      "rewards-max/forensics_tiou_reward": 0.8537329137325287,
      "rewards/forensics_sinkhorn_reward": 0.6343007683753967,
      "rewards/forensics_strict_edge_reward": 0.013537119957618415,
      "rewards/forensics_tiou_reward": 0.4352460652589798,
      "step": 532
    },
    {
      "completion_length": 24.171875,
      "epoch": 1.6359447004608296,
      "grad_norm": 7.064146518707275,
      "kl": 0.4150390625,
      "learning_rate": 5.907692307692307e-07,
      "loss": 0.1004,
      "reward": 0.7383491694927216,
      "reward_std": 0.0872659757733345,
      "reward_std_all": 0.08255316503345966,
      "rewards-max/forensics_sinkhorn_reward": 0.9716726541519165,
      "rewards-max/forensics_strict_edge_reward": 0.041760344291105866,
      "rewards-max/forensics_tiou_reward": 0.9506784379482269,
      "rewards/forensics_sinkhorn_reward": 0.6693081259727478,
      "rewards/forensics_strict_edge_reward": 0.0055246943375095725,
      "rewards/forensics_tiou_reward": 0.48487451672554016,
      "step": 533
    },
    {
      "completion_length": 22.671875,
      "epoch": 1.6390168970814132,
      "grad_norm": 7.00565242767334,
      "kl": 0.5625,
      "learning_rate": 5.9e-07,
      "loss": 0.0698,
      "reward": 0.7423146367073059,
      "reward_std": 0.04380875080823898,
      "reward_std_all": 0.05674717016518116,
      "rewards-max/forensics_sinkhorn_reward": 0.9422841668128967,
      "rewards-max/forensics_strict_edge_reward": 0.10037059988826513,
      "rewards-max/forensics_tiou_reward": 0.8917503654956818,
      "rewards/forensics_sinkhorn_reward": 0.681388795375824,
      "rewards/forensics_strict_edge_reward": 0.008930786862038076,
      "rewards/forensics_tiou_reward": 0.44360509514808655,
      "step": 534
    },
    {
      "completion_length": 24.09375,
      "epoch": 1.642089093701997,
      "grad_norm": 5.589056015014648,
      "kl": 0.2763671875,
      "learning_rate": 5.892307692307692e-07,
      "loss": 0.0845,
      "reward": 0.7335246801376343,
      "reward_std": 0.08617415651679039,
      "reward_std_all": 0.08533555269241333,
      "rewards-max/forensics_sinkhorn_reward": 0.9654151797294617,
      "rewards-max/forensics_strict_edge_reward": 0.21903837099671364,
      "rewards-max/forensics_tiou_reward": 0.9427486956119537,
      "rewards/forensics_sinkhorn_reward": 0.6809166967868805,
      "rewards/forensics_strict_edge_reward": 0.017239855602383614,
      "rewards/forensics_tiou_reward": 0.46477119624614716,
      "step": 535
    },
    {
      "completion_length": 25.71875,
      "epoch": 1.6451612903225805,
      "grad_norm": 5.11735200881958,
      "kl": 0.3408203125,
      "learning_rate": 5.884615384615385e-07,
      "loss": 0.0088,
      "reward": 0.7714394927024841,
      "reward_std": 0.04632697068154812,
      "reward_std_all": 0.04439761396497488,
      "rewards-max/forensics_sinkhorn_reward": 0.9531357288360596,
      "rewards-max/forensics_strict_edge_reward": 0.40273305773735046,
      "rewards-max/forensics_tiou_reward": 0.9087571203708649,
      "rewards/forensics_sinkhorn_reward": 0.6928305625915527,
      "rewards/forensics_strict_edge_reward": 0.02701747976243496,
      "rewards/forensics_tiou_reward": 0.4655648320913315,
      "step": 536
    },
    {
      "completion_length": 22.84375,
      "epoch": 1.6482334869431643,
      "grad_norm": 6.095886707305908,
      "kl": 0.3115234375,
      "learning_rate": 5.876923076923077e-07,
      "loss": 0.0747,
      "reward": 0.7294376492500305,
      "reward_std": 0.05884767323732376,
      "reward_std_all": 0.06612242013216019,
      "rewards-max/forensics_sinkhorn_reward": 0.9702459871768951,
      "rewards-max/forensics_strict_edge_reward": 0.3126997575163841,
      "rewards-max/forensics_tiou_reward": 0.9521156549453735,
      "rewards/forensics_sinkhorn_reward": 0.6430186629295349,
      "rewards/forensics_strict_edge_reward": 0.014811321860179305,
      "rewards/forensics_tiou_reward": 0.40749309957027435,
      "step": 537
    },
    {
      "completion_length": 21.8125,
      "epoch": 1.651305683563748,
      "grad_norm": 6.054629325866699,
      "kl": 0.396484375,
      "learning_rate": 5.869230769230768e-07,
      "loss": 0.0058,
      "reward": 0.7533762454986572,
      "reward_std": 0.05563018284738064,
      "reward_std_all": 0.05509401299059391,
      "rewards-max/forensics_sinkhorn_reward": 0.9239205420017242,
      "rewards-max/forensics_strict_edge_reward": 0.1539987213909626,
      "rewards-max/forensics_tiou_reward": 0.84527787566185,
      "rewards/forensics_sinkhorn_reward": 0.6954465508460999,
      "rewards/forensics_strict_edge_reward": 0.012173502705991268,
      "rewards/forensics_tiou_reward": 0.4367193579673767,
      "step": 538
    },
    {
      "completion_length": 26.359375,
      "epoch": 1.6543778801843319,
      "grad_norm": 4.559239387512207,
      "kl": 0.4091796875,
      "learning_rate": 5.861538461538461e-07,
      "loss": 0.0766,
      "reward": 0.745850145816803,
      "reward_std": 0.06445810571312904,
      "reward_std_all": 0.06308176275342703,
      "rewards-max/forensics_sinkhorn_reward": 0.9803051054477692,
      "rewards-max/forensics_strict_edge_reward": 0.4468829333782196,
      "rewards-max/forensics_tiou_reward": 0.9761349558830261,
      "rewards/forensics_sinkhorn_reward": 0.6242191195487976,
      "rewards/forensics_strict_edge_reward": 0.059226082637906075,
      "rewards/forensics_tiou_reward": 0.5067685544490814,
      "step": 539
    },
    {
      "completion_length": 24.375,
      "epoch": 1.6574500768049156,
      "grad_norm": 7.081478118896484,
      "kl": 0.337890625,
      "learning_rate": 5.853846153846154e-07,
      "loss": 0.0566,
      "reward": 0.7126529216766357,
      "reward_std": 0.10146580636501312,
      "reward_std_all": 0.09763317182660103,
      "rewards-max/forensics_sinkhorn_reward": 0.9871649444103241,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9945827126502991,
      "rewards/forensics_sinkhorn_reward": 0.6313905715942383,
      "rewards/forensics_strict_edge_reward": 0.07075362280011177,
      "rewards/forensics_tiou_reward": 0.42465899884700775,
      "step": 540
    },
    {
      "completion_length": 28.078125,
      "epoch": 1.6605222734254992,
      "grad_norm": 5.452181339263916,
      "kl": 0.341796875,
      "learning_rate": 5.846153846153847e-07,
      "loss": 0.1363,
      "reward": 0.7415655851364136,
      "reward_std": 0.07382855750620365,
      "reward_std_all": 0.07160178013145924,
      "rewards-max/forensics_sinkhorn_reward": 0.9373442530632019,
      "rewards-max/forensics_strict_edge_reward": 0.41692329943180084,
      "rewards-max/forensics_tiou_reward": 0.890419602394104,
      "rewards/forensics_sinkhorn_reward": 0.6464482545852661,
      "rewards/forensics_strict_edge_reward": 0.03680534288287163,
      "rewards/forensics_tiou_reward": 0.45003271102905273,
      "step": 541
    },
    {
      "completion_length": 26.171875,
      "epoch": 1.663594470046083,
      "grad_norm": 4.97829532623291,
      "kl": 0.33203125,
      "learning_rate": 5.838461538461538e-07,
      "loss": 0.133,
      "reward": 0.7703864276409149,
      "reward_std": 0.04831802286207676,
      "reward_std_all": 0.04569857753813267,
      "rewards-max/forensics_sinkhorn_reward": 0.9529300928115845,
      "rewards-max/forensics_strict_edge_reward": 0.33643411402590573,
      "rewards-max/forensics_tiou_reward": 0.9060820639133453,
      "rewards/forensics_sinkhorn_reward": 0.6644277572631836,
      "rewards/forensics_strict_edge_reward": 0.017991874570725486,
      "rewards/forensics_tiou_reward": 0.45373372733592987,
      "step": 542
    },
    {
      "completion_length": 25.46875,
      "epoch": 1.6666666666666665,
      "grad_norm": 5.331282615661621,
      "kl": 0.44140625,
      "learning_rate": 5.83076923076923e-07,
      "loss": -0.0633,
      "reward": 0.7280518412590027,
      "reward_std": 0.05733721889555454,
      "reward_std_all": 0.06520824506878853,
      "rewards-max/forensics_sinkhorn_reward": 0.9800901710987091,
      "rewards-max/forensics_strict_edge_reward": 0.6279524266719818,
      "rewards-max/forensics_tiou_reward": 0.973116546869278,
      "rewards/forensics_sinkhorn_reward": 0.717336118221283,
      "rewards/forensics_strict_edge_reward": 0.060031330212950706,
      "rewards/forensics_tiou_reward": 0.5187206268310547,
      "step": 543
    },
    {
      "completion_length": 23.75,
      "epoch": 1.6697388632872503,
      "grad_norm": 5.381534099578857,
      "kl": 0.3798828125,
      "learning_rate": 5.823076923076923e-07,
      "loss": 0.0181,
      "reward": 0.7367962300777435,
      "reward_std": 0.07020734623074532,
      "reward_std_all": 0.06981435045599937,
      "rewards-max/forensics_sinkhorn_reward": 0.9594926536083221,
      "rewards-max/forensics_strict_edge_reward": 0.37608956545591354,
      "rewards-max/forensics_tiou_reward": 0.9264669120311737,
      "rewards/forensics_sinkhorn_reward": 0.6810107827186584,
      "rewards/forensics_strict_edge_reward": 0.018337749410420656,
      "rewards/forensics_tiou_reward": 0.5141596794128418,
      "step": 544
    },
    {
      "completion_length": 29.421875,
      "epoch": 1.672811059907834,
      "grad_norm": 5.637319087982178,
      "kl": 0.3203125,
      "learning_rate": 5.815384615384615e-07,
      "loss": 0.0194,
      "reward": 0.7394558787345886,
      "reward_std": 0.03876032680273056,
      "reward_std_all": 0.05382101982831955,
      "rewards-max/forensics_sinkhorn_reward": 0.9729730784893036,
      "rewards-max/forensics_strict_edge_reward": 0.3523504212498665,
      "rewards-max/forensics_tiou_reward": 0.9560056626796722,
      "rewards/forensics_sinkhorn_reward": 0.6299072206020355,
      "rewards/forensics_strict_edge_reward": 0.027557848021388054,
      "rewards/forensics_tiou_reward": 0.44879716634750366,
      "step": 545
    },
    {
      "completion_length": 23.8125,
      "epoch": 1.6758832565284179,
      "grad_norm": 4.505317687988281,
      "kl": 0.3857421875,
      "learning_rate": 5.807692307692307e-07,
      "loss": 0.0926,
      "reward": 0.7583434879779816,
      "reward_std": 0.061736563220620155,
      "reward_std_all": 0.05863039940595627,
      "rewards-max/forensics_sinkhorn_reward": 0.9626280963420868,
      "rewards-max/forensics_strict_edge_reward": 0.4227834939956665,
      "rewards-max/forensics_tiou_reward": 0.9267497956752777,
      "rewards/forensics_sinkhorn_reward": 0.6596827208995819,
      "rewards/forensics_strict_edge_reward": 0.025631817057728767,
      "rewards/forensics_tiou_reward": 0.4439043253660202,
      "step": 546
    },
    {
      "completion_length": 22.609375,
      "epoch": 1.6789554531490016,
      "grad_norm": 6.312270641326904,
      "kl": 0.30078125,
      "learning_rate": 5.8e-07,
      "loss": 0.0157,
      "reward": 0.7648011445999146,
      "reward_std": 0.06575698405504227,
      "reward_std_all": 0.0593939945101738,
      "rewards-max/forensics_sinkhorn_reward": 0.9790797829627991,
      "rewards-max/forensics_strict_edge_reward": 0.5880055129528046,
      "rewards-max/forensics_tiou_reward": 0.9768413603305817,
      "rewards/forensics_sinkhorn_reward": 0.710275799036026,
      "rewards/forensics_strict_edge_reward": 0.03461919166147709,
      "rewards/forensics_tiou_reward": 0.5092556029558182,
      "step": 547
    },
    {
      "completion_length": 26.6875,
      "epoch": 1.6820276497695854,
      "grad_norm": 6.373841285705566,
      "kl": 0.5859375,
      "learning_rate": 5.792307692307692e-07,
      "loss": 0.1739,
      "reward": 0.7072791159152985,
      "reward_std": 0.05883318744599819,
      "reward_std_all": 0.08101744204759598,
      "rewards-max/forensics_sinkhorn_reward": 0.9387692809104919,
      "rewards-max/forensics_strict_edge_reward": 0.17858605831861496,
      "rewards-max/forensics_tiou_reward": 0.8864566385746002,
      "rewards/forensics_sinkhorn_reward": 0.634846031665802,
      "rewards/forensics_strict_edge_reward": 0.012014395091682673,
      "rewards/forensics_tiou_reward": 0.4626957178115845,
      "step": 548
    },
    {
      "completion_length": 23.34375,
      "epoch": 1.685099846390169,
      "grad_norm": 5.8698811531066895,
      "kl": 0.3525390625,
      "learning_rate": 5.784615384615385e-07,
      "loss": 0.0855,
      "reward": 0.7397603988647461,
      "reward_std": 0.08434323593974113,
      "reward_std_all": 0.07873711735010147,
      "rewards-max/forensics_sinkhorn_reward": 0.9591924846172333,
      "rewards-max/forensics_strict_edge_reward": 0.42207689583301544,
      "rewards-max/forensics_tiou_reward": 0.9236611127853394,
      "rewards/forensics_sinkhorn_reward": 0.6776601672172546,
      "rewards/forensics_strict_edge_reward": 0.027286095544695854,
      "rewards/forensics_tiou_reward": 0.46925297379493713,
      "step": 549
    },
    {
      "completion_length": 25.15625,
      "epoch": 1.6881720430107527,
      "grad_norm": 6.407999038696289,
      "kl": 0.3359375,
      "learning_rate": 5.776923076923077e-07,
      "loss": 0.0129,
      "reward": 0.7673376798629761,
      "reward_std": 0.042356349527835846,
      "reward_std_all": 0.04225606983527541,
      "rewards-max/forensics_sinkhorn_reward": 0.9608651697635651,
      "rewards-max/forensics_strict_edge_reward": 0.5900419056415558,
      "rewards-max/forensics_tiou_reward": 0.9334000647068024,
      "rewards/forensics_sinkhorn_reward": 0.705266922712326,
      "rewards/forensics_strict_edge_reward": 0.03550201095640659,
      "rewards/forensics_tiou_reward": 0.512367844581604,
      "step": 550
    },
    {
      "completion_length": 23.203125,
      "epoch": 1.6912442396313363,
      "grad_norm": 5.623291492462158,
      "kl": 0.3701171875,
      "learning_rate": 5.769230769230768e-07,
      "loss": 0.1133,
      "reward": 0.7500719130039215,
      "reward_std": 0.05951529089361429,
      "reward_std_all": 0.05867331847548485,
      "rewards-max/forensics_sinkhorn_reward": 0.9228000938892365,
      "rewards-max/forensics_strict_edge_reward": 0.48635128629393876,
      "rewards-max/forensics_tiou_reward": 0.8196358978748322,
      "rewards/forensics_sinkhorn_reward": 0.6637808084487915,
      "rewards/forensics_strict_edge_reward": 0.0306228251720313,
      "rewards/forensics_tiou_reward": 0.48361000418663025,
      "step": 551
    },
    {
      "completion_length": 26.546875,
      "epoch": 1.69431643625192,
      "grad_norm": 6.0276899337768555,
      "kl": 0.3798828125,
      "learning_rate": 5.761538461538461e-07,
      "loss": 0.1688,
      "reward": 0.7203441262245178,
      "reward_std": 0.05182589963078499,
      "reward_std_all": 0.0684599056839943,
      "rewards-max/forensics_sinkhorn_reward": 0.9485679566860199,
      "rewards-max/forensics_strict_edge_reward": 0.44488054513931274,
      "rewards-max/forensics_tiou_reward": 0.9099678993225098,
      "rewards/forensics_sinkhorn_reward": 0.6375686824321747,
      "rewards/forensics_strict_edge_reward": 0.030970517545938492,
      "rewards/forensics_tiou_reward": 0.45196403563022614,
      "step": 552
    },
    {
      "completion_length": 28.9375,
      "epoch": 1.6973886328725039,
      "grad_norm": 5.066876411437988,
      "kl": 0.349609375,
      "learning_rate": 5.753846153846153e-07,
      "loss": 0.0841,
      "reward": 0.7561882138252258,
      "reward_std": 0.0460730604827404,
      "reward_std_all": 0.04947194829583168,
      "rewards-max/forensics_sinkhorn_reward": 0.9688183665275574,
      "rewards-max/forensics_strict_edge_reward": 0.37395593523979187,
      "rewards-max/forensics_tiou_reward": 0.9471230208873749,
      "rewards/forensics_sinkhorn_reward": 0.6830993890762329,
      "rewards/forensics_strict_edge_reward": 0.03263801150023937,
      "rewards/forensics_tiou_reward": 0.4987054020166397,
      "step": 553
    },
    {
      "completion_length": 28.515625,
      "epoch": 1.7004608294930876,
      "grad_norm": 5.634526252746582,
      "kl": 0.36328125,
      "learning_rate": 5.746153846153847e-07,
      "loss": 0.1157,
      "reward": 0.7654836475849152,
      "reward_std": 0.028109611943364143,
      "reward_std_all": 0.03530100733041763,
      "rewards-max/forensics_sinkhorn_reward": 0.973230242729187,
      "rewards-max/forensics_strict_edge_reward": 0.18065907806158066,
      "rewards-max/forensics_tiou_reward": 0.9584033489227295,
      "rewards/forensics_sinkhorn_reward": 0.6600913107395172,
      "rewards/forensics_strict_edge_reward": 0.02432065736502409,
      "rewards/forensics_tiou_reward": 0.4989143908023834,
      "step": 554
    },
    {
      "completion_length": 26.859375,
      "epoch": 1.7035330261136714,
      "grad_norm": 4.86983060836792,
      "kl": 0.3701171875,
      "learning_rate": 5.738461538461539e-07,
      "loss": 0.0612,
      "reward": 0.7541717886924744,
      "reward_std": 0.06221219711005688,
      "reward_std_all": 0.05965655855834484,
      "rewards-max/forensics_sinkhorn_reward": 0.9446286261081696,
      "rewards-max/forensics_strict_edge_reward": 0.26864299550652504,
      "rewards-max/forensics_tiou_reward": 0.8901672065258026,
      "rewards/forensics_sinkhorn_reward": 0.6841333508491516,
      "rewards/forensics_strict_edge_reward": 0.01828695274889469,
      "rewards/forensics_tiou_reward": 0.44428136944770813,
      "step": 555
    },
    {
      "completion_length": 21.0625,
      "epoch": 1.706605222734255,
      "grad_norm": 6.287864685058594,
      "kl": 0.361328125,
      "learning_rate": 5.73076923076923e-07,
      "loss": 0.0994,
      "reward": 0.7275453209877014,
      "reward_std": 0.09423235431313515,
      "reward_std_all": 0.08952685818076134,
      "rewards-max/forensics_sinkhorn_reward": 0.9672621190547943,
      "rewards-max/forensics_strict_edge_reward": 0.14193252474069595,
      "rewards-max/forensics_tiou_reward": 0.9414741396903992,
      "rewards/forensics_sinkhorn_reward": 0.6987446248531342,
      "rewards/forensics_strict_edge_reward": 0.01847580261528492,
      "rewards/forensics_tiou_reward": 0.5202814489603043,
      "step": 556
    },
    {
      "completion_length": 26.375,
      "epoch": 1.7096774193548387,
      "grad_norm": 4.500039100646973,
      "kl": 0.29931640625,
      "learning_rate": 5.723076923076923e-07,
      "loss": 0.1393,
      "reward": 0.7734534740447998,
      "reward_std": 0.04050430841743946,
      "reward_std_all": 0.03899865597486496,
      "rewards-max/forensics_sinkhorn_reward": 0.943415641784668,
      "rewards-max/forensics_strict_edge_reward": 0.38529306650161743,
      "rewards-max/forensics_tiou_reward": 0.8945984244346619,
      "rewards/forensics_sinkhorn_reward": 0.6549832224845886,
      "rewards/forensics_strict_edge_reward": 0.019127548672258854,
      "rewards/forensics_tiou_reward": 0.46689191460609436,
      "step": 557
    },
    {
      "completion_length": 25.703125,
      "epoch": 1.7127496159754223,
      "grad_norm": 5.962106227874756,
      "kl": 0.34765625,
      "learning_rate": 5.715384615384615e-07,
      "loss": 0.1105,
      "reward": 0.7275654375553131,
      "reward_std": 0.04305928573012352,
      "reward_std_all": 0.06291265785694122,
      "rewards-max/forensics_sinkhorn_reward": 0.9429255127906799,
      "rewards-max/forensics_strict_edge_reward": 0.3504481874406338,
      "rewards-max/forensics_tiou_reward": 0.8885318338871002,
      "rewards/forensics_sinkhorn_reward": 0.6486914455890656,
      "rewards/forensics_strict_edge_reward": 0.021085403859615326,
      "rewards/forensics_tiou_reward": 0.4554806053638458,
      "step": 558
    },
    {
      "completion_length": 24.046875,
      "epoch": 1.715821812596006,
      "grad_norm": 5.820790767669678,
      "kl": 0.3564453125,
      "learning_rate": 5.707692307692307e-07,
      "loss": 0.0265,
      "reward": 0.7734763920307159,
      "reward_std": 0.03311374643817544,
      "reward_std_all": 0.03410703083500266,
      "rewards-max/forensics_sinkhorn_reward": 0.9671066999435425,
      "rewards-max/forensics_strict_edge_reward": 0.27228695154190063,
      "rewards-max/forensics_tiou_reward": 0.9478578865528107,
      "rewards/forensics_sinkhorn_reward": 0.7159439623355865,
      "rewards/forensics_strict_edge_reward": 0.017296348698437214,
      "rewards/forensics_tiou_reward": 0.48679205775260925,
      "step": 559
    },
    {
      "completion_length": 24.8125,
      "epoch": 1.7188940092165899,
      "grad_norm": 6.93126916885376,
      "kl": 0.515625,
      "learning_rate": 5.699999999999999e-07,
      "loss": 0.0688,
      "reward": 0.686175525188446,
      "reward_std": 0.06509839370846748,
      "reward_std_all": 0.08508431911468506,
      "rewards-max/forensics_sinkhorn_reward": 0.9471554160118103,
      "rewards-max/forensics_strict_edge_reward": 0.07131166663020849,
      "rewards-max/forensics_tiou_reward": 0.8888286054134369,
      "rewards/forensics_sinkhorn_reward": 0.5860868394374847,
      "rewards/forensics_strict_edge_reward": 0.0055891345255076885,
      "rewards/forensics_tiou_reward": 0.3928789496421814,
      "step": 560
    },
    {
      "completion_length": 26.0625,
      "epoch": 1.7219662058371736,
      "grad_norm": 6.079333782196045,
      "kl": 0.3203125,
      "learning_rate": 5.692307692307692e-07,
      "loss": 0.0059,
      "reward": 0.7882590591907501,
      "reward_std": 0.028463836759328842,
      "reward_std_all": 0.02682219911366701,
      "rewards-max/forensics_sinkhorn_reward": 0.9687439799308777,
      "rewards-max/forensics_strict_edge_reward": 0.4212051257491112,
      "rewards-max/forensics_tiou_reward": 0.9509531259536743,
      "rewards/forensics_sinkhorn_reward": 0.7420915961265564,
      "rewards/forensics_strict_edge_reward": 0.05128078255802393,
      "rewards/forensics_tiou_reward": 0.5263161659240723,
      "step": 561
    },
    {
      "completion_length": 24.171875,
      "epoch": 1.7250384024577574,
      "grad_norm": 5.328073978424072,
      "kl": 0.416015625,
      "learning_rate": 5.684615384615385e-07,
      "loss": 0.055,
      "reward": 0.7863892316818237,
      "reward_std": 0.0272222263738513,
      "reward_std_all": 0.026523496489971876,
      "rewards-max/forensics_sinkhorn_reward": 0.9701232016086578,
      "rewards-max/forensics_strict_edge_reward": 0.3894139528274536,
      "rewards-max/forensics_tiou_reward": 0.9511651992797852,
      "rewards/forensics_sinkhorn_reward": 0.7103028893470764,
      "rewards/forensics_strict_edge_reward": 0.029925810173153877,
      "rewards/forensics_tiou_reward": 0.5298685282468796,
      "step": 562
    },
    {
      "completion_length": 23.453125,
      "epoch": 1.728110599078341,
      "grad_norm": 6.062849521636963,
      "kl": 0.4228515625,
      "learning_rate": 5.676923076923077e-07,
      "loss": -0.0898,
      "reward": 0.7811692953109741,
      "reward_std": 0.0068745422177016735,
      "reward_std_all": 0.01714281179010868,
      "rewards-max/forensics_sinkhorn_reward": 0.973277747631073,
      "rewards-max/forensics_strict_edge_reward": 0.1894257366657257,
      "rewards-max/forensics_tiou_reward": 0.9624215066432953,
      "rewards/forensics_sinkhorn_reward": 0.7666839063167572,
      "rewards/forensics_strict_edge_reward": 0.02504762541502714,
      "rewards/forensics_tiou_reward": 0.5421878695487976,
      "step": 563
    },
    {
      "completion_length": 25.765625,
      "epoch": 1.7311827956989247,
      "grad_norm": 5.715612888336182,
      "kl": 0.4921875,
      "learning_rate": 5.669230769230769e-07,
      "loss": 0.0271,
      "reward": 0.7608327269554138,
      "reward_std": 0.04959084466099739,
      "reward_std_all": 0.04890480078756809,
      "rewards-max/forensics_sinkhorn_reward": 0.9304582476615906,
      "rewards-max/forensics_strict_edge_reward": 0.3035059869289398,
      "rewards-max/forensics_tiou_reward": 0.8662807941436768,
      "rewards/forensics_sinkhorn_reward": 0.6662992835044861,
      "rewards/forensics_strict_edge_reward": 0.02197258546948433,
      "rewards/forensics_tiou_reward": 0.4509269595146179,
      "step": 564
    },
    {
      "completion_length": 24.0625,
      "epoch": 1.7342549923195083,
      "grad_norm": 6.6031670570373535,
      "kl": 0.37890625,
      "learning_rate": 5.661538461538461e-07,
      "loss": -0.0325,
      "reward": 0.7516432404518127,
      "reward_std": 0.0218961302889511,
      "reward_std_all": 0.03750641900114715,
      "rewards-max/forensics_sinkhorn_reward": 0.9558887183666229,
      "rewards-max/forensics_strict_edge_reward": 0.5852221548557281,
      "rewards-max/forensics_tiou_reward": 0.9165444076061249,
      "rewards/forensics_sinkhorn_reward": 0.6657527387142181,
      "rewards/forensics_strict_edge_reward": 0.041801854968070984,
      "rewards/forensics_tiou_reward": 0.43817150592803955,
      "step": 565
    },
    {
      "completion_length": 26.734375,
      "epoch": 1.737327188940092,
      "grad_norm": 5.70039701461792,
      "kl": 0.3271484375,
      "learning_rate": 5.653846153846153e-07,
      "loss": 0.0656,
      "reward": 0.73942831158638,
      "reward_std": 0.04921055771410465,
      "reward_std_all": 0.05711930897086859,
      "rewards-max/forensics_sinkhorn_reward": 0.9460383951663971,
      "rewards-max/forensics_strict_edge_reward": 0.43554507941007614,
      "rewards-max/forensics_tiou_reward": 0.8985409140586853,
      "rewards/forensics_sinkhorn_reward": 0.6720464527606964,
      "rewards/forensics_strict_edge_reward": 0.028091846266761422,
      "rewards/forensics_tiou_reward": 0.49121618270874023,
      "step": 566
    },
    {
      "completion_length": 23.234375,
      "epoch": 1.7403993855606759,
      "grad_norm": 6.41695499420166,
      "kl": 0.21728515625,
      "learning_rate": 5.646153846153847e-07,
      "loss": 0.0564,
      "reward": 0.7832063436508179,
      "reward_std": 0.024740219581872225,
      "reward_std_all": 0.025386280845850706,
      "rewards-max/forensics_sinkhorn_reward": 0.9801755547523499,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9755765497684479,
      "rewards/forensics_sinkhorn_reward": 0.7383276224136353,
      "rewards/forensics_strict_edge_reward": 0.05836484767496586,
      "rewards/forensics_tiou_reward": 0.5326663255691528,
      "step": 567
    },
    {
      "completion_length": 26.1875,
      "epoch": 1.7434715821812596,
      "grad_norm": 5.382589817047119,
      "kl": 0.33984375,
      "learning_rate": 5.638461538461539e-07,
      "loss": 0.0327,
      "reward": 0.7643202841281891,
      "reward_std": 0.052385153248906136,
      "reward_std_all": 0.04997440939769149,
      "rewards-max/forensics_sinkhorn_reward": 0.9702120721340179,
      "rewards-max/forensics_strict_edge_reward": 0.5049464255571365,
      "rewards-max/forensics_tiou_reward": 0.9538368582725525,
      "rewards/forensics_sinkhorn_reward": 0.7098999917507172,
      "rewards/forensics_strict_edge_reward": 0.03720415197312832,
      "rewards/forensics_tiou_reward": 0.4958711713552475,
      "step": 568
    },
    {
      "completion_length": 27.125,
      "epoch": 1.7465437788018434,
      "grad_norm": 6.072665691375732,
      "kl": 0.318359375,
      "learning_rate": 5.63076923076923e-07,
      "loss": 0.0148,
      "reward": 0.7354025840759277,
      "reward_std": 0.04673360660672188,
      "reward_std_all": 0.05884273163974285,
      "rewards-max/forensics_sinkhorn_reward": 0.9790560305118561,
      "rewards-max/forensics_strict_edge_reward": 0.3226078450679779,
      "rewards-max/forensics_tiou_reward": 0.9649020135402679,
      "rewards/forensics_sinkhorn_reward": 0.6277106702327728,
      "rewards/forensics_strict_edge_reward": 0.03149683494120836,
      "rewards/forensics_tiou_reward": 0.4825698733329773,
      "step": 569
    },
    {
      "completion_length": 22.890625,
      "epoch": 1.7496159754224272,
      "grad_norm": 6.217278480529785,
      "kl": 0.3017578125,
      "learning_rate": 5.623076923076923e-07,
      "loss": 0.0855,
      "reward": 0.7768502533435822,
      "reward_std": 0.04179554618895054,
      "reward_std_all": 0.039186177775263786,
      "rewards-max/forensics_sinkhorn_reward": 0.9772132933139801,
      "rewards-max/forensics_strict_edge_reward": 0.684635654091835,
      "rewards-max/forensics_tiou_reward": 0.9702760875225067,
      "rewards/forensics_sinkhorn_reward": 0.7282100319862366,
      "rewards/forensics_strict_edge_reward": 0.059403304010629654,
      "rewards/forensics_tiou_reward": 0.5585369765758514,
      "step": 570
    },
    {
      "completion_length": 25.21875,
      "epoch": 1.7526881720430108,
      "grad_norm": 5.494638442993164,
      "kl": 0.412109375,
      "learning_rate": 5.615384615384615e-07,
      "loss": 0.1779,
      "reward": 0.7287363111972809,
      "reward_std": 0.05206853151321411,
      "reward_std_all": 0.06629656068980694,
      "rewards-max/forensics_sinkhorn_reward": 0.9832309782505035,
      "rewards-max/forensics_strict_edge_reward": 0.08982638642191887,
      "rewards-max/forensics_tiou_reward": 0.9824842214584351,
      "rewards/forensics_sinkhorn_reward": 0.6408801972866058,
      "rewards/forensics_strict_edge_reward": 0.006178711075335741,
      "rewards/forensics_tiou_reward": 0.4765927791595459,
      "step": 571
    },
    {
      "completion_length": 21.859375,
      "epoch": 1.7557603686635943,
      "grad_norm": 4.9646477699279785,
      "kl": 0.330078125,
      "learning_rate": 5.607692307692308e-07,
      "loss": 0.07,
      "reward": 0.7568888664245605,
      "reward_std": 0.05301351845264435,
      "reward_std_all": 0.05204594321548939,
      "rewards-max/forensics_sinkhorn_reward": 0.973762035369873,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9636716544628143,
      "rewards/forensics_sinkhorn_reward": 0.6931445300579071,
      "rewards/forensics_strict_edge_reward": 0.04555101692676544,
      "rewards/forensics_tiou_reward": 0.5170236527919769,
      "step": 572
    },
    {
      "completion_length": 24.859375,
      "epoch": 1.758832565284178,
      "grad_norm": 6.70456075668335,
      "kl": 0.341796875,
      "learning_rate": 5.6e-07,
      "loss": 0.0068,
      "reward": 0.7567836940288544,
      "reward_std": 0.061553750187158585,
      "reward_std_all": 0.058422431349754333,
      "rewards-max/forensics_sinkhorn_reward": 0.9019446074962616,
      "rewards-max/forensics_strict_edge_reward": 0.3736908221617341,
      "rewards-max/forensics_tiou_reward": 0.8066402673721313,
      "rewards/forensics_sinkhorn_reward": 0.670304924249649,
      "rewards/forensics_strict_edge_reward": 0.04296155359770637,
      "rewards/forensics_tiou_reward": 0.4849057048559189,
      "step": 573
    },
    {
      "completion_length": 23.71875,
      "epoch": 1.7619047619047619,
      "grad_norm": 6.2098236083984375,
      "kl": 0.3251953125,
      "learning_rate": 5.592307692307691e-07,
      "loss": -0.0774,
      "reward": 0.7470859289169312,
      "reward_std": 0.024255639873445034,
      "reward_std_all": 0.042588889598846436,
      "rewards-max/forensics_sinkhorn_reward": 0.972150057554245,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9580370187759399,
      "rewards/forensics_sinkhorn_reward": 0.6797143816947937,
      "rewards/forensics_strict_edge_reward": 0.023176842369139194,
      "rewards/forensics_tiou_reward": 0.5108073651790619,
      "step": 574
    },
    {
      "completion_length": 26.5,
      "epoch": 1.7649769585253456,
      "grad_norm": 5.471656799316406,
      "kl": 0.3349609375,
      "learning_rate": 5.584615384615385e-07,
      "loss": 0.2006,
      "reward": 0.7353164851665497,
      "reward_std": 0.07990609109401703,
      "reward_std_all": 0.07809320092201233,
      "rewards-max/forensics_sinkhorn_reward": 0.9586760103702545,
      "rewards-max/forensics_strict_edge_reward": 0.4648469015955925,
      "rewards-max/forensics_tiou_reward": 0.930749386548996,
      "rewards/forensics_sinkhorn_reward": 0.6334303021430969,
      "rewards/forensics_strict_edge_reward": 0.02839751122519374,
      "rewards/forensics_tiou_reward": 0.4534997045993805,
      "step": 575
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.7680491551459294,
      "grad_norm": 6.3100080490112305,
      "kl": 0.404296875,
      "learning_rate": 5.576923076923077e-07,
      "loss": 0.0844,
      "reward": 0.7262915968894958,
      "reward_std": 0.06462340615689754,
      "reward_std_all": 0.07517238333821297,
      "rewards-max/forensics_sinkhorn_reward": 0.9214827120304108,
      "rewards-max/forensics_strict_edge_reward": 0.13011959474533796,
      "rewards-max/forensics_tiou_reward": 0.8374117314815521,
      "rewards/forensics_sinkhorn_reward": 0.6803960800170898,
      "rewards/forensics_strict_edge_reward": 0.01100035069976002,
      "rewards/forensics_tiou_reward": 0.4877539575099945,
      "step": 576
    },
    {
      "completion_length": 25.546875,
      "epoch": 1.7711213517665132,
      "grad_norm": 6.551080703735352,
      "kl": 0.2822265625,
      "learning_rate": 5.569230769230769e-07,
      "loss": 0.0609,
      "reward": 0.6975935995578766,
      "reward_std": 0.07157494500279427,
      "reward_std_all": 0.0883454754948616,
      "rewards-max/forensics_sinkhorn_reward": 0.9777042269706726,
      "rewards-max/forensics_strict_edge_reward": 0.29145028814673424,
      "rewards-max/forensics_tiou_reward": 0.968435525894165,
      "rewards/forensics_sinkhorn_reward": 0.6218495666980743,
      "rewards/forensics_strict_edge_reward": 0.03044461621902883,
      "rewards/forensics_tiou_reward": 0.45688000321388245,
      "step": 577
    },
    {
      "completion_length": 27.203125,
      "epoch": 1.7741935483870968,
      "grad_norm": 5.904832363128662,
      "kl": 0.36328125,
      "learning_rate": 5.561538461538461e-07,
      "loss": 0.0868,
      "reward": 0.774198442697525,
      "reward_std": 0.05043832492083311,
      "reward_std_all": 0.04633930325508118,
      "rewards-max/forensics_sinkhorn_reward": 0.9714849889278412,
      "rewards-max/forensics_strict_edge_reward": 0.43584437668323517,
      "rewards-max/forensics_tiou_reward": 0.9539453685283661,
      "rewards/forensics_sinkhorn_reward": 0.7217055559158325,
      "rewards/forensics_strict_edge_reward": 0.04955552518367767,
      "rewards/forensics_tiou_reward": 0.5516284704208374,
      "step": 578
    },
    {
      "completion_length": 27.1875,
      "epoch": 1.7772657450076805,
      "grad_norm": 4.408742904663086,
      "kl": 0.4892578125,
      "learning_rate": 5.553846153846153e-07,
      "loss": 0.0819,
      "reward": 0.7769714295864105,
      "reward_std": 0.03477223217487335,
      "reward_std_all": 0.03398778010159731,
      "rewards-max/forensics_sinkhorn_reward": 0.9827292263507843,
      "rewards-max/forensics_strict_edge_reward": 0.6660739481449127,
      "rewards-max/forensics_tiou_reward": 0.9848387241363525,
      "rewards/forensics_sinkhorn_reward": 0.7339848279953003,
      "rewards/forensics_strict_edge_reward": 0.07715191692113876,
      "rewards/forensics_tiou_reward": 0.5953733623027802,
      "step": 579
    },
    {
      "completion_length": 25.21875,
      "epoch": 1.780337941628264,
      "grad_norm": 4.958686828613281,
      "kl": 0.412109375,
      "learning_rate": 5.546153846153846e-07,
      "loss": 0.049,
      "reward": 0.7666658759117126,
      "reward_std": 0.044780682772397995,
      "reward_std_all": 0.04351633973419666,
      "rewards-max/forensics_sinkhorn_reward": 0.983664482831955,
      "rewards-max/forensics_strict_edge_reward": 0.6363584399223328,
      "rewards-max/forensics_tiou_reward": 0.9846864342689514,
      "rewards/forensics_sinkhorn_reward": 0.7001948952674866,
      "rewards/forensics_strict_edge_reward": 0.06041010469198227,
      "rewards/forensics_tiou_reward": 0.5332559049129486,
      "step": 580
    },
    {
      "completion_length": 26.46875,
      "epoch": 1.7834101382488479,
      "grad_norm": 5.825222969055176,
      "kl": 0.572265625,
      "learning_rate": 5.538461538461539e-07,
      "loss": 0.0784,
      "reward": 0.7665116190910339,
      "reward_std": 0.036151159554719925,
      "reward_std_all": 0.03822445869445801,
      "rewards-max/forensics_sinkhorn_reward": 0.9573576152324677,
      "rewards-max/forensics_strict_edge_reward": 0.27862371038645506,
      "rewards-max/forensics_tiou_reward": 0.929855078458786,
      "rewards/forensics_sinkhorn_reward": 0.6996340155601501,
      "rewards/forensics_strict_edge_reward": 0.026133433682844043,
      "rewards/forensics_tiou_reward": 0.4683738052845001,
      "step": 581
    },
    {
      "completion_length": 24.65625,
      "epoch": 1.7864823348694316,
      "grad_norm": 6.693287372589111,
      "kl": 0.6572265625,
      "learning_rate": 5.53076923076923e-07,
      "loss": 0.0923,
      "reward": 0.7334997653961182,
      "reward_std": 0.05398047994822264,
      "reward_std_all": 0.06457136012613773,
      "rewards-max/forensics_sinkhorn_reward": 0.939420074224472,
      "rewards-max/forensics_strict_edge_reward": 0.18765338975936174,
      "rewards-max/forensics_tiou_reward": 0.8782519698143005,
      "rewards/forensics_sinkhorn_reward": 0.6758174896240234,
      "rewards/forensics_strict_edge_reward": 0.015156201756326482,
      "rewards/forensics_tiou_reward": 0.42929302155971527,
      "step": 582
    },
    {
      "completion_length": 24.265625,
      "epoch": 1.7895545314900154,
      "grad_norm": 5.599242687225342,
      "kl": 0.3603515625,
      "learning_rate": 5.523076923076923e-07,
      "loss": 0.0126,
      "reward": 0.7850351631641388,
      "reward_std": 0.013645224506035447,
      "reward_std_all": 0.020023949211463332,
      "rewards-max/forensics_sinkhorn_reward": 0.9777288436889648,
      "rewards-max/forensics_strict_edge_reward": 0.611439049243927,
      "rewards-max/forensics_tiou_reward": 0.9718689620494843,
      "rewards/forensics_sinkhorn_reward": 0.7621504962444305,
      "rewards/forensics_strict_edge_reward": 0.0382945965975523,
      "rewards/forensics_tiou_reward": 0.5171012580394745,
      "step": 583
    },
    {
      "completion_length": 26.1875,
      "epoch": 1.7926267281105992,
      "grad_norm": 5.707552909851074,
      "kl": 0.373046875,
      "learning_rate": 5.515384615384615e-07,
      "loss": -0.0735,
      "reward": 0.7138761878013611,
      "reward_std": 0.06315481662750244,
      "reward_std_all": 0.07593009248375893,
      "rewards-max/forensics_sinkhorn_reward": 0.9852527976036072,
      "rewards-max/forensics_strict_edge_reward": 0.776590883731842,
      "rewards-max/forensics_tiou_reward": 0.9886879920959473,
      "rewards/forensics_sinkhorn_reward": 0.6454919576644897,
      "rewards/forensics_strict_edge_reward": 0.07380173169076443,
      "rewards/forensics_tiou_reward": 0.47932136058807373,
      "step": 584
    },
    {
      "completion_length": 22.515625,
      "epoch": 1.7956989247311828,
      "grad_norm": 5.48043155670166,
      "kl": 0.296875,
      "learning_rate": 5.507692307692308e-07,
      "loss": -0.0739,
      "reward": 0.7686383128166199,
      "reward_std": 0.011181647423654795,
      "reward_std_all": 0.02688213251531124,
      "rewards-max/forensics_sinkhorn_reward": 0.9522521495819092,
      "rewards-max/forensics_strict_edge_reward": 0.4009041115641594,
      "rewards-max/forensics_tiou_reward": 0.9137061834335327,
      "rewards/forensics_sinkhorn_reward": 0.7094075381755829,
      "rewards/forensics_strict_edge_reward": 0.03947724215686321,
      "rewards/forensics_tiou_reward": 0.4654112011194229,
      "step": 585
    },
    {
      "completion_length": 24.484375,
      "epoch": 1.7987711213517665,
      "grad_norm": 5.322304725646973,
      "kl": 0.33984375,
      "learning_rate": 5.5e-07,
      "loss": 0.0459,
      "reward": 0.7769545614719391,
      "reward_std": 0.025658300146460533,
      "reward_std_all": 0.027789155021309853,
      "rewards-max/forensics_sinkhorn_reward": 0.9768470227718353,
      "rewards-max/forensics_strict_edge_reward": 0.36058833822607994,
      "rewards-max/forensics_tiou_reward": 0.9621484875679016,
      "rewards/forensics_sinkhorn_reward": 0.6834836900234222,
      "rewards/forensics_strict_edge_reward": 0.036795132560655475,
      "rewards/forensics_tiou_reward": 0.5342728197574615,
      "step": 586
    },
    {
      "completion_length": 23.40625,
      "epoch": 1.80184331797235,
      "grad_norm": 6.467336177825928,
      "kl": 0.3056640625,
      "learning_rate": 5.492307692307691e-07,
      "loss": 0.084,
      "reward": 0.7255539298057556,
      "reward_std": 0.08682087436318398,
      "reward_std_all": 0.08496689051389694,
      "rewards-max/forensics_sinkhorn_reward": 0.9718778133392334,
      "rewards-max/forensics_strict_edge_reward": 0.36638111621141434,
      "rewards-max/forensics_tiou_reward": 0.9575762450695038,
      "rewards/forensics_sinkhorn_reward": 0.6271680891513824,
      "rewards/forensics_strict_edge_reward": 0.02167621161788702,
      "rewards/forensics_tiou_reward": 0.4032072573900223,
      "step": 587
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.8049155145929339,
      "grad_norm": 5.153688907623291,
      "kl": 0.548828125,
      "learning_rate": 5.484615384615385e-07,
      "loss": -0.0613,
      "reward": 0.7569634914398193,
      "reward_std": 0.017368671484291553,
      "reward_std_all": 0.034716591238975525,
      "rewards-max/forensics_sinkhorn_reward": 0.9824220240116119,
      "rewards-max/forensics_strict_edge_reward": 0.6416448950767517,
      "rewards-max/forensics_tiou_reward": 0.9811535775661469,
      "rewards/forensics_sinkhorn_reward": 0.7187118828296661,
      "rewards/forensics_strict_edge_reward": 0.05299851670861244,
      "rewards/forensics_tiou_reward": 0.547853946685791,
      "step": 588
    },
    {
      "completion_length": 23.40625,
      "epoch": 1.8079877112135176,
      "grad_norm": 6.190130233764648,
      "kl": 0.283203125,
      "learning_rate": 5.476923076923077e-07,
      "loss": 0.0596,
      "reward": 0.7682629525661469,
      "reward_std": 0.04706692323088646,
      "reward_std_all": 0.045183971524238586,
      "rewards-max/forensics_sinkhorn_reward": 0.9688928425312042,
      "rewards-max/forensics_strict_edge_reward": 0.3619650974869728,
      "rewards-max/forensics_tiou_reward": 0.9387322664260864,
      "rewards/forensics_sinkhorn_reward": 0.7178420424461365,
      "rewards/forensics_strict_edge_reward": 0.02430390566587448,
      "rewards/forensics_tiou_reward": 0.5010594427585602,
      "step": 589
    },
    {
      "completion_length": 25.15625,
      "epoch": 1.8110599078341014,
      "grad_norm": 5.206435203552246,
      "kl": 0.3115234375,
      "learning_rate": 5.469230769230769e-07,
      "loss": 0.1249,
      "reward": 0.6891390085220337,
      "reward_std": 0.0978284515440464,
      "reward_std_all": 0.10796844214200974,
      "rewards-max/forensics_sinkhorn_reward": 0.9707871973514557,
      "rewards-max/forensics_strict_edge_reward": 0.6061199605464935,
      "rewards-max/forensics_tiou_reward": 0.9561183452606201,
      "rewards/forensics_sinkhorn_reward": 0.5933756232261658,
      "rewards/forensics_strict_edge_reward": 0.04606015980243683,
      "rewards/forensics_tiou_reward": 0.46963314712047577,
      "step": 590
    },
    {
      "completion_length": 23.9375,
      "epoch": 1.8141321044546852,
      "grad_norm": 6.023600101470947,
      "kl": 0.306640625,
      "learning_rate": 5.461538461538461e-07,
      "loss": -0.0449,
      "reward": 0.7671158909797668,
      "reward_std": 0.04137599840760231,
      "reward_std_all": 0.04124346002936363,
      "rewards-max/forensics_sinkhorn_reward": 0.967787891626358,
      "rewards-max/forensics_strict_edge_reward": 0.38249877095222473,
      "rewards-max/forensics_tiou_reward": 0.9375229179859161,
      "rewards/forensics_sinkhorn_reward": 0.7090516686439514,
      "rewards/forensics_strict_edge_reward": 0.027340351603925228,
      "rewards/forensics_tiou_reward": 0.44890616834163666,
      "step": 591
    },
    {
      "completion_length": 25.359375,
      "epoch": 1.817204301075269,
      "grad_norm": 5.032681941986084,
      "kl": 0.2861328125,
      "learning_rate": 5.453846153846153e-07,
      "loss": 0.0538,
      "reward": 0.7281026244163513,
      "reward_std": 0.06858893856406212,
      "reward_std_all": 0.07151643931865692,
      "rewards-max/forensics_sinkhorn_reward": 0.979680210351944,
      "rewards-max/forensics_strict_edge_reward": 0.5676676407456398,
      "rewards-max/forensics_tiou_reward": 0.9751456379890442,
      "rewards/forensics_sinkhorn_reward": 0.6687644720077515,
      "rewards/forensics_strict_edge_reward": 0.029085648246109486,
      "rewards/forensics_tiou_reward": 0.46714767813682556,
      "step": 592
    },
    {
      "completion_length": 25.359375,
      "epoch": 1.8202764976958525,
      "grad_norm": 4.303493499755859,
      "kl": 0.310546875,
      "learning_rate": 5.446153846153846e-07,
      "loss": -0.0441,
      "reward": 0.7335149347782135,
      "reward_std": 0.013539372012019157,
      "reward_std_all": 0.04415895789861679,
      "rewards-max/forensics_sinkhorn_reward": 0.9765210449695587,
      "rewards-max/forensics_strict_edge_reward": 0.5389915108680725,
      "rewards-max/forensics_tiou_reward": 0.9660646915435791,
      "rewards/forensics_sinkhorn_reward": 0.7026078701019287,
      "rewards/forensics_strict_edge_reward": 0.0477543119341135,
      "rewards/forensics_tiou_reward": 0.5164405107498169,
      "step": 593
    },
    {
      "completion_length": 24.625,
      "epoch": 1.823348694316436,
      "grad_norm": 4.535218715667725,
      "kl": 0.310546875,
      "learning_rate": 5.438461538461538e-07,
      "loss": 0.0772,
      "reward": 0.7716026306152344,
      "reward_std": 0.043570347130298615,
      "reward_std_all": 0.04145362786948681,
      "rewards-max/forensics_sinkhorn_reward": 0.9831089377403259,
      "rewards-max/forensics_strict_edge_reward": 0.6082770228385925,
      "rewards-max/forensics_tiou_reward": 0.9838946163654327,
      "rewards/forensics_sinkhorn_reward": 0.7289271354675293,
      "rewards/forensics_strict_edge_reward": 0.05444402992725372,
      "rewards/forensics_tiou_reward": 0.5524330139160156,
      "step": 594
    },
    {
      "completion_length": 22.921875,
      "epoch": 1.8264208909370199,
      "grad_norm": 6.755789756774902,
      "kl": 0.3564453125,
      "learning_rate": 5.430769230769231e-07,
      "loss": 0.1164,
      "reward": 0.7531170845031738,
      "reward_std": 0.03928151912987232,
      "reward_std_all": 0.04768352396786213,
      "rewards-max/forensics_sinkhorn_reward": 0.9485141336917877,
      "rewards-max/forensics_strict_edge_reward": 0.3340276926755905,
      "rewards-max/forensics_tiou_reward": 0.903405636548996,
      "rewards/forensics_sinkhorn_reward": 0.6360431909561157,
      "rewards/forensics_strict_edge_reward": 0.02199701778590679,
      "rewards/forensics_tiou_reward": 0.4201129227876663,
      "step": 595
    },
    {
      "completion_length": 24.65625,
      "epoch": 1.8294930875576036,
      "grad_norm": 6.398003578186035,
      "kl": 0.27587890625,
      "learning_rate": 5.423076923076923e-07,
      "loss": 0.0332,
      "reward": 0.755843847990036,
      "reward_std": 0.06611545197665691,
      "reward_std_all": 0.06172485277056694,
      "rewards-max/forensics_sinkhorn_reward": 0.9674757122993469,
      "rewards-max/forensics_strict_edge_reward": 0.5071525871753693,
      "rewards-max/forensics_tiou_reward": 0.9371891617774963,
      "rewards/forensics_sinkhorn_reward": 0.7611290216445923,
      "rewards/forensics_strict_edge_reward": 0.07066921889781952,
      "rewards/forensics_tiou_reward": 0.6045598089694977,
      "step": 596
    },
    {
      "completion_length": 26.6875,
      "epoch": 1.8325652841781874,
      "grad_norm": 5.701013088226318,
      "kl": 0.341796875,
      "learning_rate": 5.415384615384615e-07,
      "loss": 0.0804,
      "reward": 0.6745545268058777,
      "reward_std": 0.08443359658122063,
      "reward_std_all": 0.10397547483444214,
      "rewards-max/forensics_sinkhorn_reward": 0.9783920347690582,
      "rewards-max/forensics_strict_edge_reward": 0.5226984620094299,
      "rewards-max/forensics_tiou_reward": 0.9689276218414307,
      "rewards/forensics_sinkhorn_reward": 0.5844375193119049,
      "rewards/forensics_strict_edge_reward": 0.03746808134019375,
      "rewards/forensics_tiou_reward": 0.3846752941608429,
      "step": 597
    },
    {
      "completion_length": 27.71875,
      "epoch": 1.8356374807987712,
      "grad_norm": 4.381137371063232,
      "kl": 0.2890625,
      "learning_rate": 5.407692307692308e-07,
      "loss": 0.0735,
      "reward": 0.772357314825058,
      "reward_std": 0.0294899451546371,
      "reward_std_all": 0.03172401525080204,
      "rewards-max/forensics_sinkhorn_reward": 0.9740769267082214,
      "rewards-max/forensics_strict_edge_reward": 0.7457418143749237,
      "rewards-max/forensics_tiou_reward": 0.96659255027771,
      "rewards/forensics_sinkhorn_reward": 0.750138133764267,
      "rewards/forensics_strict_edge_reward": 0.05937489494681358,
      "rewards/forensics_tiou_reward": 0.5895671248435974,
      "step": 598
    },
    {
      "completion_length": 23.3125,
      "epoch": 1.838709677419355,
      "grad_norm": 7.127796173095703,
      "kl": 0.3017578125,
      "learning_rate": 5.4e-07,
      "loss": 0.0885,
      "reward": 0.7723469138145447,
      "reward_std": 0.023747306549921632,
      "reward_std_all": 0.031096328049898148,
      "rewards-max/forensics_sinkhorn_reward": 0.9623991549015045,
      "rewards-max/forensics_strict_edge_reward": 0.48020514845848083,
      "rewards-max/forensics_tiou_reward": 0.9205242097377777,
      "rewards/forensics_sinkhorn_reward": 0.7123187482357025,
      "rewards/forensics_strict_edge_reward": 0.032961214892566204,
      "rewards/forensics_tiou_reward": 0.49852797389030457,
      "step": 599
    },
    {
      "completion_length": 27.0,
      "epoch": 1.8417818740399385,
      "grad_norm": 5.912111282348633,
      "kl": 0.4130859375,
      "learning_rate": 5.392307692307691e-07,
      "loss": -0.0577,
      "reward": 0.7366248071193695,
      "reward_std": 0.055782223120331764,
      "reward_std_all": 0.06088607385754585,
      "rewards-max/forensics_sinkhorn_reward": 0.9644031226634979,
      "rewards-max/forensics_strict_edge_reward": 0.35114636272192,
      "rewards-max/forensics_tiou_reward": 0.9389133453369141,
      "rewards/forensics_sinkhorn_reward": 0.6759244501590729,
      "rewards/forensics_strict_edge_reward": 0.02613041689619422,
      "rewards/forensics_tiou_reward": 0.46398937702178955,
      "step": 600
    },
    {
      "completion_length": 24.03125,
      "epoch": 1.8448540706605223,
      "grad_norm": 4.635590076446533,
      "kl": 0.3505859375,
      "learning_rate": 5.384615384615384e-07,
      "loss": 0.0012,
      "reward": 0.7121998965740204,
      "reward_std": 0.06955519877374172,
      "reward_std_all": 0.08013452589511871,
      "rewards-max/forensics_sinkhorn_reward": 0.9402857422828674,
      "rewards-max/forensics_strict_edge_reward": 0.388295441865921,
      "rewards-max/forensics_tiou_reward": 0.8800345957279205,
      "rewards/forensics_sinkhorn_reward": 0.6591179668903351,
      "rewards/forensics_strict_edge_reward": 0.01960371620953083,
      "rewards/forensics_tiou_reward": 0.48944829404354095,
      "step": 601
    },
    {
      "completion_length": 26.796875,
      "epoch": 1.8479262672811059,
      "grad_norm": 6.069166660308838,
      "kl": 0.3681640625,
      "learning_rate": 5.376923076923077e-07,
      "loss": 0.0464,
      "reward": 0.7837872505187988,
      "reward_std": 0.02363153174519539,
      "reward_std_all": 0.024388063699007034,
      "rewards-max/forensics_sinkhorn_reward": 0.9775602221488953,
      "rewards-max/forensics_strict_edge_reward": 0.45414966344833374,
      "rewards-max/forensics_tiou_reward": 0.9689128994941711,
      "rewards/forensics_sinkhorn_reward": 0.7158701121807098,
      "rewards/forensics_strict_edge_reward": 0.026029760017991066,
      "rewards/forensics_tiou_reward": 0.46452100574970245,
      "step": 602
    },
    {
      "completion_length": 24.46875,
      "epoch": 1.8509984639016897,
      "grad_norm": 5.194303512573242,
      "kl": 0.3623046875,
      "learning_rate": 5.369230769230769e-07,
      "loss": 0.1419,
      "reward": 0.7566713690757751,
      "reward_std": 0.04610176011919975,
      "reward_std_all": 0.04844668321311474,
      "rewards-max/forensics_sinkhorn_reward": 0.8963572382926941,
      "rewards-max/forensics_strict_edge_reward": 0.195231594145298,
      "rewards-max/forensics_tiou_reward": 0.7824496924877167,
      "rewards/forensics_sinkhorn_reward": 0.6385655999183655,
      "rewards/forensics_strict_edge_reward": 0.008534164284355938,
      "rewards/forensics_tiou_reward": 0.4109325110912323,
      "step": 603
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.8540706605222734,
      "grad_norm": 6.864537715911865,
      "kl": 0.3359375,
      "learning_rate": 5.361538461538462e-07,
      "loss": 0.0466,
      "reward": 0.7601584792137146,
      "reward_std": 0.038964045234024525,
      "reward_std_all": 0.04247613623738289,
      "rewards-max/forensics_sinkhorn_reward": 0.9798062443733215,
      "rewards-max/forensics_strict_edge_reward": 0.5553485602140427,
      "rewards-max/forensics_tiou_reward": 0.977484792470932,
      "rewards/forensics_sinkhorn_reward": 0.6730460226535797,
      "rewards/forensics_strict_edge_reward": 0.03022720292210579,
      "rewards/forensics_tiou_reward": 0.46570947766304016,
      "step": 604
    },
    {
      "completion_length": 24.8125,
      "epoch": 1.8571428571428572,
      "grad_norm": 6.372686386108398,
      "kl": 0.44140625,
      "learning_rate": 5.353846153846153e-07,
      "loss": -0.0185,
      "reward": 0.7793271839618683,
      "reward_std": 0.019520874600857496,
      "reward_std_all": 0.024183155968785286,
      "rewards-max/forensics_sinkhorn_reward": 0.968726396560669,
      "rewards-max/forensics_strict_edge_reward": 0.4663326218724251,
      "rewards-max/forensics_tiou_reward": 0.9493678212165833,
      "rewards/forensics_sinkhorn_reward": 0.708999752998352,
      "rewards/forensics_strict_edge_reward": 0.0295755488332361,
      "rewards/forensics_tiou_reward": 0.4633172005414963,
      "step": 605
    },
    {
      "completion_length": 23.546875,
      "epoch": 1.860215053763441,
      "grad_norm": 6.035883903503418,
      "kl": 0.373046875,
      "learning_rate": 5.346153846153846e-07,
      "loss": 0.0489,
      "reward": 0.6865594685077667,
      "reward_std": 0.08985811471939087,
      "reward_std_all": 0.10894635692238808,
      "rewards-max/forensics_sinkhorn_reward": 0.9624957740306854,
      "rewards-max/forensics_strict_edge_reward": 0.556238055229187,
      "rewards-max/forensics_tiou_reward": 0.9380442500114441,
      "rewards/forensics_sinkhorn_reward": 0.6393338739871979,
      "rewards/forensics_strict_edge_reward": 0.03824541624635458,
      "rewards/forensics_tiou_reward": 0.4798298180103302,
      "step": 606
    },
    {
      "completion_length": 24.40625,
      "epoch": 1.8632872503840245,
      "grad_norm": 5.3167314529418945,
      "kl": 0.357421875,
      "learning_rate": 5.338461538461538e-07,
      "loss": 0.0154,
      "reward": 0.7600616216659546,
      "reward_std": 0.04193333350121975,
      "reward_std_all": 0.04395700804889202,
      "rewards-max/forensics_sinkhorn_reward": 0.9627819657325745,
      "rewards-max/forensics_strict_edge_reward": 0.4043564051389694,
      "rewards-max/forensics_tiou_reward": 0.9316298365592957,
      "rewards/forensics_sinkhorn_reward": 0.6976471543312073,
      "rewards/forensics_strict_edge_reward": 0.03880543075501919,
      "rewards/forensics_tiou_reward": 0.46063660085201263,
      "step": 607
    },
    {
      "completion_length": 22.015625,
      "epoch": 1.8663594470046083,
      "grad_norm": 5.844497203826904,
      "kl": 0.3359375,
      "learning_rate": 5.33076923076923e-07,
      "loss": 0.0096,
      "reward": 0.7253571152687073,
      "reward_std": 0.06558633595705032,
      "reward_std_all": 0.07330674305558205,
      "rewards-max/forensics_sinkhorn_reward": 0.955572634935379,
      "rewards-max/forensics_strict_edge_reward": 0.37694883346557617,
      "rewards-max/forensics_tiou_reward": 0.918354868888855,
      "rewards/forensics_sinkhorn_reward": 0.6667578220367432,
      "rewards/forensics_strict_edge_reward": 0.04354915674775839,
      "rewards/forensics_tiou_reward": 0.4703223407268524,
      "step": 608
    },
    {
      "completion_length": 25.78125,
      "epoch": 1.8694316436251919,
      "grad_norm": 6.336006164550781,
      "kl": 0.4541015625,
      "learning_rate": 5.323076923076923e-07,
      "loss": 0.1135,
      "reward": 0.7226261496543884,
      "reward_std": 0.06425093114376068,
      "reward_std_all": 0.07616548798978329,
      "rewards-max/forensics_sinkhorn_reward": 0.9811908006668091,
      "rewards-max/forensics_strict_edge_reward": 0.594437800347805,
      "rewards-max/forensics_tiou_reward": 0.9819591641426086,
      "rewards/forensics_sinkhorn_reward": 0.638291209936142,
      "rewards/forensics_strict_edge_reward": 0.030577962286770344,
      "rewards/forensics_tiou_reward": 0.4421689808368683,
      "step": 609
    },
    {
      "completion_length": 24.9375,
      "epoch": 1.8725038402457757,
      "grad_norm": 5.663148880004883,
      "kl": 0.36328125,
      "learning_rate": 5.315384615384615e-07,
      "loss": 0.0048,
      "reward": 0.77577343583107,
      "reward_std": 0.03281582146883011,
      "reward_std_all": 0.03304901486262679,
      "rewards-max/forensics_sinkhorn_reward": 0.9791020750999451,
      "rewards-max/forensics_strict_edge_reward": 0.6398434042930603,
      "rewards-max/forensics_tiou_reward": 0.9739238321781158,
      "rewards/forensics_sinkhorn_reward": 0.7830994129180908,
      "rewards/forensics_strict_edge_reward": 0.04765457473695278,
      "rewards/forensics_tiou_reward": 0.6403902769088745,
      "step": 610
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.8755760368663594,
      "grad_norm": 5.322821140289307,
      "kl": 0.333984375,
      "learning_rate": 5.307692307692308e-07,
      "loss": 0.0549,
      "reward": 0.7429405748844147,
      "reward_std": 0.06752989813685417,
      "reward_std_all": 0.06709489226341248,
      "rewards-max/forensics_sinkhorn_reward": 0.976709634065628,
      "rewards-max/forensics_strict_edge_reward": 0.2573440223932266,
      "rewards-max/forensics_tiou_reward": 0.9640927314758301,
      "rewards/forensics_sinkhorn_reward": 0.6701858043670654,
      "rewards/forensics_strict_edge_reward": 0.016363886184990406,
      "rewards/forensics_tiou_reward": 0.4866492450237274,
      "step": 611
    },
    {
      "completion_length": 26.140625,
      "epoch": 1.8786482334869432,
      "grad_norm": 4.9992194175720215,
      "kl": 0.3427734375,
      "learning_rate": 5.3e-07,
      "loss": 0.0209,
      "reward": 0.7726648151874542,
      "reward_std": 0.054852189496159554,
      "reward_std_all": 0.04991690814495087,
      "rewards-max/forensics_sinkhorn_reward": 0.9310436546802521,
      "rewards-max/forensics_strict_edge_reward": 0.36084495298564434,
      "rewards-max/forensics_tiou_reward": 0.8546944558620453,
      "rewards/forensics_sinkhorn_reward": 0.7049553990364075,
      "rewards/forensics_strict_edge_reward": 0.018321439623832703,
      "rewards/forensics_tiou_reward": 0.5281513929367065,
      "step": 612
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.881720430107527,
      "grad_norm": 6.760980606079102,
      "kl": 0.39453125,
      "learning_rate": 5.292307692307692e-07,
      "loss": 0.0175,
      "reward": 0.7655564546585083,
      "reward_std": 0.05542369931936264,
      "reward_std_all": 0.0521415863186121,
      "rewards-max/forensics_sinkhorn_reward": 0.9530943036079407,
      "rewards-max/forensics_strict_edge_reward": 0.24503491073846817,
      "rewards-max/forensics_tiou_reward": 0.901502251625061,
      "rewards/forensics_sinkhorn_reward": 0.7187425494194031,
      "rewards/forensics_strict_edge_reward": 0.019084544852375984,
      "rewards/forensics_tiou_reward": 0.5269640982151031,
      "step": 613
    },
    {
      "completion_length": 22.71875,
      "epoch": 1.8847926267281108,
      "grad_norm": 7.463662624359131,
      "kl": 0.341796875,
      "learning_rate": 5.284615384615384e-07,
      "loss": 0.0663,
      "reward": 0.733159065246582,
      "reward_std": 0.0546959787607193,
      "reward_std_all": 0.06696350127458572,
      "rewards-max/forensics_sinkhorn_reward": 0.9569305777549744,
      "rewards-max/forensics_strict_edge_reward": 0.2694893293082714,
      "rewards-max/forensics_tiou_reward": 0.9170509874820709,
      "rewards/forensics_sinkhorn_reward": 0.6608225703239441,
      "rewards/forensics_strict_edge_reward": 0.05213048530276865,
      "rewards/forensics_tiou_reward": 0.48046815395355225,
      "step": 614
    },
    {
      "completion_length": 25.5625,
      "epoch": 1.8878648233486943,
      "grad_norm": 5.454308032989502,
      "kl": 0.3681640625,
      "learning_rate": 5.276923076923076e-07,
      "loss": 0.1722,
      "reward": 0.6855090260505676,
      "reward_std": 0.05911200866103172,
      "reward_std_all": 0.089480210095644,
      "rewards-max/forensics_sinkhorn_reward": 0.9799886643886566,
      "rewards-max/forensics_strict_edge_reward": 0.3487136736512184,
      "rewards-max/forensics_tiou_reward": 0.9796157479286194,
      "rewards/forensics_sinkhorn_reward": 0.5633345544338226,
      "rewards/forensics_strict_edge_reward": 0.03414586931467056,
      "rewards/forensics_tiou_reward": 0.43178893625736237,
      "step": 615
    },
    {
      "completion_length": 22.328125,
      "epoch": 1.8909370199692779,
      "grad_norm": 6.193352699279785,
      "kl": 0.357421875,
      "learning_rate": 5.269230769230769e-07,
      "loss": 0.0299,
      "reward": 0.7393682599067688,
      "reward_std": 0.04569630231708288,
      "reward_std_all": 0.05813905969262123,
      "rewards-max/forensics_sinkhorn_reward": 0.969221442937851,
      "rewards-max/forensics_strict_edge_reward": 0.4346180260181427,
      "rewards-max/forensics_tiou_reward": 0.938929408788681,
      "rewards/forensics_sinkhorn_reward": 0.6989019811153412,
      "rewards/forensics_strict_edge_reward": 0.02850915677845478,
      "rewards/forensics_tiou_reward": 0.4614500552415848,
      "step": 616
    },
    {
      "completion_length": 23.734375,
      "epoch": 1.8940092165898617,
      "grad_norm": 5.87831974029541,
      "kl": 0.4287109375,
      "learning_rate": 5.261538461538462e-07,
      "loss": 0.1816,
      "reward": 0.7299697995185852,
      "reward_std": 0.0635837223380804,
      "reward_std_all": 0.07082163169980049,
      "rewards-max/forensics_sinkhorn_reward": 0.9714240729808807,
      "rewards-max/forensics_strict_edge_reward": 0.25160736590623856,
      "rewards-max/forensics_tiou_reward": 0.9535458385944366,
      "rewards/forensics_sinkhorn_reward": 0.6551247835159302,
      "rewards/forensics_strict_edge_reward": 0.023587927222251892,
      "rewards/forensics_tiou_reward": 0.4514537900686264,
      "step": 617
    },
    {
      "completion_length": 22.265625,
      "epoch": 1.8970814132104454,
      "grad_norm": 4.8492536544799805,
      "kl": 0.373046875,
      "learning_rate": 5.253846153846153e-07,
      "loss": 0.0107,
      "reward": 0.775368720293045,
      "reward_std": 0.043156778905540705,
      "reward_std_all": 0.04088739678263664,
      "rewards-max/forensics_sinkhorn_reward": 0.9872221946716309,
      "rewards-max/forensics_strict_edge_reward": 0.5569239109754562,
      "rewards-max/forensics_tiou_reward": 0.9948059320449829,
      "rewards/forensics_sinkhorn_reward": 0.76315838098526,
      "rewards/forensics_strict_edge_reward": 0.039378732442855835,
      "rewards/forensics_tiou_reward": 0.5680136382579803,
      "step": 618
    },
    {
      "completion_length": 24.1875,
      "epoch": 1.9001536098310292,
      "grad_norm": 5.5016255378723145,
      "kl": 0.431640625,
      "learning_rate": 5.246153846153846e-07,
      "loss": 0.1046,
      "reward": 0.7860445976257324,
      "reward_std": 0.02274798136204481,
      "reward_std_all": 0.023521140683442354,
      "rewards-max/forensics_sinkhorn_reward": 0.9776131212711334,
      "rewards-max/forensics_strict_edge_reward": 0.24543391168117523,
      "rewards-max/forensics_tiou_reward": 0.96688511967659,
      "rewards/forensics_sinkhorn_reward": 0.7373983561992645,
      "rewards/forensics_strict_edge_reward": 0.014307274017482996,
      "rewards/forensics_tiou_reward": 0.5041756331920624,
      "step": 619
    },
    {
      "completion_length": 20.984375,
      "epoch": 1.903225806451613,
      "grad_norm": 5.412417888641357,
      "kl": 0.314453125,
      "learning_rate": 5.238461538461538e-07,
      "loss": -0.0964,
      "reward": 0.7830935716629028,
      "reward_std": 0.02392993913963437,
      "reward_std_all": 0.024909948464483023,
      "rewards-max/forensics_sinkhorn_reward": 0.9638634026050568,
      "rewards-max/forensics_strict_edge_reward": 0.5618371963500977,
      "rewards-max/forensics_tiou_reward": 0.9293356835842133,
      "rewards/forensics_sinkhorn_reward": 0.759316474199295,
      "rewards/forensics_strict_edge_reward": 0.04004437290132046,
      "rewards/forensics_tiou_reward": 0.5110891759395599,
      "step": 620
    },
    {
      "completion_length": 22.546875,
      "epoch": 1.9062980030721968,
      "grad_norm": 5.393112659454346,
      "kl": 0.3427734375,
      "learning_rate": 5.23076923076923e-07,
      "loss": 0.0708,
      "reward": 0.7147315442562103,
      "reward_std": 0.056505266577005386,
      "reward_std_all": 0.07352430745959282,
      "rewards-max/forensics_sinkhorn_reward": 0.9575136303901672,
      "rewards-max/forensics_strict_edge_reward": 0.3018318712711334,
      "rewards-max/forensics_tiou_reward": 0.9179708659648895,
      "rewards/forensics_sinkhorn_reward": 0.6526797711849213,
      "rewards/forensics_strict_edge_reward": 0.03349191788583994,
      "rewards/forensics_tiou_reward": 0.46126100420951843,
      "step": 621
    },
    {
      "completion_length": 26.59375,
      "epoch": 1.9093701996927803,
      "grad_norm": 7.306897163391113,
      "kl": 0.42578125,
      "learning_rate": 5.223076923076924e-07,
      "loss": -0.0616,
      "reward": 0.7655181586742401,
      "reward_std": 0.021014095284044743,
      "reward_std_all": 0.03311929851770401,
      "rewards-max/forensics_sinkhorn_reward": 0.9742676317691803,
      "rewards-max/forensics_strict_edge_reward": 0.3550187572836876,
      "rewards-max/forensics_tiou_reward": 0.9605253636837006,
      "rewards/forensics_sinkhorn_reward": 0.7412566244602203,
      "rewards/forensics_strict_edge_reward": 0.031646134331822395,
      "rewards/forensics_tiou_reward": 0.5466812252998352,
      "step": 622
    },
    {
      "completion_length": 24.046875,
      "epoch": 1.912442396313364,
      "grad_norm": 6.044737815856934,
      "kl": 0.5146484375,
      "learning_rate": 5.215384615384615e-07,
      "loss": 0.0507,
      "reward": 0.7471510767936707,
      "reward_std": 0.07058626972138882,
      "reward_std_all": 0.06774167902767658,
      "rewards-max/forensics_sinkhorn_reward": 0.9215418100357056,
      "rewards-max/forensics_strict_edge_reward": 0.14174341037869453,
      "rewards-max/forensics_tiou_reward": 0.8373045325279236,
      "rewards/forensics_sinkhorn_reward": 0.6481623649597168,
      "rewards/forensics_strict_edge_reward": 0.008766304468736053,
      "rewards/forensics_tiou_reward": 0.4314611852169037,
      "step": 623
    },
    {
      "completion_length": 23.953125,
      "epoch": 1.9155145929339477,
      "grad_norm": 5.812639236450195,
      "kl": 0.5869140625,
      "learning_rate": 5.207692307692308e-07,
      "loss": -0.0252,
      "reward": 0.7497419714927673,
      "reward_std": 0.022559751756489277,
      "reward_std_all": 0.0377285722643137,
      "rewards-max/forensics_sinkhorn_reward": 0.9710647165775299,
      "rewards-max/forensics_strict_edge_reward": 0.38056008517742157,
      "rewards-max/forensics_tiou_reward": 0.9544687867164612,
      "rewards/forensics_sinkhorn_reward": 0.7248103618621826,
      "rewards/forensics_strict_edge_reward": 0.030787145718932152,
      "rewards/forensics_tiou_reward": 0.5198446214199066,
      "step": 624
    },
    {
      "completion_length": 23.234375,
      "epoch": 1.9185867895545314,
      "grad_norm": 5.424780368804932,
      "kl": 0.3935546875,
      "learning_rate": 5.2e-07,
      "loss": 0.0036,
      "reward": 0.743678480386734,
      "reward_std": 0.02976570464670658,
      "reward_std_all": 0.04726348724216223,
      "rewards-max/forensics_sinkhorn_reward": 0.9812994003295898,
      "rewards-max/forensics_strict_edge_reward": 0.6271176636219025,
      "rewards-max/forensics_tiou_reward": 0.97939732670784,
      "rewards/forensics_sinkhorn_reward": 0.6954173445701599,
      "rewards/forensics_strict_edge_reward": 0.07531203329563141,
      "rewards/forensics_tiou_reward": 0.48456570506095886,
      "step": 625
    },
    {
      "completion_length": 22.46875,
      "epoch": 1.9216589861751152,
      "grad_norm": 6.252349376678467,
      "kl": 0.3388671875,
      "learning_rate": 5.192307692307692e-07,
      "loss": 0.0085,
      "reward": 0.7606673240661621,
      "reward_std": 0.036107310093939304,
      "reward_std_all": 0.044157618656754494,
      "rewards-max/forensics_sinkhorn_reward": 0.9829190671443939,
      "rewards-max/forensics_strict_edge_reward": 0.670350581407547,
      "rewards-max/forensics_tiou_reward": 0.9825224876403809,
      "rewards/forensics_sinkhorn_reward": 0.7076754570007324,
      "rewards/forensics_strict_edge_reward": 0.0497251320630312,
      "rewards/forensics_tiou_reward": 0.4610886871814728,
      "step": 626
    },
    {
      "completion_length": 24.171875,
      "epoch": 1.924731182795699,
      "grad_norm": 5.729938983917236,
      "kl": 0.4169921875,
      "learning_rate": 5.184615384615384e-07,
      "loss": 0.1053,
      "reward": 0.7436459362506866,
      "reward_std": 0.07526424899697304,
      "reward_std_all": 0.07135133817791939,
      "rewards-max/forensics_sinkhorn_reward": 0.9639497399330139,
      "rewards-max/forensics_strict_edge_reward": 0.36790603399276733,
      "rewards-max/forensics_tiou_reward": 0.938362330198288,
      "rewards/forensics_sinkhorn_reward": 0.6797934174537659,
      "rewards/forensics_strict_edge_reward": 0.03171293064951897,
      "rewards/forensics_tiou_reward": 0.47684675455093384,
      "step": 627
    },
    {
      "completion_length": 23.265625,
      "epoch": 1.9278033794162828,
      "grad_norm": 5.908327579498291,
      "kl": 0.4306640625,
      "learning_rate": 5.176923076923076e-07,
      "loss": 0.0465,
      "reward": 0.7782111763954163,
      "reward_std": 0.0403341855853796,
      "reward_std_all": 0.03781651519238949,
      "rewards-max/forensics_sinkhorn_reward": 0.9800806641578674,
      "rewards-max/forensics_strict_edge_reward": 0.16989751905202866,
      "rewards-max/forensics_tiou_reward": 0.9768811166286469,
      "rewards/forensics_sinkhorn_reward": 0.7354428172111511,
      "rewards/forensics_strict_edge_reward": 0.018684151582419872,
      "rewards/forensics_tiou_reward": 0.5619697272777557,
      "step": 628
    },
    {
      "completion_length": 21.6875,
      "epoch": 1.9308755760368663,
      "grad_norm": 7.992784023284912,
      "kl": 0.484375,
      "learning_rate": 5.16923076923077e-07,
      "loss": -0.027,
      "reward": 0.7446722388267517,
      "reward_std": 0.038996852934360504,
      "reward_std_all": 0.04979780130088329,
      "rewards-max/forensics_sinkhorn_reward": 0.9841029942035675,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9863228797912598,
      "rewards/forensics_sinkhorn_reward": 0.734203577041626,
      "rewards/forensics_strict_edge_reward": 0.062116000801324844,
      "rewards/forensics_tiou_reward": 0.5499524772167206,
      "step": 629
    },
    {
      "completion_length": 23.703125,
      "epoch": 1.93394777265745,
      "grad_norm": 5.7499542236328125,
      "kl": 0.375,
      "learning_rate": 5.161538461538462e-07,
      "loss": -0.0739,
      "reward": 0.7547685503959656,
      "reward_std": 0.03614088334143162,
      "reward_std_all": 0.04607540927827358,
      "rewards-max/forensics_sinkhorn_reward": 0.9715285003185272,
      "rewards-max/forensics_strict_edge_reward": 0.31773683428764343,
      "rewards-max/forensics_tiou_reward": 0.9529917538166046,
      "rewards/forensics_sinkhorn_reward": 0.6993516385555267,
      "rewards/forensics_strict_edge_reward": 0.018826759420335293,
      "rewards/forensics_tiou_reward": 0.47572270035743713,
      "step": 630
    },
    {
      "completion_length": 23.96875,
      "epoch": 1.9370199692780337,
      "grad_norm": 6.366598129272461,
      "kl": 0.357421875,
      "learning_rate": 5.153846153846153e-07,
      "loss": 0.0005,
      "reward": 0.6598859429359436,
      "reward_std": 0.05200802534818649,
      "reward_std_all": 0.09289185702800751,
      "rewards-max/forensics_sinkhorn_reward": 0.9447126686573029,
      "rewards-max/forensics_strict_edge_reward": 0.19332606927491724,
      "rewards-max/forensics_tiou_reward": 0.873431384563446,
      "rewards/forensics_sinkhorn_reward": 0.6202393174171448,
      "rewards/forensics_strict_edge_reward": 0.008541995572159067,
      "rewards/forensics_tiou_reward": 0.3853808790445328,
      "step": 631
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.9400921658986174,
      "grad_norm": 6.125938892364502,
      "kl": 0.3759765625,
      "learning_rate": 5.146153846153846e-07,
      "loss": 0.0687,
      "reward": 0.7863548994064331,
      "reward_std": 0.02655080985277891,
      "reward_std_all": 0.026618897914886475,
      "rewards-max/forensics_sinkhorn_reward": 0.9623780846595764,
      "rewards-max/forensics_strict_edge_reward": 0.14149524457752705,
      "rewards-max/forensics_tiou_reward": 0.9286180138587952,
      "rewards/forensics_sinkhorn_reward": 0.6973210871219635,
      "rewards/forensics_strict_edge_reward": 0.013466372271068394,
      "rewards/forensics_tiou_reward": 0.5208268612623215,
      "step": 632
    },
    {
      "completion_length": 25.859375,
      "epoch": 1.9431643625192012,
      "grad_norm": 6.428800106048584,
      "kl": 0.333984375,
      "learning_rate": 5.138461538461538e-07,
      "loss": 0.1508,
      "reward": 0.7091503739356995,
      "reward_std": 0.05700212623924017,
      "reward_std_all": 0.07629280351102352,
      "rewards-max/forensics_sinkhorn_reward": 0.9843067824840546,
      "rewards-max/forensics_strict_edge_reward": 0.606867641210556,
      "rewards-max/forensics_tiou_reward": 0.9873754382133484,
      "rewards/forensics_sinkhorn_reward": 0.6132956147193909,
      "rewards/forensics_strict_edge_reward": 0.04603304900228977,
      "rewards/forensics_tiou_reward": 0.46089014410972595,
      "step": 633
    },
    {
      "completion_length": 24.203125,
      "epoch": 1.946236559139785,
      "grad_norm": 6.008572578430176,
      "kl": 0.3662109375,
      "learning_rate": 5.13076923076923e-07,
      "loss": 0.0033,
      "reward": 0.7374498546123505,
      "reward_std": 0.0636730007827282,
      "reward_std_all": 0.06856447830796242,
      "rewards-max/forensics_sinkhorn_reward": 0.9665351212024689,
      "rewards-max/forensics_strict_edge_reward": 0.43805210292339325,
      "rewards-max/forensics_tiou_reward": 0.9394465088844299,
      "rewards/forensics_sinkhorn_reward": 0.6531864106655121,
      "rewards/forensics_strict_edge_reward": 0.04134219139814377,
      "rewards/forensics_tiou_reward": 0.4538380205631256,
      "step": 634
    },
    {
      "completion_length": 22.09375,
      "epoch": 1.9493087557603688,
      "grad_norm": 5.261870861053467,
      "kl": 0.2939453125,
      "learning_rate": 5.123076923076923e-07,
      "loss": 0.0503,
      "reward": 0.7702584266662598,
      "reward_std": 0.04419420287013054,
      "reward_std_all": 0.042406097054481506,
      "rewards-max/forensics_sinkhorn_reward": 0.9786931574344635,
      "rewards-max/forensics_strict_edge_reward": 0.6317985653877258,
      "rewards-max/forensics_tiou_reward": 0.969660609960556,
      "rewards/forensics_sinkhorn_reward": 0.746563047170639,
      "rewards/forensics_strict_edge_reward": 0.0745338462293148,
      "rewards/forensics_tiou_reward": 0.5740505158901215,
      "step": 635
    },
    {
      "completion_length": 22.96875,
      "epoch": 1.9523809523809523,
      "grad_norm": 5.5847272872924805,
      "kl": 0.37109375,
      "learning_rate": 5.115384615384615e-07,
      "loss": 0.0375,
      "reward": 0.7424050271511078,
      "reward_std": 0.04831336811184883,
      "reward_std_all": 0.0564850028604269,
      "rewards-max/forensics_sinkhorn_reward": 0.9584426283836365,
      "rewards-max/forensics_strict_edge_reward": 0.27330425195395947,
      "rewards-max/forensics_tiou_reward": 0.908871591091156,
      "rewards/forensics_sinkhorn_reward": 0.6690503060817719,
      "rewards/forensics_strict_edge_reward": 0.014571588253602386,
      "rewards/forensics_tiou_reward": 0.4315932095050812,
      "step": 636
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.955453149001536,
      "grad_norm": 7.156543254852295,
      "kl": 0.55859375,
      "learning_rate": 5.107692307692308e-07,
      "loss": 0.0458,
      "reward": 0.7473863363265991,
      "reward_std": 0.053442858159542084,
      "reward_std_all": 0.05652150325477123,
      "rewards-max/forensics_sinkhorn_reward": 0.94551682472229,
      "rewards-max/forensics_strict_edge_reward": 0.24788997322320938,
      "rewards-max/forensics_tiou_reward": 0.9026142954826355,
      "rewards/forensics_sinkhorn_reward": 0.6716234087944031,
      "rewards/forensics_strict_edge_reward": 0.029230501211714,
      "rewards/forensics_tiou_reward": 0.5052449405193329,
      "step": 637
    },
    {
      "completion_length": 24.6875,
      "epoch": 1.9585253456221197,
      "grad_norm": 4.9763875007629395,
      "kl": 0.3798828125,
      "learning_rate": 5.1e-07,
      "loss": 0.0295,
      "reward": 0.7572146654129028,
      "reward_std": 0.06683658808469772,
      "reward_std_all": 0.062168143689632416,
      "rewards-max/forensics_sinkhorn_reward": 0.9351977109909058,
      "rewards-max/forensics_strict_edge_reward": 0.2951139658689499,
      "rewards-max/forensics_tiou_reward": 0.8600393235683441,
      "rewards/forensics_sinkhorn_reward": 0.6945615708827972,
      "rewards/forensics_strict_edge_reward": 0.031208226922899485,
      "rewards/forensics_tiou_reward": 0.5103632211685181,
      "step": 638
    },
    {
      "completion_length": 28.125,
      "epoch": 1.9615975422427034,
      "grad_norm": 3.984262228012085,
      "kl": 0.2998046875,
      "learning_rate": 5.092307692307692e-07,
      "loss": -0.05,
      "reward": 0.7546897530555725,
      "reward_std": 0.05359949730336666,
      "reward_std_all": 0.054170072078704834,
      "rewards-max/forensics_sinkhorn_reward": 0.9741558730602264,
      "rewards-max/forensics_strict_edge_reward": 0.4038039594888687,
      "rewards-max/forensics_tiou_reward": 0.957338809967041,
      "rewards/forensics_sinkhorn_reward": 0.7161616384983063,
      "rewards/forensics_strict_edge_reward": 0.028833206743001938,
      "rewards/forensics_tiou_reward": 0.5337491631507874,
      "step": 639
    },
    {
      "completion_length": 25.515625,
      "epoch": 1.9646697388632872,
      "grad_norm": 5.487133979797363,
      "kl": 0.32421875,
      "learning_rate": 5.084615384615384e-07,
      "loss": 0.1166,
      "reward": 0.7499412894248962,
      "reward_std": 0.06781509518623352,
      "reward_std_all": 0.06448493432253599,
      "rewards-max/forensics_sinkhorn_reward": 0.9672250747680664,
      "rewards-max/forensics_strict_edge_reward": 0.1693999469280243,
      "rewards-max/forensics_tiou_reward": 0.9407178461551666,
      "rewards/forensics_sinkhorn_reward": 0.6618532836437225,
      "rewards/forensics_strict_edge_reward": 0.017907829955220222,
      "rewards/forensics_tiou_reward": 0.4611731916666031,
      "step": 640
    },
    {
      "completion_length": 28.234375,
      "epoch": 1.967741935483871,
      "grad_norm": 5.315008640289307,
      "kl": 0.3017578125,
      "learning_rate": 5.076923076923076e-07,
      "loss": -0.0216,
      "reward": 0.7437680661678314,
      "reward_std": 0.050768520683050156,
      "reward_std_all": 0.054854052141308784,
      "rewards-max/forensics_sinkhorn_reward": 0.9851274192333221,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9901753664016724,
      "rewards/forensics_sinkhorn_reward": 0.7126314342021942,
      "rewards/forensics_strict_edge_reward": 0.06107403337955475,
      "rewards/forensics_tiou_reward": 0.5233891308307648,
      "step": 641
    },
    {
      "completion_length": 23.765625,
      "epoch": 1.9708141321044548,
      "grad_norm": 5.892560958862305,
      "kl": 0.44921875,
      "learning_rate": 5.06923076923077e-07,
      "loss": 0.1278,
      "reward": 0.7051052451133728,
      "reward_std": 0.08006692677736282,
      "reward_std_all": 0.09455565735697746,
      "rewards-max/forensics_sinkhorn_reward": 0.9675032198429108,
      "rewards-max/forensics_strict_edge_reward": 0.43677301704883575,
      "rewards-max/forensics_tiou_reward": 0.9412364363670349,
      "rewards/forensics_sinkhorn_reward": 0.6550478637218475,
      "rewards/forensics_strict_edge_reward": 0.029727941378951073,
      "rewards/forensics_tiou_reward": 0.46901988983154297,
      "step": 642
    },
    {
      "completion_length": 27.546875,
      "epoch": 1.9738863287250386,
      "grad_norm": 6.6394219398498535,
      "kl": 0.3740234375,
      "learning_rate": 5.061538461538462e-07,
      "loss": 0.2081,
      "reward": 0.758729100227356,
      "reward_std": 0.060400715097784996,
      "reward_std_all": 0.056985875591635704,
      "rewards-max/forensics_sinkhorn_reward": 0.9617418944835663,
      "rewards-max/forensics_strict_edge_reward": 0.3713260293006897,
      "rewards-max/forensics_tiou_reward": 0.935612291097641,
      "rewards/forensics_sinkhorn_reward": 0.6833216547966003,
      "rewards/forensics_strict_edge_reward": 0.029450954869389534,
      "rewards/forensics_tiou_reward": 0.5188498795032501,
      "step": 643
    },
    {
      "completion_length": 24.125,
      "epoch": 1.976958525345622,
      "grad_norm": 5.170452117919922,
      "kl": 0.357421875,
      "learning_rate": 5.053846153846154e-07,
      "loss": 0.1633,
      "reward": 0.7039646804332733,
      "reward_std": 0.09652465581893921,
      "reward_std_all": 0.0977279394865036,
      "rewards-max/forensics_sinkhorn_reward": 0.9767210483551025,
      "rewards-max/forensics_strict_edge_reward": 0.5099464654922485,
      "rewards-max/forensics_tiou_reward": 0.965118944644928,
      "rewards/forensics_sinkhorn_reward": 0.626878172159195,
      "rewards/forensics_strict_edge_reward": 0.05174644663929939,
      "rewards/forensics_tiou_reward": 0.4586459845304489,
      "step": 644
    },
    {
      "completion_length": 24.421875,
      "epoch": 1.9800307219662059,
      "grad_norm": 6.596863746643066,
      "kl": 0.4501953125,
      "learning_rate": 5.046153846153846e-07,
      "loss": 0.0476,
      "reward": 0.7251000702381134,
      "reward_std": 0.09317167103290558,
      "reward_std_all": 0.08867675438523293,
      "rewards-max/forensics_sinkhorn_reward": 0.9685781896114349,
      "rewards-max/forensics_strict_edge_reward": 0.1968984380364418,
      "rewards-max/forensics_tiou_reward": 0.942301332950592,
      "rewards/forensics_sinkhorn_reward": 0.6030357182025909,
      "rewards/forensics_strict_edge_reward": 0.017873473349027336,
      "rewards/forensics_tiou_reward": 0.4173012673854828,
      "step": 645
    },
    {
      "completion_length": 23.859375,
      "epoch": 1.9831029185867894,
      "grad_norm": 5.561154842376709,
      "kl": 0.3095703125,
      "learning_rate": 5.038461538461538e-07,
      "loss": 0.0981,
      "reward": 0.748184859752655,
      "reward_std": 0.04807370901107788,
      "reward_std_all": 0.05444145016372204,
      "rewards-max/forensics_sinkhorn_reward": 0.9750393927097321,
      "rewards-max/forensics_strict_edge_reward": 0.48544541001319885,
      "rewards-max/forensics_tiou_reward": 0.9634908437728882,
      "rewards/forensics_sinkhorn_reward": 0.6524086594581604,
      "rewards/forensics_strict_edge_reward": 0.04545935522764921,
      "rewards/forensics_tiou_reward": 0.4728098660707474,
      "step": 646
    },
    {
      "completion_length": 30.9375,
      "epoch": 1.9861751152073732,
      "grad_norm": 4.3636345863342285,
      "kl": 0.451171875,
      "learning_rate": 5.03076923076923e-07,
      "loss": 0.1775,
      "reward": 0.6914364695549011,
      "reward_std": 0.1148470789194107,
      "reward_std_all": 0.11761182174086571,
      "rewards-max/forensics_sinkhorn_reward": 0.9034033715724945,
      "rewards-max/forensics_strict_edge_reward": 0.24242953391512856,
      "rewards-max/forensics_tiou_reward": 0.7977764010429382,
      "rewards/forensics_sinkhorn_reward": 0.5747857093811035,
      "rewards/forensics_strict_edge_reward": 0.016019956117816037,
      "rewards/forensics_tiou_reward": 0.38277600705623627,
      "step": 647
    },
    {
      "completion_length": 26.078125,
      "epoch": 1.989247311827957,
      "grad_norm": 5.813856601715088,
      "kl": 0.306640625,
      "learning_rate": 5.023076923076923e-07,
      "loss": 0.1822,
      "reward": 0.7079676985740662,
      "reward_std": 0.10406635329127312,
      "reward_std_all": 0.10101324878633022,
      "rewards-max/forensics_sinkhorn_reward": 0.9793969094753265,
      "rewards-max/forensics_strict_edge_reward": 0.4533289223909378,
      "rewards-max/forensics_tiou_reward": 0.9766879975795746,
      "rewards/forensics_sinkhorn_reward": 0.6452237963676453,
      "rewards/forensics_strict_edge_reward": 0.041238851845264435,
      "rewards/forensics_tiou_reward": 0.5077167898416519,
      "step": 648
    },
    {
      "completion_length": 26.578125,
      "epoch": 1.9923195084485408,
      "grad_norm": 4.942193031311035,
      "kl": 0.3427734375,
      "learning_rate": 5.015384615384614e-07,
      "loss": 0.0022,
      "reward": 0.7532391548156738,
      "reward_std": 0.05392560735344887,
      "reward_std_all": 0.05387907102704048,
      "rewards-max/forensics_sinkhorn_reward": 0.9830065369606018,
      "rewards-max/forensics_strict_edge_reward": 0.6112430393695831,
      "rewards-max/forensics_tiou_reward": 0.9838539361953735,
      "rewards/forensics_sinkhorn_reward": 0.7251310348510742,
      "rewards/forensics_strict_edge_reward": 0.045168060809373856,
      "rewards/forensics_tiou_reward": 0.5433337688446045,
      "step": 649
    },
    {
      "completion_length": 22.46875,
      "epoch": 1.9953917050691246,
      "grad_norm": 5.561490058898926,
      "kl": 0.3251953125,
      "learning_rate": 5.007692307692308e-07,
      "loss": 0.0388,
      "reward": 0.7359333038330078,
      "reward_std": 0.07953383028507233,
      "reward_std_all": 0.0759697612375021,
      "rewards-max/forensics_sinkhorn_reward": 0.9576567709445953,
      "rewards-max/forensics_strict_edge_reward": 0.502871572971344,
      "rewards-max/forensics_tiou_reward": 0.9073815643787384,
      "rewards/forensics_sinkhorn_reward": 0.6722316741943359,
      "rewards/forensics_strict_edge_reward": 0.04298247862607241,
      "rewards/forensics_tiou_reward": 0.4337015002965927,
      "step": 650
    },
    {
      "completion_length": 26.296875,
      "epoch": 1.998463901689708,
      "grad_norm": 6.653395175933838,
      "kl": 0.359375,
      "learning_rate": 5e-07,
      "loss": 0.0877,
      "reward": 0.7304840385913849,
      "reward_std": 0.03789366502314806,
      "reward_std_all": 0.05785398185253143,
      "rewards-max/forensics_sinkhorn_reward": 0.971373975276947,
      "rewards-max/forensics_strict_edge_reward": 0.6548857688903809,
      "rewards-max/forensics_tiou_reward": 0.9550952613353729,
      "rewards/forensics_sinkhorn_reward": 0.6639386415481567,
      "rewards/forensics_strict_edge_reward": 0.04728224314749241,
      "rewards/forensics_tiou_reward": 0.47429215908050537,
      "step": 651
    },
    {
      "completion_length": 18.200000762939453,
      "epoch": 2.0,
      "grad_norm": 3.216265916824341,
      "kl": 0.318359375,
      "learning_rate": 4.992307692307692e-07,
      "loss": -0.0085,
      "reward": 0.8028743863105774,
      "reward_std": 0.002027079463005066,
      "reward_std_all": 0.004560449160635471,
      "rewards-max/forensics_sinkhorn_reward": 0.8665762543678284,
      "rewards-max/forensics_strict_edge_reward": 0.011874840594828129,
      "rewards-max/forensics_tiou_reward": 0.7118862271308899,
      "rewards/forensics_sinkhorn_reward": 0.7304008603096008,
      "rewards/forensics_strict_edge_reward": 0.002375541254878044,
      "rewards/forensics_tiou_reward": 0.41789132356643677,
      "step": 652
    },
    {
      "completion_length": 25.984375,
      "epoch": 2.003072196620584,
      "grad_norm": 5.020893096923828,
      "kl": 0.4072265625,
      "learning_rate": 4.984615384615384e-07,
      "loss": 0.066,
      "reward": 0.7711257040500641,
      "reward_std": 0.03098741453140974,
      "reward_std_all": 0.03558139689266682,
      "rewards-max/forensics_sinkhorn_reward": 0.9598931968212128,
      "rewards-max/forensics_strict_edge_reward": 0.29996346682310104,
      "rewards-max/forensics_tiou_reward": 0.934171736240387,
      "rewards/forensics_sinkhorn_reward": 0.6639716327190399,
      "rewards/forensics_strict_edge_reward": 0.025316565297544003,
      "rewards/forensics_tiou_reward": 0.4879189729690552,
      "step": 653
    },
    {
      "completion_length": 27.953125,
      "epoch": 2.0061443932411676,
      "grad_norm": 4.7349324226379395,
      "kl": 0.521484375,
      "learning_rate": 4.976923076923077e-07,
      "loss": 0.1217,
      "reward": 0.7605102956295013,
      "reward_std": 0.053215671330690384,
      "reward_std_all": 0.052642740309238434,
      "rewards-max/forensics_sinkhorn_reward": 0.9646831452846527,
      "rewards-max/forensics_strict_edge_reward": 0.36184918880462646,
      "rewards-max/forensics_tiou_reward": 0.9330812692642212,
      "rewards/forensics_sinkhorn_reward": 0.6676186323165894,
      "rewards/forensics_strict_edge_reward": 0.02960611041635275,
      "rewards/forensics_tiou_reward": 0.5493954122066498,
      "step": 654
    },
    {
      "completion_length": 24.640625,
      "epoch": 2.0092165898617513,
      "grad_norm": 5.749842643737793,
      "kl": 0.3740234375,
      "learning_rate": 4.969230769230769e-07,
      "loss": 0.0287,
      "reward": 0.7896538972854614,
      "reward_std": 0.014389273535925895,
      "reward_std_all": 0.016787463682703674,
      "rewards-max/forensics_sinkhorn_reward": 0.9542061686515808,
      "rewards-max/forensics_strict_edge_reward": 0.3897520452737808,
      "rewards-max/forensics_tiou_reward": 0.91313835978508,
      "rewards/forensics_sinkhorn_reward": 0.748590886592865,
      "rewards/forensics_strict_edge_reward": 0.018816116265952587,
      "rewards/forensics_tiou_reward": 0.5517599284648895,
      "step": 655
    },
    {
      "completion_length": 26.296875,
      "epoch": 2.0122887864823347,
      "grad_norm": 5.328112602233887,
      "kl": 0.3994140625,
      "learning_rate": 4.961538461538462e-07,
      "loss": 0.1244,
      "reward": 0.7711369097232819,
      "reward_std": 0.03514622384682298,
      "reward_std_all": 0.03636102471500635,
      "rewards-max/forensics_sinkhorn_reward": 0.9614810049533844,
      "rewards-max/forensics_strict_edge_reward": 0.19456010311841965,
      "rewards-max/forensics_tiou_reward": 0.931710809469223,
      "rewards/forensics_sinkhorn_reward": 0.6846626996994019,
      "rewards/forensics_strict_edge_reward": 0.012569596525281668,
      "rewards/forensics_tiou_reward": 0.5413039922714233,
      "step": 656
    },
    {
      "completion_length": 25.59375,
      "epoch": 2.0153609831029184,
      "grad_norm": 5.32203483581543,
      "kl": 0.3154296875,
      "learning_rate": 4.953846153846154e-07,
      "loss": 0.0056,
      "reward": 0.7397976815700531,
      "reward_std": 0.04451012518256903,
      "reward_std_all": 0.058904360979795456,
      "rewards-max/forensics_sinkhorn_reward": 0.9829197824001312,
      "rewards-max/forensics_strict_edge_reward": 0.36972566694021225,
      "rewards-max/forensics_tiou_reward": 0.9847327172756195,
      "rewards/forensics_sinkhorn_reward": 0.734983891248703,
      "rewards/forensics_strict_edge_reward": 0.037241848185658455,
      "rewards/forensics_tiou_reward": 0.5645635426044464,
      "step": 657
    },
    {
      "completion_length": 23.9375,
      "epoch": 2.0184331797235022,
      "grad_norm": 5.1506428718566895,
      "kl": 0.451171875,
      "learning_rate": 4.946153846153846e-07,
      "loss": 0.1203,
      "reward": 0.7461795508861542,
      "reward_std": 0.05705544725060463,
      "reward_std_all": 0.05849980469793081,
      "rewards-max/forensics_sinkhorn_reward": 0.9705550968647003,
      "rewards-max/forensics_strict_edge_reward": 0.47332076728343964,
      "rewards-max/forensics_tiou_reward": 0.9554271996021271,
      "rewards/forensics_sinkhorn_reward": 0.714130312204361,
      "rewards/forensics_strict_edge_reward": 0.059784132055938244,
      "rewards/forensics_tiou_reward": 0.516027569770813,
      "step": 658
    },
    {
      "completion_length": 24.0625,
      "epoch": 2.021505376344086,
      "grad_norm": 5.013936996459961,
      "kl": 0.3955078125,
      "learning_rate": 4.938461538461538e-07,
      "loss": 0.0947,
      "reward": 0.7628488540649414,
      "reward_std": 0.035863734781742096,
      "reward_std_all": 0.04195781052112579,
      "rewards-max/forensics_sinkhorn_reward": 0.964963972568512,
      "rewards-max/forensics_strict_edge_reward": 0.4613940045237541,
      "rewards-max/forensics_tiou_reward": 0.9365738928318024,
      "rewards/forensics_sinkhorn_reward": 0.7142758369445801,
      "rewards/forensics_strict_edge_reward": 0.023815018124878407,
      "rewards/forensics_tiou_reward": 0.5095179527997971,
      "step": 659
    },
    {
      "completion_length": 23.140625,
      "epoch": 2.02457757296467,
      "grad_norm": 6.20326566696167,
      "kl": 0.416015625,
      "learning_rate": 4.93076923076923e-07,
      "loss": 0.023,
      "reward": 0.8013458847999573,
      "reward_std": 0.0072174667729996145,
      "reward_std_all": 0.0084776267176494,
      "rewards-max/forensics_sinkhorn_reward": 0.9858180582523346,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9921968281269073,
      "rewards/forensics_sinkhorn_reward": 0.8094125092029572,
      "rewards/forensics_strict_edge_reward": 0.08152930438518524,
      "rewards/forensics_tiou_reward": 0.6438049077987671,
      "step": 660
    },
    {
      "completion_length": 24.515625,
      "epoch": 2.0276497695852536,
      "grad_norm": 5.255027770996094,
      "kl": 0.353515625,
      "learning_rate": 4.923076923076923e-07,
      "loss": -0.0102,
      "reward": 0.7953201234340668,
      "reward_std": 0.017409421037882566,
      "reward_std_all": 0.01756253419443965,
      "rewards-max/forensics_sinkhorn_reward": 0.9824007749557495,
      "rewards-max/forensics_strict_edge_reward": 0.5686750710010529,
      "rewards-max/forensics_tiou_reward": 0.9819507300853729,
      "rewards/forensics_sinkhorn_reward": 0.7632474005222321,
      "rewards/forensics_strict_edge_reward": 0.04789992980659008,
      "rewards/forensics_tiou_reward": 0.5835382342338562,
      "step": 661
    },
    {
      "completion_length": 22.0625,
      "epoch": 2.0307219662058373,
      "grad_norm": 5.793727874755859,
      "kl": 0.314453125,
      "learning_rate": 4.915384615384615e-07,
      "loss": -0.0114,
      "reward": 0.7524654865264893,
      "reward_std": 0.041575477458536625,
      "reward_std_all": 0.048168658278882504,
      "rewards-max/forensics_sinkhorn_reward": 0.9544274508953094,
      "rewards-max/forensics_strict_edge_reward": 0.4855904970318079,
      "rewards-max/forensics_tiou_reward": 0.908496618270874,
      "rewards/forensics_sinkhorn_reward": 0.731269508600235,
      "rewards/forensics_strict_edge_reward": 0.04818215942941606,
      "rewards/forensics_tiou_reward": 0.5211688578128815,
      "step": 662
    },
    {
      "completion_length": 23.65625,
      "epoch": 2.033794162826421,
      "grad_norm": 7.141790866851807,
      "kl": 0.3173828125,
      "learning_rate": 4.907692307692308e-07,
      "loss": -0.0201,
      "reward": 0.7803420424461365,
      "reward_std": 0.029673250392079353,
      "reward_std_all": 0.030308766290545464,
      "rewards-max/forensics_sinkhorn_reward": 0.9777865707874298,
      "rewards-max/forensics_strict_edge_reward": 0.48898807168006897,
      "rewards-max/forensics_tiou_reward": 0.9726254045963287,
      "rewards/forensics_sinkhorn_reward": 0.784085214138031,
      "rewards/forensics_strict_edge_reward": 0.041431570425629616,
      "rewards/forensics_tiou_reward": 0.6351664066314697,
      "step": 663
    },
    {
      "completion_length": 26.359375,
      "epoch": 2.0368663594470044,
      "grad_norm": 6.128605365753174,
      "kl": 0.3505859375,
      "learning_rate": 4.9e-07,
      "loss": -0.0327,
      "reward": 0.7701829969882965,
      "reward_std": 0.04046042822301388,
      "reward_std_all": 0.03996534924954176,
      "rewards-max/forensics_sinkhorn_reward": 0.9783118069171906,
      "rewards-max/forensics_strict_edge_reward": 0.5583494007587433,
      "rewards-max/forensics_tiou_reward": 0.972753256559372,
      "rewards/forensics_sinkhorn_reward": 0.7333165407180786,
      "rewards/forensics_strict_edge_reward": 0.049909669905900955,
      "rewards/forensics_tiou_reward": 0.5433868765830994,
      "step": 664
    },
    {
      "completion_length": 26.34375,
      "epoch": 2.0399385560675882,
      "grad_norm": 6.421109676361084,
      "kl": 0.544921875,
      "learning_rate": 4.892307692307692e-07,
      "loss": 0.0677,
      "reward": 0.7368133366107941,
      "reward_std": 0.06550872512161732,
      "reward_std_all": 0.06614888086915016,
      "rewards-max/forensics_sinkhorn_reward": 0.9746414721012115,
      "rewards-max/forensics_strict_edge_reward": 0.3277502804994583,
      "rewards-max/forensics_tiou_reward": 0.9648540019989014,
      "rewards/forensics_sinkhorn_reward": 0.6410223543643951,
      "rewards/forensics_strict_edge_reward": 0.03369510266929865,
      "rewards/forensics_tiou_reward": 0.4950628876686096,
      "step": 665
    },
    {
      "completion_length": 21.390625,
      "epoch": 2.043010752688172,
      "grad_norm": 5.546144485473633,
      "kl": 0.3427734375,
      "learning_rate": 4.884615384615384e-07,
      "loss": 0.0387,
      "reward": 0.7570953071117401,
      "reward_std": 0.0279676029458642,
      "reward_std_all": 0.038145143538713455,
      "rewards-max/forensics_sinkhorn_reward": 0.9854718744754791,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9907469749450684,
      "rewards/forensics_sinkhorn_reward": 0.7260177731513977,
      "rewards/forensics_strict_edge_reward": 0.07187318429350853,
      "rewards/forensics_tiou_reward": 0.4759429097175598,
      "step": 666
    },
    {
      "completion_length": 24.828125,
      "epoch": 2.046082949308756,
      "grad_norm": 5.543783187866211,
      "kl": 0.4765625,
      "learning_rate": 4.876923076923077e-07,
      "loss": 0.0173,
      "reward": 0.7642806470394135,
      "reward_std": 0.0442879032343626,
      "reward_std_all": 0.04406045749783516,
      "rewards-max/forensics_sinkhorn_reward": 0.9831781387329102,
      "rewards-max/forensics_strict_edge_reward": 0.5820994228124619,
      "rewards-max/forensics_tiou_reward": 0.9829354286193848,
      "rewards/forensics_sinkhorn_reward": 0.6939463913440704,
      "rewards/forensics_strict_edge_reward": 0.06829847395420074,
      "rewards/forensics_tiou_reward": 0.4830315113067627,
      "step": 667
    },
    {
      "completion_length": 26.921875,
      "epoch": 2.0491551459293396,
      "grad_norm": 4.981592178344727,
      "kl": 0.369140625,
      "learning_rate": 4.869230769230769e-07,
      "loss": 0.0224,
      "reward": 0.7778767943382263,
      "reward_std": 0.034704593010246754,
      "reward_std_all": 0.0341911013238132,
      "rewards-max/forensics_sinkhorn_reward": 0.9784652590751648,
      "rewards-max/forensics_strict_edge_reward": 0.5037532150745392,
      "rewards-max/forensics_tiou_reward": 0.975646585226059,
      "rewards/forensics_sinkhorn_reward": 0.7525964379310608,
      "rewards/forensics_strict_edge_reward": 0.06093626655638218,
      "rewards/forensics_tiou_reward": 0.5920470952987671,
      "step": 668
    },
    {
      "completion_length": 22.453125,
      "epoch": 2.0522273425499233,
      "grad_norm": 5.349271297454834,
      "kl": 0.3984375,
      "learning_rate": 4.861538461538461e-07,
      "loss": -0.0436,
      "reward": 0.7676408886909485,
      "reward_std": 0.021069398382678628,
      "reward_std_all": 0.029053795151412487,
      "rewards-max/forensics_sinkhorn_reward": 0.9841226637363434,
      "rewards-max/forensics_strict_edge_reward": 0.7727466821670532,
      "rewards-max/forensics_tiou_reward": 0.9872617125511169,
      "rewards/forensics_sinkhorn_reward": 0.780633956193924,
      "rewards/forensics_strict_edge_reward": 0.08696696162223816,
      "rewards/forensics_tiou_reward": 0.6120937764644623,
      "step": 669
    },
    {
      "completion_length": 23.8125,
      "epoch": 2.055299539170507,
      "grad_norm": 6.794642448425293,
      "kl": 0.42578125,
      "learning_rate": 4.853846153846154e-07,
      "loss": 0.006,
      "reward": 0.7640391290187836,
      "reward_std": 0.03323754481971264,
      "reward_std_all": 0.038736844435334206,
      "rewards-max/forensics_sinkhorn_reward": 0.9684377908706665,
      "rewards-max/forensics_strict_edge_reward": 0.3667459934949875,
      "rewards-max/forensics_tiou_reward": 0.9494785070419312,
      "rewards/forensics_sinkhorn_reward": 0.734412431716919,
      "rewards/forensics_strict_edge_reward": 0.024664711207151413,
      "rewards/forensics_tiou_reward": 0.5247392058372498,
      "step": 670
    },
    {
      "completion_length": 25.828125,
      "epoch": 2.0583717357910905,
      "grad_norm": 5.232473850250244,
      "kl": 0.4326171875,
      "learning_rate": 4.846153846153846e-07,
      "loss": 0.1807,
      "reward": 0.7743873000144958,
      "reward_std": 0.04156231367960572,
      "reward_std_all": 0.03960189130157232,
      "rewards-max/forensics_sinkhorn_reward": 0.9819530844688416,
      "rewards-max/forensics_strict_edge_reward": 0.2861788719892502,
      "rewards-max/forensics_tiou_reward": 0.9786447882652283,
      "rewards/forensics_sinkhorn_reward": 0.6890750527381897,
      "rewards/forensics_strict_edge_reward": 0.0180709776468575,
      "rewards/forensics_tiou_reward": 0.5274934768676758,
      "step": 671
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.0614439324116742,
      "grad_norm": 5.750974655151367,
      "kl": 0.3564453125,
      "learning_rate": 4.838461538461539e-07,
      "loss": 0.0307,
      "reward": 0.7817159295082092,
      "reward_std": 0.02776514319702983,
      "reward_std_all": 0.027991097420454025,
      "rewards-max/forensics_sinkhorn_reward": 0.9824605584144592,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9809284806251526,
      "rewards/forensics_sinkhorn_reward": 0.7701334059238434,
      "rewards/forensics_strict_edge_reward": 0.05108881555497646,
      "rewards/forensics_tiou_reward": 0.570813924074173,
      "step": 672
    },
    {
      "completion_length": 24.484375,
      "epoch": 2.064516129032258,
      "grad_norm": 6.037294387817383,
      "kl": 0.3818359375,
      "learning_rate": 4.83076923076923e-07,
      "loss": 0.1325,
      "reward": 0.7651330530643463,
      "reward_std": 0.047081878408789635,
      "reward_std_all": 0.045895567163825035,
      "rewards-max/forensics_sinkhorn_reward": 0.9771781265735626,
      "rewards-max/forensics_strict_edge_reward": 0.5905451774597168,
      "rewards-max/forensics_tiou_reward": 0.9690744876861572,
      "rewards/forensics_sinkhorn_reward": 0.6877857148647308,
      "rewards/forensics_strict_edge_reward": 0.07927847653627396,
      "rewards/forensics_tiou_reward": 0.5034136772155762,
      "step": 673
    },
    {
      "completion_length": 21.546875,
      "epoch": 2.067588325652842,
      "grad_norm": 5.744025230407715,
      "kl": 0.3330078125,
      "learning_rate": 4.823076923076923e-07,
      "loss": -0.0397,
      "reward": 0.79951012134552,
      "reward_std": 0.011407339887227863,
      "reward_std_all": 0.012098844861611724,
      "rewards-max/forensics_sinkhorn_reward": 0.9760497808456421,
      "rewards-max/forensics_strict_edge_reward": 0.6092528700828552,
      "rewards-max/forensics_tiou_reward": 0.9611881077289581,
      "rewards/forensics_sinkhorn_reward": 0.8270289599895477,
      "rewards/forensics_strict_edge_reward": 0.06127926241606474,
      "rewards/forensics_tiou_reward": 0.6366404592990875,
      "step": 674
    },
    {
      "completion_length": 23.453125,
      "epoch": 2.0706605222734256,
      "grad_norm": 7.004373550415039,
      "kl": 0.4296875,
      "learning_rate": 4.815384615384615e-07,
      "loss": -0.0036,
      "reward": 0.7783187925815582,
      "reward_std": 0.03011118248105049,
      "reward_std_all": 0.030394083820283413,
      "rewards-max/forensics_sinkhorn_reward": 0.9608074724674225,
      "rewards-max/forensics_strict_edge_reward": 0.6321712285280228,
      "rewards-max/forensics_tiou_reward": 0.9286422431468964,
      "rewards/forensics_sinkhorn_reward": 0.753559947013855,
      "rewards/forensics_strict_edge_reward": 0.07681480050086975,
      "rewards/forensics_tiou_reward": 0.5630025267601013,
      "step": 675
    },
    {
      "completion_length": 24.515625,
      "epoch": 2.0737327188940093,
      "grad_norm": 6.155462741851807,
      "kl": 0.3701171875,
      "learning_rate": 4.807692307692307e-07,
      "loss": -0.0256,
      "reward": 0.7674347460269928,
      "reward_std": 0.031817901879549026,
      "reward_std_all": 0.03493468835949898,
      "rewards-max/forensics_sinkhorn_reward": 0.9814818203449249,
      "rewards-max/forensics_strict_edge_reward": 0.6078790128231049,
      "rewards-max/forensics_tiou_reward": 0.9816002249717712,
      "rewards/forensics_sinkhorn_reward": 0.7490690648555756,
      "rewards/forensics_strict_edge_reward": 0.07909658923745155,
      "rewards/forensics_tiou_reward": 0.5791574567556381,
      "step": 676
    },
    {
      "completion_length": 24.609375,
      "epoch": 2.076804915514593,
      "grad_norm": 4.9622039794921875,
      "kl": 0.388671875,
      "learning_rate": 4.8e-07,
      "loss": 0.0018,
      "reward": 0.7861943542957306,
      "reward_std": 0.031016552820801735,
      "reward_std_all": 0.029412182047963142,
      "rewards-max/forensics_sinkhorn_reward": 0.9672823548316956,
      "rewards-max/forensics_strict_edge_reward": 0.36893177032470703,
      "rewards-max/forensics_tiou_reward": 0.9448032081127167,
      "rewards/forensics_sinkhorn_reward": 0.721947968006134,
      "rewards/forensics_strict_edge_reward": 0.05144459381699562,
      "rewards/forensics_tiou_reward": 0.5507501065731049,
      "step": 677
    },
    {
      "completion_length": 21.125,
      "epoch": 2.0798771121351765,
      "grad_norm": 7.706777095794678,
      "kl": 0.3056640625,
      "learning_rate": 4.792307692307692e-07,
      "loss": -0.0453,
      "reward": 0.767181932926178,
      "reward_std": 0.0021688806591555476,
      "reward_std_all": 0.02254269691184163,
      "rewards-max/forensics_sinkhorn_reward": 0.9535709619522095,
      "rewards-max/forensics_strict_edge_reward": 0.19261780381202698,
      "rewards-max/forensics_tiou_reward": 0.9136668145656586,
      "rewards/forensics_sinkhorn_reward": 0.7489223182201385,
      "rewards/forensics_strict_edge_reward": 0.024386676028370857,
      "rewards/forensics_tiou_reward": 0.5244416296482086,
      "step": 678
    },
    {
      "completion_length": 26.484375,
      "epoch": 2.0829493087557602,
      "grad_norm": 7.441048622131348,
      "kl": 0.4501953125,
      "learning_rate": 4.784615384615384e-07,
      "loss": -0.0218,
      "reward": 0.7944026291370392,
      "reward_std": 0.014705394976772368,
      "reward_std_all": 0.015448322519659996,
      "rewards-max/forensics_sinkhorn_reward": 0.9830676019191742,
      "rewards-max/forensics_strict_edge_reward": 0.3683512359857559,
      "rewards-max/forensics_tiou_reward": 0.9837125837802887,
      "rewards/forensics_sinkhorn_reward": 0.7635237574577332,
      "rewards/forensics_strict_edge_reward": 0.07263191975653172,
      "rewards/forensics_tiou_reward": 0.5808859169483185,
      "step": 679
    },
    {
      "completion_length": 29.125,
      "epoch": 2.086021505376344,
      "grad_norm": 6.434823513031006,
      "kl": 0.3369140625,
      "learning_rate": 4.776923076923077e-07,
      "loss": -0.0153,
      "reward": 0.779335081577301,
      "reward_std": 0.01416543172672391,
      "reward_std_all": 0.0222492516040802,
      "rewards-max/forensics_sinkhorn_reward": 0.9460191130638123,
      "rewards-max/forensics_strict_edge_reward": 0.2023453414440155,
      "rewards-max/forensics_tiou_reward": 0.9020358920097351,
      "rewards/forensics_sinkhorn_reward": 0.7388316988945007,
      "rewards/forensics_strict_edge_reward": 0.019345424138009548,
      "rewards/forensics_tiou_reward": 0.558422178030014,
      "step": 680
    },
    {
      "completion_length": 25.0625,
      "epoch": 2.089093701996928,
      "grad_norm": 5.256580352783203,
      "kl": 0.462890625,
      "learning_rate": 4.769230769230769e-07,
      "loss": 0.0339,
      "reward": 0.7864809632301331,
      "reward_std": 0.03230239078402519,
      "reward_std_all": 0.03048084769397974,
      "rewards-max/forensics_sinkhorn_reward": 0.980289876461029,
      "rewards-max/forensics_strict_edge_reward": 0.18857790529727936,
      "rewards-max/forensics_tiou_reward": 0.9738172590732574,
      "rewards/forensics_sinkhorn_reward": 0.7593114078044891,
      "rewards/forensics_strict_edge_reward": 0.02368901250883937,
      "rewards/forensics_tiou_reward": 0.5743534564971924,
      "step": 681
    },
    {
      "completion_length": 23.625,
      "epoch": 2.0921658986175116,
      "grad_norm": 6.662400245666504,
      "kl": 0.388671875,
      "learning_rate": 4.7615384615384613e-07,
      "loss": -0.0514,
      "reward": 0.7975768446922302,
      "reward_std": 0.011382835451513529,
      "reward_std_all": 0.012427665293216705,
      "rewards-max/forensics_sinkhorn_reward": 0.9412274956703186,
      "rewards-max/forensics_strict_edge_reward": 0.5053723999299109,
      "rewards-max/forensics_tiou_reward": 0.8791050314903259,
      "rewards/forensics_sinkhorn_reward": 0.7603525519371033,
      "rewards/forensics_strict_edge_reward": 0.048244983481708914,
      "rewards/forensics_tiou_reward": 0.5334300398826599,
      "step": 682
    },
    {
      "completion_length": 27.21875,
      "epoch": 2.0952380952380953,
      "grad_norm": 5.203677177429199,
      "kl": 0.4052734375,
      "learning_rate": 4.753846153846154e-07,
      "loss": 0.1215,
      "reward": 0.7852047383785248,
      "reward_std": 0.026839648140594363,
      "reward_std_all": 0.026702266186475754,
      "rewards-max/forensics_sinkhorn_reward": 0.9709131717681885,
      "rewards-max/forensics_strict_edge_reward": 0.3102388232946396,
      "rewards-max/forensics_tiou_reward": 0.9517736434936523,
      "rewards/forensics_sinkhorn_reward": 0.7147130370140076,
      "rewards/forensics_strict_edge_reward": 0.022687687538564205,
      "rewards/forensics_tiou_reward": 0.5843837857246399,
      "step": 683
    },
    {
      "completion_length": 24.234375,
      "epoch": 2.098310291858679,
      "grad_norm": 5.7482008934021,
      "kl": 0.412109375,
      "learning_rate": 4.7461538461538456e-07,
      "loss": 0.0291,
      "reward": 0.7717690169811249,
      "reward_std": 0.04764716699719429,
      "reward_std_all": 0.044677495025098324,
      "rewards-max/forensics_sinkhorn_reward": 0.9542110860347748,
      "rewards-max/forensics_strict_edge_reward": 0.4579825960099697,
      "rewards-max/forensics_tiou_reward": 0.9210802912712097,
      "rewards/forensics_sinkhorn_reward": 0.6592164933681488,
      "rewards/forensics_strict_edge_reward": 0.022789115784689784,
      "rewards/forensics_tiou_reward": 0.4807400107383728,
      "step": 684
    },
    {
      "completion_length": 25.390625,
      "epoch": 2.1013824884792625,
      "grad_norm": 6.837858200073242,
      "kl": 0.3779296875,
      "learning_rate": 4.738461538461538e-07,
      "loss": 0.0246,
      "reward": 0.7682484984397888,
      "reward_std": 0.035520028322935104,
      "reward_std_all": 0.037346324883401394,
      "rewards-max/forensics_sinkhorn_reward": 0.9731687903404236,
      "rewards-max/forensics_strict_edge_reward": 0.5881353616714478,
      "rewards-max/forensics_tiou_reward": 0.9599300622940063,
      "rewards/forensics_sinkhorn_reward": 0.6936996579170227,
      "rewards/forensics_strict_edge_reward": 0.044886671006679535,
      "rewards/forensics_tiou_reward": 0.4705323129892349,
      "step": 685
    },
    {
      "completion_length": 26.5,
      "epoch": 2.1044546850998462,
      "grad_norm": 5.859909534454346,
      "kl": 0.3515625,
      "learning_rate": 4.7307692307692304e-07,
      "loss": 0.0127,
      "reward": 0.7956378161907196,
      "reward_std": 0.00989415223011747,
      "reward_std_all": 0.01206527091562748,
      "rewards-max/forensics_sinkhorn_reward": 0.9565039873123169,
      "rewards-max/forensics_strict_edge_reward": 0.3166895806789398,
      "rewards-max/forensics_tiou_reward": 0.9123151004314423,
      "rewards/forensics_sinkhorn_reward": 0.7230867445468903,
      "rewards/forensics_strict_edge_reward": 0.019834873732179403,
      "rewards/forensics_tiou_reward": 0.5125174969434738,
      "step": 686
    },
    {
      "completion_length": 26.453125,
      "epoch": 2.10752688172043,
      "grad_norm": 6.77433443069458,
      "kl": 0.50390625,
      "learning_rate": 4.723076923076923e-07,
      "loss": 0.035,
      "reward": 0.789518415927887,
      "reward_std": 0.01837752596475184,
      "reward_std_all": 0.019173337146639824,
      "rewards-max/forensics_sinkhorn_reward": 0.9811879992485046,
      "rewards-max/forensics_strict_edge_reward": 0.6175537705421448,
      "rewards-max/forensics_tiou_reward": 0.9797897040843964,
      "rewards/forensics_sinkhorn_reward": 0.7614184617996216,
      "rewards/forensics_strict_edge_reward": 0.0724487230181694,
      "rewards/forensics_tiou_reward": 0.596285492181778,
      "step": 687
    },
    {
      "completion_length": 24.296875,
      "epoch": 2.110599078341014,
      "grad_norm": 5.551341533660889,
      "kl": 0.412109375,
      "learning_rate": 4.715384615384615e-07,
      "loss": -0.0408,
      "reward": 0.7953372597694397,
      "reward_std": 0.01440890203230083,
      "reward_std_all": 0.01507786987349391,
      "rewards-max/forensics_sinkhorn_reward": 0.9717441201210022,
      "rewards-max/forensics_strict_edge_reward": 0.5100585669279099,
      "rewards-max/forensics_tiou_reward": 0.9546493291854858,
      "rewards/forensics_sinkhorn_reward": 0.7749777138233185,
      "rewards/forensics_strict_edge_reward": 0.04962421581149101,
      "rewards/forensics_tiou_reward": 0.614896684885025,
      "step": 688
    },
    {
      "completion_length": 26.015625,
      "epoch": 2.1136712749615976,
      "grad_norm": 5.702138423919678,
      "kl": 0.3056640625,
      "learning_rate": 4.7076923076923073e-07,
      "loss": -0.0012,
      "reward": 0.7551742792129517,
      "reward_std": 0.042986318469047546,
      "reward_std_all": 0.045964525546878576,
      "rewards-max/forensics_sinkhorn_reward": 0.9860103130340576,
      "rewards-max/forensics_strict_edge_reward": 0.679555356502533,
      "rewards-max/forensics_tiou_reward": 0.9930330812931061,
      "rewards/forensics_sinkhorn_reward": 0.6904090344905853,
      "rewards/forensics_strict_edge_reward": 0.0611073300242424,
      "rewards/forensics_tiou_reward": 0.5230407416820526,
      "step": 689
    },
    {
      "completion_length": 29.390625,
      "epoch": 2.1167434715821813,
      "grad_norm": 4.925398826599121,
      "kl": 0.328125,
      "learning_rate": 4.6999999999999995e-07,
      "loss": 0.0606,
      "reward": 0.7340693473815918,
      "reward_std": 0.04894182272255421,
      "reward_std_all": 0.0651413556188345,
      "rewards-max/forensics_sinkhorn_reward": 0.9556737244129181,
      "rewards-max/forensics_strict_edge_reward": 0.31437574326992035,
      "rewards-max/forensics_tiou_reward": 0.9142962098121643,
      "rewards/forensics_sinkhorn_reward": 0.6980376541614532,
      "rewards/forensics_strict_edge_reward": 0.043070070911198854,
      "rewards/forensics_tiou_reward": 0.5261856317520142,
      "step": 690
    },
    {
      "completion_length": 28.09375,
      "epoch": 2.119815668202765,
      "grad_norm": 6.188426971435547,
      "kl": 0.4638671875,
      "learning_rate": 4.692307692307692e-07,
      "loss": 0.2111,
      "reward": 0.7614595293998718,
      "reward_std": 0.041258641984313726,
      "reward_std_all": 0.0453169634565711,
      "rewards-max/forensics_sinkhorn_reward": 0.9828290939331055,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9856761991977692,
      "rewards/forensics_sinkhorn_reward": 0.6845691502094269,
      "rewards/forensics_strict_edge_reward": 0.04723895527422428,
      "rewards/forensics_tiou_reward": 0.5498560070991516,
      "step": 691
    },
    {
      "completion_length": 25.421875,
      "epoch": 2.122887864823349,
      "grad_norm": 12.747164726257324,
      "kl": 0.546875,
      "learning_rate": 4.684615384615385e-07,
      "loss": -0.0109,
      "reward": 0.7303738594055176,
      "reward_std": 0.04730231687426567,
      "reward_std_all": 0.06270843371748924,
      "rewards-max/forensics_sinkhorn_reward": 0.9210792481899261,
      "rewards-max/forensics_strict_edge_reward": 0.34284206479787827,
      "rewards-max/forensics_tiou_reward": 0.8397204577922821,
      "rewards/forensics_sinkhorn_reward": 0.6837210059165955,
      "rewards/forensics_strict_edge_reward": 0.02855884749442339,
      "rewards/forensics_tiou_reward": 0.5462197661399841,
      "step": 692
    },
    {
      "completion_length": 24.015625,
      "epoch": 2.1259600614439322,
      "grad_norm": 7.582066059112549,
      "kl": 0.5224609375,
      "learning_rate": 4.6769230769230764e-07,
      "loss": 0.001,
      "reward": 0.7530255615711212,
      "reward_std": 0.04362284857779741,
      "reward_std_all": 0.04901360347867012,
      "rewards-max/forensics_sinkhorn_reward": 0.9639275074005127,
      "rewards-max/forensics_strict_edge_reward": 0.4726305305957794,
      "rewards-max/forensics_tiou_reward": 0.9380297064781189,
      "rewards/forensics_sinkhorn_reward": 0.707865834236145,
      "rewards/forensics_strict_edge_reward": 0.047490546479821205,
      "rewards/forensics_tiou_reward": 0.5145977139472961,
      "step": 693
    },
    {
      "completion_length": 24.859375,
      "epoch": 2.129032258064516,
      "grad_norm": 6.669990062713623,
      "kl": 0.4599609375,
      "learning_rate": 4.669230769230769e-07,
      "loss": -0.0226,
      "reward": 0.7855381965637207,
      "reward_std": 0.012288725934922695,
      "reward_std_all": 0.018862538039684296,
      "rewards-max/forensics_sinkhorn_reward": 0.9588893055915833,
      "rewards-max/forensics_strict_edge_reward": 0.41155266761779785,
      "rewards-max/forensics_tiou_reward": 0.930169016122818,
      "rewards/forensics_sinkhorn_reward": 0.7328020930290222,
      "rewards/forensics_strict_edge_reward": 0.04445425979793072,
      "rewards/forensics_tiou_reward": 0.518514096736908,
      "step": 694
    },
    {
      "completion_length": 25.375,
      "epoch": 2.1321044546851,
      "grad_norm": 4.61439847946167,
      "kl": 0.490234375,
      "learning_rate": 4.661538461538461e-07,
      "loss": 0.033,
      "reward": 0.766772449016571,
      "reward_std": 0.0348364133387804,
      "reward_std_all": 0.03676725085824728,
      "rewards-max/forensics_sinkhorn_reward": 0.9762517511844635,
      "rewards-max/forensics_strict_edge_reward": 0.5873754322528839,
      "rewards-max/forensics_tiou_reward": 0.9675202667713165,
      "rewards/forensics_sinkhorn_reward": 0.7521666884422302,
      "rewards/forensics_strict_edge_reward": 0.0432270634919405,
      "rewards/forensics_tiou_reward": 0.5864619612693787,
      "step": 695
    },
    {
      "completion_length": 22.921875,
      "epoch": 2.1351766513056836,
      "grad_norm": 6.59634256362915,
      "kl": 0.3154296875,
      "learning_rate": 4.653846153846154e-07,
      "loss": -0.0407,
      "reward": 0.7601171433925629,
      "reward_std": 0.057022787630558014,
      "reward_std_all": 0.05433841422200203,
      "rewards-max/forensics_sinkhorn_reward": 0.9718262553215027,
      "rewards-max/forensics_strict_edge_reward": 0.5742200464010239,
      "rewards-max/forensics_tiou_reward": 0.9575502276420593,
      "rewards/forensics_sinkhorn_reward": 0.7591904997825623,
      "rewards/forensics_strict_edge_reward": 0.05613113008439541,
      "rewards/forensics_tiou_reward": 0.5643147826194763,
      "step": 696
    },
    {
      "completion_length": 24.453125,
      "epoch": 2.1382488479262673,
      "grad_norm": 4.797981262207031,
      "kl": 0.501953125,
      "learning_rate": 4.646153846153846e-07,
      "loss": 0.027,
      "reward": 0.7969471514225006,
      "reward_std": 0.016788756474852562,
      "reward_std_all": 0.016267860308289528,
      "rewards-max/forensics_sinkhorn_reward": 0.9863792359828949,
      "rewards-max/forensics_strict_edge_reward": 0.6924179792404175,
      "rewards-max/forensics_tiou_reward": 0.9910253882408142,
      "rewards/forensics_sinkhorn_reward": 0.8134369552135468,
      "rewards/forensics_strict_edge_reward": 0.08093732595443726,
      "rewards/forensics_tiou_reward": 0.6824132800102234,
      "step": 697
    },
    {
      "completion_length": 24.796875,
      "epoch": 2.141321044546851,
      "grad_norm": 9.2638578414917,
      "kl": 0.6044921875,
      "learning_rate": 4.638461538461538e-07,
      "loss": 0.0614,
      "reward": 0.7941315770149231,
      "reward_std": 0.016920043155550957,
      "reward_std_all": 0.016834702342748642,
      "rewards-max/forensics_sinkhorn_reward": 0.9773492217063904,
      "rewards-max/forensics_strict_edge_reward": 0.5641956627368927,
      "rewards-max/forensics_tiou_reward": 0.9650619029998779,
      "rewards/forensics_sinkhorn_reward": 0.752010703086853,
      "rewards/forensics_strict_edge_reward": 0.04607773572206497,
      "rewards/forensics_tiou_reward": 0.5733922123908997,
      "step": 698
    },
    {
      "completion_length": 30.09375,
      "epoch": 2.144393241167435,
      "grad_norm": 4.840202331542969,
      "kl": 0.4189453125,
      "learning_rate": 4.6307692307692303e-07,
      "loss": -0.0233,
      "reward": 0.7856810092926025,
      "reward_std": 0.021372949704527855,
      "reward_std_all": 0.022335028275847435,
      "rewards-max/forensics_sinkhorn_reward": 0.9730056822299957,
      "rewards-max/forensics_strict_edge_reward": 0.6979723870754242,
      "rewards-max/forensics_tiou_reward": 0.961161732673645,
      "rewards/forensics_sinkhorn_reward": 0.7728745341300964,
      "rewards/forensics_strict_edge_reward": 0.07222940400242805,
      "rewards/forensics_tiou_reward": 0.6356962621212006,
      "step": 699
    },
    {
      "completion_length": 23.75,
      "epoch": 2.1474654377880182,
      "grad_norm": 5.470643043518066,
      "kl": 0.392578125,
      "learning_rate": 4.623076923076923e-07,
      "loss": -0.0596,
      "reward": 0.7867778241634369,
      "reward_std": 0.016875123255886137,
      "reward_std_all": 0.019034080673009157,
      "rewards-max/forensics_sinkhorn_reward": 0.9664246737957001,
      "rewards-max/forensics_strict_edge_reward": 0.5515579879283905,
      "rewards-max/forensics_tiou_reward": 0.9442552328109741,
      "rewards/forensics_sinkhorn_reward": 0.7596653401851654,
      "rewards/forensics_strict_edge_reward": 0.053863607347011566,
      "rewards/forensics_tiou_reward": 0.5673722326755524,
      "step": 700
    },
    {
      "completion_length": 24.171875,
      "epoch": 2.150537634408602,
      "grad_norm": 7.185388088226318,
      "kl": 0.400390625,
      "learning_rate": 4.6153846153846156e-07,
      "loss": -0.0239,
      "reward": 0.7542409598827362,
      "reward_std": 0.025478347204625607,
      "reward_std_all": 0.038078922778367996,
      "rewards-max/forensics_sinkhorn_reward": 0.9543513059616089,
      "rewards-max/forensics_strict_edge_reward": 0.23620246909558773,
      "rewards-max/forensics_tiou_reward": 0.9187324643135071,
      "rewards/forensics_sinkhorn_reward": 0.712353527545929,
      "rewards/forensics_strict_edge_reward": 0.015169837512075901,
      "rewards/forensics_tiou_reward": 0.4947931319475174,
      "step": 701
    },
    {
      "completion_length": 27.625,
      "epoch": 2.153609831029186,
      "grad_norm": 5.608293533325195,
      "kl": 0.365234375,
      "learning_rate": 4.607692307692307e-07,
      "loss": 0.0893,
      "reward": 0.7854580581188202,
      "reward_std": 0.030081079341471195,
      "reward_std_all": 0.028738392516970634,
      "rewards-max/forensics_sinkhorn_reward": 0.9668695330619812,
      "rewards-max/forensics_strict_edge_reward": 0.3672405108809471,
      "rewards-max/forensics_tiou_reward": 0.943444550037384,
      "rewards/forensics_sinkhorn_reward": 0.7259869873523712,
      "rewards/forensics_strict_edge_reward": 0.0466619748622179,
      "rewards/forensics_tiou_reward": 0.5600429177284241,
      "step": 702
    },
    {
      "completion_length": 23.5625,
      "epoch": 2.1566820276497696,
      "grad_norm": 6.399641990661621,
      "kl": 0.5087890625,
      "learning_rate": 4.6e-07,
      "loss": 0.0755,
      "reward": 0.7467302083969116,
      "reward_std": 0.06972568854689598,
      "reward_std_all": 0.06944184005260468,
      "rewards-max/forensics_sinkhorn_reward": 0.972239077091217,
      "rewards-max/forensics_strict_edge_reward": 0.4755072444677353,
      "rewards-max/forensics_tiou_reward": 0.9613397419452667,
      "rewards/forensics_sinkhorn_reward": 0.6590532660484314,
      "rewards/forensics_strict_edge_reward": 0.035288828425109386,
      "rewards/forensics_tiou_reward": 0.5067228227853775,
      "step": 703
    },
    {
      "completion_length": 25.671875,
      "epoch": 2.1597542242703534,
      "grad_norm": 8.674038887023926,
      "kl": 0.4365234375,
      "learning_rate": 4.592307692307692e-07,
      "loss": 0.0507,
      "reward": 0.7769809365272522,
      "reward_std": 0.04814308322966099,
      "reward_std_all": 0.04401983693242073,
      "rewards-max/forensics_sinkhorn_reward": 0.9581728279590607,
      "rewards-max/forensics_strict_edge_reward": 0.6222288608551025,
      "rewards-max/forensics_tiou_reward": 0.9259491562843323,
      "rewards/forensics_sinkhorn_reward": 0.6642994284629822,
      "rewards/forensics_strict_edge_reward": 0.09035462141036987,
      "rewards/forensics_tiou_reward": 0.49694332480430603,
      "step": 704
    },
    {
      "completion_length": 25.203125,
      "epoch": 2.162826420890937,
      "grad_norm": 5.354875564575195,
      "kl": 0.361328125,
      "learning_rate": 4.5846153846153847e-07,
      "loss": 0.0574,
      "reward": 0.7386569380760193,
      "reward_std": 0.08597889542579651,
      "reward_std_all": 0.08045633137226105,
      "rewards-max/forensics_sinkhorn_reward": 0.9401615262031555,
      "rewards-max/forensics_strict_edge_reward": 0.15126588940620422,
      "rewards-max/forensics_tiou_reward": 0.8769669532775879,
      "rewards/forensics_sinkhorn_reward": 0.6891166269779205,
      "rewards/forensics_strict_edge_reward": 0.016078193672001362,
      "rewards/forensics_tiou_reward": 0.47106410562992096,
      "step": 705
    },
    {
      "completion_length": 21.75,
      "epoch": 2.165898617511521,
      "grad_norm": 8.452033996582031,
      "kl": 0.263671875,
      "learning_rate": 4.5769230769230763e-07,
      "loss": -0.0241,
      "reward": 0.7777745127677917,
      "reward_std": 0.03387463465332985,
      "reward_std_all": 0.033959406428039074,
      "rewards-max/forensics_sinkhorn_reward": 0.9607173800468445,
      "rewards-max/forensics_strict_edge_reward": 0.23242570832371712,
      "rewards-max/forensics_tiou_reward": 0.9269010722637177,
      "rewards/forensics_sinkhorn_reward": 0.7631315588951111,
      "rewards/forensics_strict_edge_reward": 0.02697958517819643,
      "rewards/forensics_tiou_reward": 0.551916241645813,
      "step": 706
    },
    {
      "completion_length": 26.75,
      "epoch": 2.1689708141321047,
      "grad_norm": 5.954933166503906,
      "kl": 0.3642578125,
      "learning_rate": 4.569230769230769e-07,
      "loss": 0.0193,
      "reward": 0.7901183068752289,
      "reward_std": 0.02277557598426938,
      "reward_std_all": 0.022511450573801994,
      "rewards-max/forensics_sinkhorn_reward": 0.983776330947876,
      "rewards-max/forensics_strict_edge_reward": 0.5062626451253891,
      "rewards-max/forensics_tiou_reward": 0.9849265217781067,
      "rewards/forensics_sinkhorn_reward": 0.7244177758693695,
      "rewards/forensics_strict_edge_reward": 0.06033847853541374,
      "rewards/forensics_tiou_reward": 0.5919649302959442,
      "step": 707
    },
    {
      "completion_length": 24.328125,
      "epoch": 2.172043010752688,
      "grad_norm": 6.676597595214844,
      "kl": 0.482421875,
      "learning_rate": 4.561538461538461e-07,
      "loss": 0.0257,
      "reward": 0.7833764851093292,
      "reward_std": 0.01682988414540887,
      "reward_std_all": 0.0229013841599226,
      "rewards-max/forensics_sinkhorn_reward": 0.9727939069271088,
      "rewards-max/forensics_strict_edge_reward": 0.2485707327723503,
      "rewards-max/forensics_tiou_reward": 0.9525197744369507,
      "rewards/forensics_sinkhorn_reward": 0.766561895608902,
      "rewards/forensics_strict_edge_reward": 0.013833908131346107,
      "rewards/forensics_tiou_reward": 0.6226375102996826,
      "step": 708
    },
    {
      "completion_length": 24.046875,
      "epoch": 2.175115207373272,
      "grad_norm": 5.988928318023682,
      "kl": 0.521484375,
      "learning_rate": 4.553846153846154e-07,
      "loss": -0.062,
      "reward": 0.7985536754131317,
      "reward_std": 0.006715084949973971,
      "reward_std_all": 0.009090386563912034,
      "rewards-max/forensics_sinkhorn_reward": 0.9717249274253845,
      "rewards-max/forensics_strict_edge_reward": 0.40167564153671265,
      "rewards-max/forensics_tiou_reward": 0.94810551404953,
      "rewards/forensics_sinkhorn_reward": 0.7830943465232849,
      "rewards/forensics_strict_edge_reward": 0.026425521355122328,
      "rewards/forensics_tiou_reward": 0.5873130261898041,
      "step": 709
    },
    {
      "completion_length": 25.609375,
      "epoch": 2.1781874039938556,
      "grad_norm": 7.131364822387695,
      "kl": 0.390625,
      "learning_rate": 4.546153846153846e-07,
      "loss": 0.0026,
      "reward": 0.753911554813385,
      "reward_std": 0.03982479590922594,
      "reward_std_all": 0.046225191093981266,
      "rewards-max/forensics_sinkhorn_reward": 0.9785464704036713,
      "rewards-max/forensics_strict_edge_reward": 0.4533289223909378,
      "rewards-max/forensics_tiou_reward": 0.968969076871872,
      "rewards/forensics_sinkhorn_reward": 0.7236286401748657,
      "rewards/forensics_strict_edge_reward": 0.06411448121070862,
      "rewards/forensics_tiou_reward": 0.5674233436584473,
      "step": 710
    },
    {
      "completion_length": 29.671875,
      "epoch": 2.1812596006144394,
      "grad_norm": 4.917402744293213,
      "kl": 0.423828125,
      "learning_rate": 4.538461538461538e-07,
      "loss": -0.042,
      "reward": 0.7583741545677185,
      "reward_std": 0.030744836665689945,
      "reward_std_all": 0.04021631879732013,
      "rewards-max/forensics_sinkhorn_reward": 0.9865356385707855,
      "rewards-max/forensics_strict_edge_reward": 0.718811422586441,
      "rewards-max/forensics_tiou_reward": 0.9947079718112946,
      "rewards/forensics_sinkhorn_reward": 0.7039739787578583,
      "rewards/forensics_strict_edge_reward": 0.0664457306265831,
      "rewards/forensics_tiou_reward": 0.5616534650325775,
      "step": 711
    },
    {
      "completion_length": 25.78125,
      "epoch": 2.184331797235023,
      "grad_norm": 5.883525371551514,
      "kl": 0.484375,
      "learning_rate": 4.5307692307692307e-07,
      "loss": -0.0376,
      "reward": 0.7683768272399902,
      "reward_std": 0.03876568842679262,
      "reward_std_all": 0.03948191553354263,
      "rewards-max/forensics_sinkhorn_reward": 0.9723235964775085,
      "rewards-max/forensics_strict_edge_reward": 0.5064684301614761,
      "rewards-max/forensics_tiou_reward": 0.9565038084983826,
      "rewards/forensics_sinkhorn_reward": 0.6929868459701538,
      "rewards/forensics_strict_edge_reward": 0.039182282984256744,
      "rewards/forensics_tiou_reward": 0.48953332006931305,
      "step": 712
    },
    {
      "completion_length": 24.046875,
      "epoch": 2.187403993855607,
      "grad_norm": 5.583851337432861,
      "kl": 0.4169921875,
      "learning_rate": 4.523076923076923e-07,
      "loss": 0.0089,
      "reward": 0.787784993648529,
      "reward_std": 0.029706287663429976,
      "reward_std_all": 0.028265522327274084,
      "rewards-max/forensics_sinkhorn_reward": 0.9492484629154205,
      "rewards-max/forensics_strict_edge_reward": 0.22773585096001625,
      "rewards-max/forensics_tiou_reward": 0.8970129489898682,
      "rewards/forensics_sinkhorn_reward": 0.748288094997406,
      "rewards/forensics_strict_edge_reward": 0.02592054521664977,
      "rewards/forensics_tiou_reward": 0.5467528998851776,
      "step": 713
    },
    {
      "completion_length": 22.78125,
      "epoch": 2.1904761904761907,
      "grad_norm": 5.730156421661377,
      "kl": 0.3876953125,
      "learning_rate": 4.5153846153846155e-07,
      "loss": -0.0892,
      "reward": 0.7989052534103394,
      "reward_std": 0.009510643314570189,
      "reward_std_all": 0.010698865633457899,
      "rewards-max/forensics_sinkhorn_reward": 0.9685227274894714,
      "rewards-max/forensics_strict_edge_reward": 0.42901745438575745,
      "rewards-max/forensics_tiou_reward": 0.9468106031417847,
      "rewards/forensics_sinkhorn_reward": 0.8022814393043518,
      "rewards/forensics_strict_edge_reward": 0.04250025749206543,
      "rewards/forensics_tiou_reward": 0.6274591088294983,
      "step": 714
    },
    {
      "completion_length": 24.453125,
      "epoch": 2.193548387096774,
      "grad_norm": 7.366287708282471,
      "kl": 0.6494140625,
      "learning_rate": 4.507692307692307e-07,
      "loss": 0.0319,
      "reward": 0.7568025588989258,
      "reward_std": 0.052047792822122574,
      "reward_std_all": 0.051488446071743965,
      "rewards-max/forensics_sinkhorn_reward": 0.9687137007713318,
      "rewards-max/forensics_strict_edge_reward": 0.37022389471530914,
      "rewards-max/forensics_tiou_reward": 0.9503158330917358,
      "rewards/forensics_sinkhorn_reward": 0.7533467411994934,
      "rewards/forensics_strict_edge_reward": 0.056719355285167694,
      "rewards/forensics_tiou_reward": 0.5989727079868317,
      "step": 715
    },
    {
      "completion_length": 26.046875,
      "epoch": 2.196620583717358,
      "grad_norm": 8.262236595153809,
      "kl": 0.4345703125,
      "learning_rate": 4.5e-07,
      "loss": -0.0486,
      "reward": 0.7994999289512634,
      "reward_std": 0.00829389609862119,
      "reward_std_all": 0.009706131415441632,
      "rewards-max/forensics_sinkhorn_reward": 0.9726937413215637,
      "rewards-max/forensics_strict_edge_reward": 0.4304373078048229,
      "rewards-max/forensics_tiou_reward": 0.9619491696357727,
      "rewards/forensics_sinkhorn_reward": 0.7863399386405945,
      "rewards/forensics_strict_edge_reward": 0.0732333823107183,
      "rewards/forensics_tiou_reward": 0.6630579233169556,
      "step": 716
    },
    {
      "completion_length": 26.328125,
      "epoch": 2.1996927803379416,
      "grad_norm": 6.088972568511963,
      "kl": 0.4931640625,
      "learning_rate": 4.492307692307692e-07,
      "loss": -0.041,
      "reward": 0.7877046167850494,
      "reward_std": 0.013814244884997606,
      "reward_std_all": 0.017819581553339958,
      "rewards-max/forensics_sinkhorn_reward": 0.9776821434497833,
      "rewards-max/forensics_strict_edge_reward": 0.49357786774635315,
      "rewards-max/forensics_tiou_reward": 0.9699394702911377,
      "rewards/forensics_sinkhorn_reward": 0.7672595381736755,
      "rewards/forensics_strict_edge_reward": 0.06965701654553413,
      "rewards/forensics_tiou_reward": 0.6201310157775879,
      "step": 717
    },
    {
      "completion_length": 23.84375,
      "epoch": 2.2027649769585254,
      "grad_norm": 8.591289520263672,
      "kl": 0.3974609375,
      "learning_rate": 4.4846153846153846e-07,
      "loss": 0.0349,
      "reward": 0.7561661005020142,
      "reward_std": 0.057063568383455276,
      "reward_std_all": 0.05642237886786461,
      "rewards-max/forensics_sinkhorn_reward": 0.987278401851654,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.9953853487968445,
      "rewards/forensics_sinkhorn_reward": 0.6978514492511749,
      "rewards/forensics_strict_edge_reward": 0.08930931985378265,
      "rewards/forensics_tiou_reward": 0.5124901533126831,
      "step": 718
    },
    {
      "completion_length": 28.15625,
      "epoch": 2.205837173579109,
      "grad_norm": 5.764827251434326,
      "kl": 0.43359375,
      "learning_rate": 4.476923076923076e-07,
      "loss": 0.007,
      "reward": 0.7805266678333282,
      "reward_std": 0.04053965210914612,
      "reward_std_all": 0.03749088291078806,
      "rewards-max/forensics_sinkhorn_reward": 0.9825976490974426,
      "rewards-max/forensics_strict_edge_reward": 0.5284900367259979,
      "rewards-max/forensics_tiou_reward": 0.981953501701355,
      "rewards/forensics_sinkhorn_reward": 0.7829378545284271,
      "rewards/forensics_strict_edge_reward": 0.07282141223549843,
      "rewards/forensics_tiou_reward": 0.6585833728313446,
      "step": 719
    },
    {
      "completion_length": 24.515625,
      "epoch": 2.208909370199693,
      "grad_norm": 5.808948040008545,
      "kl": 0.384765625,
      "learning_rate": 4.469230769230769e-07,
      "loss": 0.0409,
      "reward": 0.7569679915904999,
      "reward_std": 0.030503375455737114,
      "reward_std_all": 0.04014410264790058,
      "rewards-max/forensics_sinkhorn_reward": 0.9753202199935913,
      "rewards-max/forensics_strict_edge_reward": 0.594437800347805,
      "rewards-max/forensics_tiou_reward": 0.962850958108902,
      "rewards/forensics_sinkhorn_reward": 0.7278454899787903,
      "rewards/forensics_strict_edge_reward": 0.04150011017918587,
      "rewards/forensics_tiou_reward": 0.5332741439342499,
      "step": 720
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.2119815668202767,
      "grad_norm": 6.656505107879639,
      "kl": 0.384765625,
      "learning_rate": 4.4615384615384615e-07,
      "loss": -0.0164,
      "reward": 0.7814767956733704,
      "reward_std": 0.038304075598716736,
      "reward_std_all": 0.03574481047689915,
      "rewards-max/forensics_sinkhorn_reward": 0.9763091206550598,
      "rewards-max/forensics_strict_edge_reward": 0.533583253622055,
      "rewards-max/forensics_tiou_reward": 0.965651661157608,
      "rewards/forensics_sinkhorn_reward": 0.781903088092804,
      "rewards/forensics_strict_edge_reward": 0.05332434922456741,
      "rewards/forensics_tiou_reward": 0.6365457475185394,
      "step": 721
    },
    {
      "completion_length": 26.71875,
      "epoch": 2.21505376344086,
      "grad_norm": 5.5636820793151855,
      "kl": 0.357421875,
      "learning_rate": 4.4538461538461537e-07,
      "loss": 0.0824,
      "reward": 0.7687008082866669,
      "reward_std": 0.05157400295138359,
      "reward_std_all": 0.048033252358436584,
      "rewards-max/forensics_sinkhorn_reward": 0.9732188582420349,
      "rewards-max/forensics_strict_edge_reward": 0.3907022997736931,
      "rewards-max/forensics_tiou_reward": 0.9520735442638397,
      "rewards/forensics_sinkhorn_reward": 0.7550942599773407,
      "rewards/forensics_strict_edge_reward": 0.04381908942013979,
      "rewards/forensics_tiou_reward": 0.5701294839382172,
      "step": 722
    },
    {
      "completion_length": 28.5625,
      "epoch": 2.218125960061444,
      "grad_norm": 5.4342522621154785,
      "kl": 0.3369140625,
      "learning_rate": 4.4461538461538463e-07,
      "loss": -0.0163,
      "reward": 0.791738897562027,
      "reward_std": 0.018667510710656643,
      "reward_std_all": 0.018954175524413586,
      "rewards-max/forensics_sinkhorn_reward": 0.9799623787403107,
      "rewards-max/forensics_strict_edge_reward": 0.4899478554725647,
      "rewards-max/forensics_tiou_reward": 0.9775971174240112,
      "rewards/forensics_sinkhorn_reward": 0.7523944973945618,
      "rewards/forensics_strict_edge_reward": 0.062087275087833405,
      "rewards/forensics_tiou_reward": 0.652613639831543,
      "step": 723
    },
    {
      "completion_length": 21.1875,
      "epoch": 2.2211981566820276,
      "grad_norm": 9.528464317321777,
      "kl": 0.47265625,
      "learning_rate": 4.438461538461538e-07,
      "loss": 0.0541,
      "reward": 0.7568827569484711,
      "reward_std": 0.04869934730231762,
      "reward_std_all": 0.05083394702523947,
      "rewards-max/forensics_sinkhorn_reward": 0.9805446565151215,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9774567782878876,
      "rewards/forensics_sinkhorn_reward": 0.7305974960327148,
      "rewards/forensics_strict_edge_reward": 0.06463788077235222,
      "rewards/forensics_tiou_reward": 0.5521925687789917,
      "step": 724
    },
    {
      "completion_length": 27.546875,
      "epoch": 2.2242703533026114,
      "grad_norm": 4.868298530578613,
      "kl": 0.3095703125,
      "learning_rate": 4.4307692307692306e-07,
      "loss": -0.0378,
      "reward": 0.7916252613067627,
      "reward_std": 0.01838365849107504,
      "reward_std_all": 0.018758196267299354,
      "rewards-max/forensics_sinkhorn_reward": 0.9777829647064209,
      "rewards-max/forensics_strict_edge_reward": 0.4661732092499733,
      "rewards-max/forensics_tiou_reward": 0.9653541147708893,
      "rewards/forensics_sinkhorn_reward": 0.7658927142620087,
      "rewards/forensics_strict_edge_reward": 0.056367010809481144,
      "rewards/forensics_tiou_reward": 0.6332631707191467,
      "step": 725
    },
    {
      "completion_length": 28.78125,
      "epoch": 2.227342549923195,
      "grad_norm": 6.300488471984863,
      "kl": 0.4189453125,
      "learning_rate": 4.423076923076923e-07,
      "loss": 0.0748,
      "reward": 0.7871401906013489,
      "reward_std": 0.02233736962080002,
      "reward_std_all": 0.022570776753127575,
      "rewards-max/forensics_sinkhorn_reward": 0.9841409921646118,
      "rewards-max/forensics_strict_edge_reward": 0.4568247050046921,
      "rewards-max/forensics_tiou_reward": 0.9855461120605469,
      "rewards/forensics_sinkhorn_reward": 0.7547836303710938,
      "rewards/forensics_strict_edge_reward": 0.06771190464496613,
      "rewards/forensics_tiou_reward": 0.5639396905899048,
      "step": 726
    },
    {
      "completion_length": 23.109375,
      "epoch": 2.230414746543779,
      "grad_norm": 7.112868785858154,
      "kl": 0.55078125,
      "learning_rate": 4.4153846153846154e-07,
      "loss": 0.134,
      "reward": 0.7333895564079285,
      "reward_std": 0.06847002357244492,
      "reward_std_all": 0.07355039566755295,
      "rewards-max/forensics_sinkhorn_reward": 0.9589241147041321,
      "rewards-max/forensics_strict_edge_reward": 0.5289340615272522,
      "rewards-max/forensics_tiou_reward": 0.9283121228218079,
      "rewards/forensics_sinkhorn_reward": 0.6909130811691284,
      "rewards/forensics_strict_edge_reward": 0.03519195318222046,
      "rewards/forensics_tiou_reward": 0.46885502338409424,
      "step": 727
    },
    {
      "completion_length": 22.359375,
      "epoch": 2.2334869431643627,
      "grad_norm": 6.445511817932129,
      "kl": 0.4169921875,
      "learning_rate": 4.407692307692307e-07,
      "loss": -0.0531,
      "reward": 0.7927368879318237,
      "reward_std": 0.02282602881314233,
      "reward_std_all": 0.021980069810524583,
      "rewards-max/forensics_sinkhorn_reward": 0.971198320388794,
      "rewards-max/forensics_strict_edge_reward": 0.5755647718906403,
      "rewards-max/forensics_tiou_reward": 0.9543934166431427,
      "rewards/forensics_sinkhorn_reward": 0.7921417057514191,
      "rewards/forensics_strict_edge_reward": 0.07370295748114586,
      "rewards/forensics_tiou_reward": 0.5963520705699921,
      "step": 728
    },
    {
      "completion_length": 29.75,
      "epoch": 2.236559139784946,
      "grad_norm": 5.438632011413574,
      "kl": 0.3583984375,
      "learning_rate": 4.3999999999999997e-07,
      "loss": 0.0655,
      "reward": 0.7172223627567291,
      "reward_std": 0.053519947454333305,
      "reward_std_all": 0.07207798399031162,
      "rewards-max/forensics_sinkhorn_reward": 0.965625524520874,
      "rewards-max/forensics_strict_edge_reward": 0.3836943060159683,
      "rewards-max/forensics_tiou_reward": 0.947312980890274,
      "rewards/forensics_sinkhorn_reward": 0.6460642218589783,
      "rewards/forensics_strict_edge_reward": 0.05999170243740082,
      "rewards/forensics_tiou_reward": 0.520659863948822,
      "step": 729
    },
    {
      "completion_length": 24.828125,
      "epoch": 2.23963133640553,
      "grad_norm": 5.437568187713623,
      "kl": 0.4453125,
      "learning_rate": 4.3923076923076924e-07,
      "loss": 0.0863,
      "reward": 0.767869234085083,
      "reward_std": 0.04513349011540413,
      "reward_std_all": 0.04378102906048298,
      "rewards-max/forensics_sinkhorn_reward": 0.9791643619537354,
      "rewards-max/forensics_strict_edge_reward": 0.5532653331756592,
      "rewards-max/forensics_tiou_reward": 0.9729883968830109,
      "rewards/forensics_sinkhorn_reward": 0.7092185616493225,
      "rewards/forensics_strict_edge_reward": 0.049484895542263985,
      "rewards/forensics_tiou_reward": 0.542779952287674,
      "step": 730
    },
    {
      "completion_length": 25.125,
      "epoch": 2.2427035330261136,
      "grad_norm": 7.096614360809326,
      "kl": 0.39453125,
      "learning_rate": 4.3846153846153845e-07,
      "loss": 0.0948,
      "reward": 0.7160183489322662,
      "reward_std": 0.0693739727139473,
      "reward_std_all": 0.07535610720515251,
      "rewards-max/forensics_sinkhorn_reward": 0.9647443294525146,
      "rewards-max/forensics_strict_edge_reward": 0.6174952685832977,
      "rewards-max/forensics_tiou_reward": 0.9454127252101898,
      "rewards/forensics_sinkhorn_reward": 0.6516163647174835,
      "rewards/forensics_strict_edge_reward": 0.04812822677195072,
      "rewards/forensics_tiou_reward": 0.47253143787384033,
      "step": 731
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.2457757296466974,
      "grad_norm": 6.561961650848389,
      "kl": 0.46875,
      "learning_rate": 4.376923076923077e-07,
      "loss": 0.0674,
      "reward": 0.735670268535614,
      "reward_std": 0.05359142646193504,
      "reward_std_all": 0.0631584357470274,
      "rewards-max/forensics_sinkhorn_reward": 0.9758685529232025,
      "rewards-max/forensics_strict_edge_reward": 0.37814464420080185,
      "rewards-max/forensics_tiou_reward": 0.9599629044532776,
      "rewards/forensics_sinkhorn_reward": 0.6497967541217804,
      "rewards/forensics_strict_edge_reward": 0.042396734934300184,
      "rewards/forensics_tiou_reward": 0.4633451998233795,
      "step": 732
    },
    {
      "completion_length": 23.21875,
      "epoch": 2.248847926267281,
      "grad_norm": 5.711012840270996,
      "kl": 0.3984375,
      "learning_rate": 4.369230769230769e-07,
      "loss": -0.0015,
      "reward": 0.7678163647651672,
      "reward_std": 0.02866026759147644,
      "reward_std_all": 0.036130569875240326,
      "rewards-max/forensics_sinkhorn_reward": 0.9777842462062836,
      "rewards-max/forensics_strict_edge_reward": 0.3954024575650692,
      "rewards-max/forensics_tiou_reward": 0.9736299216747284,
      "rewards/forensics_sinkhorn_reward": 0.7140125632286072,
      "rewards/forensics_strict_edge_reward": 0.049473707331344485,
      "rewards/forensics_tiou_reward": 0.546402096748352,
      "step": 733
    },
    {
      "completion_length": 24.9375,
      "epoch": 2.251920122887865,
      "grad_norm": 8.382305145263672,
      "kl": 0.3642578125,
      "learning_rate": 4.3615384615384614e-07,
      "loss": -0.013,
      "reward": 0.7382820546627045,
      "reward_std": 0.050497455755248666,
      "reward_std_all": 0.05633561441209167,
      "rewards-max/forensics_sinkhorn_reward": 0.9836497008800507,
      "rewards-max/forensics_strict_edge_reward": 0.7578510344028473,
      "rewards-max/forensics_tiou_reward": 0.9862297773361206,
      "rewards/forensics_sinkhorn_reward": 0.6874217391014099,
      "rewards/forensics_strict_edge_reward": 0.0700339786708355,
      "rewards/forensics_tiou_reward": 0.5191773176193237,
      "step": 734
    },
    {
      "completion_length": 26.703125,
      "epoch": 2.2549923195084487,
      "grad_norm": 6.558043479919434,
      "kl": 0.3427734375,
      "learning_rate": 4.3538461538461536e-07,
      "loss": -0.0728,
      "reward": 0.7151886820793152,
      "reward_std": 0.02937194239348173,
      "reward_std_all": 0.06039882730692625,
      "rewards-max/forensics_sinkhorn_reward": 0.9707714915275574,
      "rewards-max/forensics_strict_edge_reward": 0.5097470879554749,
      "rewards-max/forensics_tiou_reward": 0.9550033211708069,
      "rewards/forensics_sinkhorn_reward": 0.7025669813156128,
      "rewards/forensics_strict_edge_reward": 0.05831548757851124,
      "rewards/forensics_tiou_reward": 0.5537860691547394,
      "step": 735
    },
    {
      "completion_length": 25.234375,
      "epoch": 2.258064516129032,
      "grad_norm": 7.01978063583374,
      "kl": 0.33203125,
      "learning_rate": 4.346153846153846e-07,
      "loss": -0.0367,
      "reward": 0.7915603518486023,
      "reward_std": 0.012979283928871155,
      "reward_std_all": 0.01593217719346285,
      "rewards-max/forensics_sinkhorn_reward": 0.9858063757419586,
      "rewards-max/forensics_strict_edge_reward": 0.7094893157482147,
      "rewards-max/forensics_tiou_reward": 0.9899512529373169,
      "rewards/forensics_sinkhorn_reward": 0.7539312243461609,
      "rewards/forensics_strict_edge_reward": 0.059426199644804,
      "rewards/forensics_tiou_reward": 0.6042097210884094,
      "step": 736
    },
    {
      "completion_length": 25.875,
      "epoch": 2.261136712749616,
      "grad_norm": 7.902054786682129,
      "kl": 0.32763671875,
      "learning_rate": 4.338461538461538e-07,
      "loss": 0.0101,
      "reward": 0.7632349729537964,
      "reward_std": 0.028478778898715973,
      "reward_std_all": 0.034446386620402336,
      "rewards-max/forensics_sinkhorn_reward": 0.9731391370296478,
      "rewards-max/forensics_strict_edge_reward": 0.4724321663379669,
      "rewards-max/forensics_tiou_reward": 0.9589546024799347,
      "rewards/forensics_sinkhorn_reward": 0.7113783955574036,
      "rewards/forensics_strict_edge_reward": 0.04547343589365482,
      "rewards/forensics_tiou_reward": 0.5422686040401459,
      "step": 737
    },
    {
      "completion_length": 29.046875,
      "epoch": 2.2642089093701996,
      "grad_norm": 5.5534162521362305,
      "kl": 0.42578125,
      "learning_rate": 4.3307692307692305e-07,
      "loss": -0.0675,
      "reward": 0.7178319990634918,
      "reward_std": 0.016038172179833055,
      "reward_std_all": 0.05183953791856766,
      "rewards-max/forensics_sinkhorn_reward": 0.9755763709545135,
      "rewards-max/forensics_strict_edge_reward": 0.4599933922290802,
      "rewards-max/forensics_tiou_reward": 0.9623980820178986,
      "rewards/forensics_sinkhorn_reward": 0.6756385564804077,
      "rewards/forensics_strict_edge_reward": 0.05795428901910782,
      "rewards/forensics_tiou_reward": 0.5542003810405731,
      "step": 738
    },
    {
      "completion_length": 24.5625,
      "epoch": 2.2672811059907834,
      "grad_norm": 6.0001959800720215,
      "kl": 0.4541015625,
      "learning_rate": 4.3230769230769227e-07,
      "loss": 0.0531,
      "reward": 0.755159467458725,
      "reward_std": 0.05958430655300617,
      "reward_std_all": 0.05862043425440788,
      "rewards-max/forensics_sinkhorn_reward": 0.970994621515274,
      "rewards-max/forensics_strict_edge_reward": 0.4357372522354126,
      "rewards-max/forensics_tiou_reward": 0.9593773186206818,
      "rewards/forensics_sinkhorn_reward": 0.6695126891136169,
      "rewards/forensics_strict_edge_reward": 0.03239305317401886,
      "rewards/forensics_tiou_reward": 0.5087318122386932,
      "step": 739
    },
    {
      "completion_length": 23.875,
      "epoch": 2.270353302611367,
      "grad_norm": 6.304168224334717,
      "kl": 0.3037109375,
      "learning_rate": 4.3153846153846153e-07,
      "loss": 0.0432,
      "reward": 0.7968638837337494,
      "reward_std": 0.014202549471519887,
      "reward_std_all": 0.01478186808526516,
      "rewards-max/forensics_sinkhorn_reward": 0.9666239023208618,
      "rewards-max/forensics_strict_edge_reward": 0.2449169121682644,
      "rewards-max/forensics_tiou_reward": 0.9447180032730103,
      "rewards/forensics_sinkhorn_reward": 0.7870893478393555,
      "rewards/forensics_strict_edge_reward": 0.045273247407749295,
      "rewards/forensics_tiou_reward": 0.5860334634780884,
      "step": 740
    },
    {
      "completion_length": 25.390625,
      "epoch": 2.273425499231951,
      "grad_norm": 7.41320276260376,
      "kl": 0.69140625,
      "learning_rate": 4.307692307692308e-07,
      "loss": 0.0548,
      "reward": 0.7867611646652222,
      "reward_std": 0.028788088355213404,
      "reward_std_all": 0.02761638630181551,
      "rewards-max/forensics_sinkhorn_reward": 0.9595091938972473,
      "rewards-max/forensics_strict_edge_reward": 0.6133978515863419,
      "rewards-max/forensics_tiou_reward": 0.9338657855987549,
      "rewards/forensics_sinkhorn_reward": 0.7300238013267517,
      "rewards/forensics_strict_edge_reward": 0.03574374224990606,
      "rewards/forensics_tiou_reward": 0.5348232686519623,
      "step": 741
    },
    {
      "completion_length": 24.65625,
      "epoch": 2.2764976958525347,
      "grad_norm": 6.171875476837158,
      "kl": 0.3916015625,
      "learning_rate": 4.2999999999999996e-07,
      "loss": 0.0964,
      "reward": 0.7411514222621918,
      "reward_std": 0.03032756457105279,
      "reward_std_all": 0.048034096136689186,
      "rewards-max/forensics_sinkhorn_reward": 0.9623628556728363,
      "rewards-max/forensics_strict_edge_reward": 0.3723905999213457,
      "rewards-max/forensics_tiou_reward": 0.9413988292217255,
      "rewards/forensics_sinkhorn_reward": 0.6826887428760529,
      "rewards/forensics_strict_edge_reward": 0.01800367096439004,
      "rewards/forensics_tiou_reward": 0.5033731311559677,
      "step": 742
    },
    {
      "completion_length": 23.734375,
      "epoch": 2.279569892473118,
      "grad_norm": 7.837794303894043,
      "kl": 0.380859375,
      "learning_rate": 4.292307692307692e-07,
      "loss": 0.0107,
      "reward": 0.7859037220478058,
      "reward_std": 0.026523126289248466,
      "reward_std_all": 0.026320738717913628,
      "rewards-max/forensics_sinkhorn_reward": 0.9662739932537079,
      "rewards-max/forensics_strict_edge_reward": 0.38191211223602295,
      "rewards-max/forensics_tiou_reward": 0.944808304309845,
      "rewards/forensics_sinkhorn_reward": 0.7563915550708771,
      "rewards/forensics_strict_edge_reward": 0.04283926449716091,
      "rewards/forensics_tiou_reward": 0.5090727508068085,
      "step": 743
    },
    {
      "completion_length": 22.703125,
      "epoch": 2.282642089093702,
      "grad_norm": 6.793100357055664,
      "kl": 0.4521484375,
      "learning_rate": 4.2846153846153844e-07,
      "loss": 0.1071,
      "reward": 0.7650095820426941,
      "reward_std": 0.04049805551767349,
      "reward_std_all": 0.04340881109237671,
      "rewards-max/forensics_sinkhorn_reward": 0.9696285724639893,
      "rewards-max/forensics_strict_edge_reward": 0.4322323799133301,
      "rewards-max/forensics_tiou_reward": 0.9553960561752319,
      "rewards/forensics_sinkhorn_reward": 0.6904313862323761,
      "rewards/forensics_strict_edge_reward": 0.02883538417518139,
      "rewards/forensics_tiou_reward": 0.4706008583307266,
      "step": 744
    },
    {
      "completion_length": 23.5625,
      "epoch": 2.2857142857142856,
      "grad_norm": 6.670623302459717,
      "kl": 0.4423828125,
      "learning_rate": 4.276923076923077e-07,
      "loss": -0.0691,
      "reward": 0.7282392978668213,
      "reward_std": 0.03520653024315834,
      "reward_std_all": 0.05978449247777462,
      "rewards-max/forensics_sinkhorn_reward": 0.9823698997497559,
      "rewards-max/forensics_strict_edge_reward": 0.364276260137558,
      "rewards-max/forensics_tiou_reward": 0.9826718270778656,
      "rewards/forensics_sinkhorn_reward": 0.6517572402954102,
      "rewards/forensics_strict_edge_reward": 0.020604433491826057,
      "rewards/forensics_tiou_reward": 0.4458530843257904,
      "step": 745
    },
    {
      "completion_length": 28.6875,
      "epoch": 2.2887864823348694,
      "grad_norm": 6.187060832977295,
      "kl": 0.4072265625,
      "learning_rate": 4.2692307692307687e-07,
      "loss": 0.0702,
      "reward": 0.7751272022724152,
      "reward_std": 0.037098757922649384,
      "reward_std_all": 0.03660101443529129,
      "rewards-max/forensics_sinkhorn_reward": 0.9712158441543579,
      "rewards-max/forensics_strict_edge_reward": 0.402368500828743,
      "rewards-max/forensics_tiou_reward": 0.9560469090938568,
      "rewards/forensics_sinkhorn_reward": 0.7119672298431396,
      "rewards/forensics_strict_edge_reward": 0.0339598823338747,
      "rewards/forensics_tiou_reward": 0.5621523708105087,
      "step": 746
    },
    {
      "completion_length": 24.640625,
      "epoch": 2.291858678955453,
      "grad_norm": 6.2274370193481445,
      "kl": 0.3427734375,
      "learning_rate": 4.2615384615384613e-07,
      "loss": 0.0334,
      "reward": 0.7573534548282623,
      "reward_std": 0.030084888450801373,
      "reward_std_all": 0.040964774787425995,
      "rewards-max/forensics_sinkhorn_reward": 0.9570081532001495,
      "rewards-max/forensics_strict_edge_reward": 0.14539159461855888,
      "rewards-max/forensics_tiou_reward": 0.9189859926700592,
      "rewards/forensics_sinkhorn_reward": 0.727016270160675,
      "rewards/forensics_strict_edge_reward": 0.017728411592543125,
      "rewards/forensics_tiou_reward": 0.5458424687385559,
      "step": 747
    },
    {
      "completion_length": 24.15625,
      "epoch": 2.294930875576037,
      "grad_norm": 7.971622943878174,
      "kl": 0.341796875,
      "learning_rate": 4.2538461538461535e-07,
      "loss": 0.0942,
      "reward": 0.7571305632591248,
      "reward_std": 0.03488034475594759,
      "reward_std_all": 0.04402429983019829,
      "rewards-max/forensics_sinkhorn_reward": 0.9604544341564178,
      "rewards-max/forensics_strict_edge_reward": 0.23678485862910748,
      "rewards-max/forensics_tiou_reward": 0.9269954860210419,
      "rewards/forensics_sinkhorn_reward": 0.7045678794384003,
      "rewards/forensics_strict_edge_reward": 0.019770185463130474,
      "rewards/forensics_tiou_reward": 0.5177553594112396,
      "step": 748
    },
    {
      "completion_length": 22.421875,
      "epoch": 2.2980030721966207,
      "grad_norm": 5.388754844665527,
      "kl": 0.4794921875,
      "learning_rate": 4.246153846153846e-07,
      "loss": 0.1069,
      "reward": 0.7498691976070404,
      "reward_std": 0.07314616441726685,
      "reward_std_all": 0.06838703900575638,
      "rewards-max/forensics_sinkhorn_reward": 0.9298222661018372,
      "rewards-max/forensics_strict_edge_reward": 0.38386269845068455,
      "rewards-max/forensics_tiou_reward": 0.8651913404464722,
      "rewards/forensics_sinkhorn_reward": 0.6936195492744446,
      "rewards/forensics_strict_edge_reward": 0.05188327236101031,
      "rewards/forensics_tiou_reward": 0.5237247496843338,
      "step": 749
    },
    {
      "completion_length": 26.625,
      "epoch": 2.3010752688172045,
      "grad_norm": 5.843987464904785,
      "kl": 0.431640625,
      "learning_rate": 4.2384615384615383e-07,
      "loss": 0.1108,
      "reward": 0.7791863977909088,
      "reward_std": 0.018493880284950137,
      "reward_std_all": 0.02467432152479887,
      "rewards-max/forensics_sinkhorn_reward": 0.9718137681484222,
      "rewards-max/forensics_strict_edge_reward": 0.17728428915143013,
      "rewards-max/forensics_tiou_reward": 0.9393729269504547,
      "rewards/forensics_sinkhorn_reward": 0.7268509864807129,
      "rewards/forensics_strict_edge_reward": 0.016685456037521362,
      "rewards/forensics_tiou_reward": 0.5665504038333893,
      "step": 750
    },
    {
      "completion_length": 27.25,
      "epoch": 2.3041474654377883,
      "grad_norm": 7.105732440948486,
      "kl": 0.4052734375,
      "learning_rate": 4.2307692307692304e-07,
      "loss": 0.1328,
      "reward": 0.7429733276367188,
      "reward_std": 0.06183583475649357,
      "reward_std_all": 0.06249236688017845,
      "rewards-max/forensics_sinkhorn_reward": 0.957717627286911,
      "rewards-max/forensics_strict_edge_reward": 0.4167146235704422,
      "rewards-max/forensics_tiou_reward": 0.9260204434394836,
      "rewards/forensics_sinkhorn_reward": 0.7105075716972351,
      "rewards/forensics_strict_edge_reward": 0.04603066295385361,
      "rewards/forensics_tiou_reward": 0.5582882016897202,
      "step": 751
    },
    {
      "completion_length": 25.140625,
      "epoch": 2.3072196620583716,
      "grad_norm": 6.373408794403076,
      "kl": 0.40625,
      "learning_rate": 4.223076923076923e-07,
      "loss": -0.0103,
      "reward": 0.7900071442127228,
      "reward_std": 0.024196932092308998,
      "reward_std_all": 0.023479143157601357,
      "rewards-max/forensics_sinkhorn_reward": 0.961463451385498,
      "rewards-max/forensics_strict_edge_reward": 0.34398502856492996,
      "rewards-max/forensics_tiou_reward": 0.9295450747013092,
      "rewards/forensics_sinkhorn_reward": 0.7606014907360077,
      "rewards/forensics_strict_edge_reward": 0.02918953262269497,
      "rewards/forensics_tiou_reward": 0.6022010147571564,
      "step": 752
    },
    {
      "completion_length": 23.140625,
      "epoch": 2.3102918586789554,
      "grad_norm": 8.465609550476074,
      "kl": 0.6044921875,
      "learning_rate": 4.215384615384615e-07,
      "loss": -0.0468,
      "reward": 0.7557233273983002,
      "reward_std": 0.03272359608672559,
      "reward_std_all": 0.04072472336702049,
      "rewards-max/forensics_sinkhorn_reward": 0.9500429928302765,
      "rewards-max/forensics_strict_edge_reward": 0.45358169078826904,
      "rewards-max/forensics_tiou_reward": 0.9067034721374512,
      "rewards/forensics_sinkhorn_reward": 0.6993069648742676,
      "rewards/forensics_strict_edge_reward": 0.04297730978578329,
      "rewards/forensics_tiou_reward": 0.47046077251434326,
      "step": 753
    },
    {
      "completion_length": 24.671875,
      "epoch": 2.313364055299539,
      "grad_norm": 7.273371696472168,
      "kl": 0.3466796875,
      "learning_rate": 4.207692307692308e-07,
      "loss": 0.0051,
      "reward": 0.7801373898983002,
      "reward_std": 0.03921466879546642,
      "reward_std_all": 0.03654642216861248,
      "rewards-max/forensics_sinkhorn_reward": 0.9630723893642426,
      "rewards-max/forensics_strict_edge_reward": 0.6791328340768814,
      "rewards-max/forensics_tiou_reward": 0.9305214285850525,
      "rewards/forensics_sinkhorn_reward": 0.7437945604324341,
      "rewards/forensics_strict_edge_reward": 0.06881086062639952,
      "rewards/forensics_tiou_reward": 0.5368698835372925,
      "step": 754
    },
    {
      "completion_length": 22.703125,
      "epoch": 2.316436251920123,
      "grad_norm": 11.2376070022583,
      "kl": 0.3564453125,
      "learning_rate": 4.1999999999999995e-07,
      "loss": 0.0337,
      "reward": 0.7491464018821716,
      "reward_std": 0.05152638256549835,
      "reward_std_all": 0.05595949664711952,
      "rewards-max/forensics_sinkhorn_reward": 0.9645367562770844,
      "rewards-max/forensics_strict_edge_reward": 0.4361710250377655,
      "rewards-max/forensics_tiou_reward": 0.9425612688064575,
      "rewards/forensics_sinkhorn_reward": 0.6515766382217407,
      "rewards/forensics_strict_edge_reward": 0.04864950478076935,
      "rewards/forensics_tiou_reward": 0.4722703993320465,
      "step": 755
    },
    {
      "completion_length": 26.03125,
      "epoch": 2.3195084485407067,
      "grad_norm": 6.627137184143066,
      "kl": 0.517578125,
      "learning_rate": 4.192307692307692e-07,
      "loss": 0.0998,
      "reward": 0.7788823843002319,
      "reward_std": 0.03153357491828501,
      "reward_std_all": 0.03134468081407249,
      "rewards-max/forensics_sinkhorn_reward": 0.9719153642654419,
      "rewards-max/forensics_strict_edge_reward": 0.5238595008850098,
      "rewards-max/forensics_tiou_reward": 0.9603115618228912,
      "rewards/forensics_sinkhorn_reward": 0.7245675325393677,
      "rewards/forensics_strict_edge_reward": 0.06616299599409103,
      "rewards/forensics_tiou_reward": 0.5904570519924164,
      "step": 756
    },
    {
      "completion_length": 25.34375,
      "epoch": 2.3225806451612905,
      "grad_norm": 7.900700092315674,
      "kl": 0.5146484375,
      "learning_rate": 4.1846153846153843e-07,
      "loss": 0.0724,
      "reward": 0.7426805198192596,
      "reward_std": 0.060880282893776894,
      "reward_std_all": 0.06500444561243057,
      "rewards-max/forensics_sinkhorn_reward": 0.9775318801403046,
      "rewards-max/forensics_strict_edge_reward": 0.29750871658325195,
      "rewards-max/forensics_tiou_reward": 0.9671319127082825,
      "rewards/forensics_sinkhorn_reward": 0.702347606420517,
      "rewards/forensics_strict_edge_reward": 0.024172832258045673,
      "rewards/forensics_tiou_reward": 0.5050095617771149,
      "step": 757
    },
    {
      "completion_length": 26.0625,
      "epoch": 2.3256528417818743,
      "grad_norm": 7.306189060211182,
      "kl": 0.4072265625,
      "learning_rate": 4.176923076923077e-07,
      "loss": 0.0424,
      "reward": 0.7673728168010712,
      "reward_std": 0.04184718104079366,
      "reward_std_all": 0.04184638662263751,
      "rewards-max/forensics_sinkhorn_reward": 0.9803731441497803,
      "rewards-max/forensics_strict_edge_reward": 0.5396573096513748,
      "rewards-max/forensics_tiou_reward": 0.9792531132698059,
      "rewards/forensics_sinkhorn_reward": 0.7248961925506592,
      "rewards/forensics_strict_edge_reward": 0.06213845871388912,
      "rewards/forensics_tiou_reward": 0.5614364743232727,
      "step": 758
    },
    {
      "completion_length": 24.578125,
      "epoch": 2.3287250384024576,
      "grad_norm": 7.013620853424072,
      "kl": 0.384765625,
      "learning_rate": 4.1692307692307686e-07,
      "loss": 0.037,
      "reward": 0.7884140908718109,
      "reward_std": 0.02065748255699873,
      "reward_std_all": 0.021429527550935745,
      "rewards-max/forensics_sinkhorn_reward": 0.9321179986000061,
      "rewards-max/forensics_strict_edge_reward": 0.17760160192847252,
      "rewards-max/forensics_tiou_reward": 0.8665570318698883,
      "rewards/forensics_sinkhorn_reward": 0.7403577268123627,
      "rewards/forensics_strict_edge_reward": 0.011177344247698784,
      "rewards/forensics_tiou_reward": 0.4910877197980881,
      "step": 759
    },
    {
      "completion_length": 25.78125,
      "epoch": 2.3317972350230414,
      "grad_norm": 6.408485412597656,
      "kl": 0.4609375,
      "learning_rate": 4.161538461538461e-07,
      "loss": 0.1553,
      "reward": 0.7589548230171204,
      "reward_std": 0.043380213901400566,
      "reward_std_all": 0.044516537338495255,
      "rewards-max/forensics_sinkhorn_reward": 0.9883145987987518,
      "rewards-max/forensics_strict_edge_reward": 0.835382878780365,
      "rewards-max/forensics_tiou_reward": 0.9984047412872314,
      "rewards/forensics_sinkhorn_reward": 0.6872001886367798,
      "rewards/forensics_strict_edge_reward": 0.11099071055650711,
      "rewards/forensics_tiou_reward": 0.5558187067508698,
      "step": 760
    },
    {
      "completion_length": 26.703125,
      "epoch": 2.334869431643625,
      "grad_norm": 5.246341705322266,
      "kl": 0.4169921875,
      "learning_rate": 4.153846153846154e-07,
      "loss": 0.2018,
      "reward": 0.7370559871196747,
      "reward_std": 0.08813337609171867,
      "reward_std_all": 0.0821937844157219,
      "rewards-max/forensics_sinkhorn_reward": 0.9708910286426544,
      "rewards-max/forensics_strict_edge_reward": 0.4680478796362877,
      "rewards-max/forensics_tiou_reward": 0.9534582793712616,
      "rewards/forensics_sinkhorn_reward": 0.7001531720161438,
      "rewards/forensics_strict_edge_reward": 0.03455750085413456,
      "rewards/forensics_tiou_reward": 0.470846027135849,
      "step": 761
    },
    {
      "completion_length": 24.6875,
      "epoch": 2.337941628264209,
      "grad_norm": 5.298004150390625,
      "kl": 0.384765625,
      "learning_rate": 4.146153846153846e-07,
      "loss": 0.1065,
      "reward": 0.7503692507743835,
      "reward_std": 0.043079935014247894,
      "reward_std_all": 0.051479130052030087,
      "rewards-max/forensics_sinkhorn_reward": 0.9784089922904968,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9739598035812378,
      "rewards/forensics_sinkhorn_reward": 0.7099977433681488,
      "rewards/forensics_strict_edge_reward": 0.054722173139452934,
      "rewards/forensics_tiou_reward": 0.562647670507431,
      "step": 762
    },
    {
      "completion_length": 22.734375,
      "epoch": 2.3410138248847927,
      "grad_norm": 5.999513149261475,
      "kl": 0.333984375,
      "learning_rate": 4.1384615384615387e-07,
      "loss": 0.0426,
      "reward": 0.7788205742835999,
      "reward_std": 0.029687007889151573,
      "reward_std_all": 0.030076720751821995,
      "rewards-max/forensics_sinkhorn_reward": 0.9686141908168793,
      "rewards-max/forensics_strict_edge_reward": 0.15853514336049557,
      "rewards-max/forensics_tiou_reward": 0.9461611807346344,
      "rewards/forensics_sinkhorn_reward": 0.7541840672492981,
      "rewards/forensics_strict_edge_reward": 0.01508915494196117,
      "rewards/forensics_tiou_reward": 0.5991677641868591,
      "step": 763
    },
    {
      "completion_length": 26.703125,
      "epoch": 2.3440860215053765,
      "grad_norm": 5.901421070098877,
      "kl": 0.3583984375,
      "learning_rate": 4.1307692307692303e-07,
      "loss": 0.003,
      "reward": 0.7606759369373322,
      "reward_std": 0.041582899168133736,
      "reward_std_all": 0.042855534702539444,
      "rewards-max/forensics_sinkhorn_reward": 0.9688752889633179,
      "rewards-max/forensics_strict_edge_reward": 0.5089065283536911,
      "rewards-max/forensics_tiou_reward": 0.9479415714740753,
      "rewards/forensics_sinkhorn_reward": 0.6756750345230103,
      "rewards/forensics_strict_edge_reward": 0.059242261573672295,
      "rewards/forensics_tiou_reward": 0.57387974858284,
      "step": 764
    },
    {
      "completion_length": 24.5625,
      "epoch": 2.3471582181259603,
      "grad_norm": 5.189489364624023,
      "kl": 0.4677734375,
      "learning_rate": 4.123076923076923e-07,
      "loss": 0.0704,
      "reward": 0.7765743434429169,
      "reward_std": 0.013325230684131384,
      "reward_std_all": 0.023296697065234184,
      "rewards-max/forensics_sinkhorn_reward": 0.9822808504104614,
      "rewards-max/forensics_strict_edge_reward": 0.5132913738489151,
      "rewards-max/forensics_tiou_reward": 0.9776737987995148,
      "rewards/forensics_sinkhorn_reward": 0.7635300755500793,
      "rewards/forensics_strict_edge_reward": 0.0406820559874177,
      "rewards/forensics_tiou_reward": 0.5775569081306458,
      "step": 765
    },
    {
      "completion_length": 21.4375,
      "epoch": 2.3502304147465436,
      "grad_norm": 7.678463935852051,
      "kl": 0.4169921875,
      "learning_rate": 4.115384615384615e-07,
      "loss": -0.057,
      "reward": 0.7749208509922028,
      "reward_std": 0.035734571516513824,
      "reward_std_all": 0.03540355525910854,
      "rewards-max/forensics_sinkhorn_reward": 0.9777452945709229,
      "rewards-max/forensics_strict_edge_reward": 0.6741649508476257,
      "rewards-max/forensics_tiou_reward": 0.9723503887653351,
      "rewards/forensics_sinkhorn_reward": 0.7657004296779633,
      "rewards/forensics_strict_edge_reward": 0.08822494000196457,
      "rewards/forensics_tiou_reward": 0.5386427342891693,
      "step": 766
    },
    {
      "completion_length": 26.0625,
      "epoch": 2.3533026113671274,
      "grad_norm": 5.427978992462158,
      "kl": 0.4892578125,
      "learning_rate": 4.107692307692308e-07,
      "loss": 0.0245,
      "reward": 0.7903529703617096,
      "reward_std": 0.023800859693437815,
      "reward_std_all": 0.02306722616776824,
      "rewards-max/forensics_sinkhorn_reward": 0.9844934046268463,
      "rewards-max/forensics_strict_edge_reward": 0.5892573893070221,
      "rewards-max/forensics_tiou_reward": 0.9852820634841919,
      "rewards/forensics_sinkhorn_reward": 0.7625680267810822,
      "rewards/forensics_strict_edge_reward": 0.07514587417244911,
      "rewards/forensics_tiou_reward": 0.6182110905647278,
      "step": 767
    },
    {
      "completion_length": 23.984375,
      "epoch": 2.356374807987711,
      "grad_norm": 8.593097686767578,
      "kl": 0.716796875,
      "learning_rate": 4.0999999999999994e-07,
      "loss": -0.0199,
      "reward": 0.7824336290359497,
      "reward_std": 0.04018814116716385,
      "reward_std_all": 0.036910535767674446,
      "rewards-max/forensics_sinkhorn_reward": 0.9789687395095825,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9737599194049835,
      "rewards/forensics_sinkhorn_reward": 0.7615277767181396,
      "rewards/forensics_strict_edge_reward": 0.05369794741272926,
      "rewards/forensics_tiou_reward": 0.6156695783138275,
      "step": 768
    },
    {
      "completion_length": 25.140625,
      "epoch": 2.359447004608295,
      "grad_norm": 7.409857273101807,
      "kl": 0.365234375,
      "learning_rate": 4.092307692307692e-07,
      "loss": -0.0227,
      "reward": 0.8027742505073547,
      "reward_std": 0.0037581626093015075,
      "reward_std_all": 0.005884757032617927,
      "rewards-max/forensics_sinkhorn_reward": 0.9724408984184265,
      "rewards-max/forensics_strict_edge_reward": 0.6974508315324783,
      "rewards-max/forensics_tiou_reward": 0.9620094895362854,
      "rewards/forensics_sinkhorn_reward": 0.7112038135528564,
      "rewards/forensics_strict_edge_reward": 0.07286588475108147,
      "rewards/forensics_tiou_reward": 0.5904408097267151,
      "step": 769
    },
    {
      "completion_length": 25.765625,
      "epoch": 2.3625192012288787,
      "grad_norm": 5.664425373077393,
      "kl": 0.3671875,
      "learning_rate": 4.0846153846153847e-07,
      "loss": 0.0732,
      "reward": 0.799947202205658,
      "reward_std": 0.007686563767492771,
      "reward_std_all": 0.009075177367776632,
      "rewards-max/forensics_sinkhorn_reward": 0.9840812981128693,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9857354760169983,
      "rewards/forensics_sinkhorn_reward": 0.7537771165370941,
      "rewards/forensics_strict_edge_reward": 0.0656195841729641,
      "rewards/forensics_tiou_reward": 0.5804667472839355,
      "step": 770
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.3655913978494625,
      "grad_norm": 7.017309188842773,
      "kl": 0.3349609375,
      "learning_rate": 4.076923076923077e-07,
      "loss": -0.0576,
      "reward": 0.7702130079269409,
      "reward_std": 0.056810250505805016,
      "reward_std_all": 0.05176554247736931,
      "rewards-max/forensics_sinkhorn_reward": 0.958364725112915,
      "rewards-max/forensics_strict_edge_reward": 0.48555219173431396,
      "rewards-max/forensics_tiou_reward": 0.9239600300788879,
      "rewards/forensics_sinkhorn_reward": 0.7335274815559387,
      "rewards/forensics_strict_edge_reward": 0.030860764905810356,
      "rewards/forensics_tiou_reward": 0.5725512206554413,
      "step": 771
    },
    {
      "completion_length": 23.4375,
      "epoch": 2.3686635944700463,
      "grad_norm": 6.089517116546631,
      "kl": 0.4091796875,
      "learning_rate": 4.069230769230769e-07,
      "loss": 0.0083,
      "reward": 0.7948211133480072,
      "reward_std": 0.020173864322714508,
      "reward_std_all": 0.019255684921517968,
      "rewards-max/forensics_sinkhorn_reward": 0.9838680028915405,
      "rewards-max/forensics_strict_edge_reward": 0.6529513001441956,
      "rewards-max/forensics_tiou_reward": 0.9859558045864105,
      "rewards/forensics_sinkhorn_reward": 0.7885211408138275,
      "rewards/forensics_strict_edge_reward": 0.07571323961019516,
      "rewards/forensics_tiou_reward": 0.6558026075363159,
      "step": 772
    },
    {
      "completion_length": 27.1875,
      "epoch": 2.3717357910906296,
      "grad_norm": 6.080165863037109,
      "kl": 0.3515625,
      "learning_rate": 4.061538461538461e-07,
      "loss": -0.1261,
      "reward": 0.7645072340965271,
      "reward_std": 0.020067408389877528,
      "reward_std_all": 0.030455863918177783,
      "rewards-max/forensics_sinkhorn_reward": 0.976590096950531,
      "rewards-max/forensics_strict_edge_reward": 0.7639631927013397,
      "rewards-max/forensics_tiou_reward": 0.9706386923789978,
      "rewards/forensics_sinkhorn_reward": 0.7646628022193909,
      "rewards/forensics_strict_edge_reward": 0.08008560910820961,
      "rewards/forensics_tiou_reward": 0.5523113757371902,
      "step": 773
    },
    {
      "completion_length": 24.21875,
      "epoch": 2.3748079877112134,
      "grad_norm": 8.17920970916748,
      "kl": 0.32421875,
      "learning_rate": 4.053846153846154e-07,
      "loss": 0.0577,
      "reward": 0.7651422321796417,
      "reward_std": 0.05106595903635025,
      "reward_std_all": 0.04903717711567879,
      "rewards-max/forensics_sinkhorn_reward": 0.9675675332546234,
      "rewards-max/forensics_strict_edge_reward": 0.3116760104894638,
      "rewards-max/forensics_tiou_reward": 0.9424063265323639,
      "rewards/forensics_sinkhorn_reward": 0.7389780879020691,
      "rewards/forensics_strict_edge_reward": 0.031157067976891994,
      "rewards/forensics_tiou_reward": 0.561825305223465,
      "step": 774
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.377880184331797,
      "grad_norm": 4.880976676940918,
      "kl": 0.3486328125,
      "learning_rate": 4.046153846153846e-07,
      "loss": -0.0296,
      "reward": 0.7640979290008545,
      "reward_std": 0.04824026953428984,
      "reward_std_all": 0.049434637650847435,
      "rewards-max/forensics_sinkhorn_reward": 0.981723815202713,
      "rewards-max/forensics_strict_edge_reward": 0.5136521756649017,
      "rewards-max/forensics_tiou_reward": 0.9803282618522644,
      "rewards/forensics_sinkhorn_reward": 0.7550319731235504,
      "rewards/forensics_strict_edge_reward": 0.056309083476662636,
      "rewards/forensics_tiou_reward": 0.606590062379837,
      "step": 775
    },
    {
      "completion_length": 28.21875,
      "epoch": 2.380952380952381,
      "grad_norm": 19.061447143554688,
      "kl": 1.1298828125,
      "learning_rate": 4.0384615384615386e-07,
      "loss": 0.0606,
      "reward": 0.7832722663879395,
      "reward_std": 0.03929832857102156,
      "reward_std_all": 0.03585115959867835,
      "rewards-max/forensics_sinkhorn_reward": 0.9761016070842743,
      "rewards-max/forensics_strict_edge_reward": 0.38373464345932007,
      "rewards-max/forensics_tiou_reward": 0.96476149559021,
      "rewards/forensics_sinkhorn_reward": 0.7696205973625183,
      "rewards/forensics_strict_edge_reward": 0.035134103149175644,
      "rewards/forensics_tiou_reward": 0.684964656829834,
      "step": 776
    },
    {
      "completion_length": 26.28125,
      "epoch": 2.3840245775729647,
      "grad_norm": 5.298277378082275,
      "kl": 0.3857421875,
      "learning_rate": 4.03076923076923e-07,
      "loss": 0.0913,
      "reward": 0.772121250629425,
      "reward_std": 0.034410178661346436,
      "reward_std_all": 0.03497171960771084,
      "rewards-max/forensics_sinkhorn_reward": 0.9575859606266022,
      "rewards-max/forensics_strict_edge_reward": 0.30791324749588966,
      "rewards-max/forensics_tiou_reward": 0.9214225709438324,
      "rewards/forensics_sinkhorn_reward": 0.6960814297199249,
      "rewards/forensics_strict_edge_reward": 0.04257646203041077,
      "rewards/forensics_tiou_reward": 0.5108375549316406,
      "step": 777
    },
    {
      "completion_length": 26.609375,
      "epoch": 2.3870967741935485,
      "grad_norm": 5.328033447265625,
      "kl": 0.38671875,
      "learning_rate": 4.023076923076923e-07,
      "loss": 0.0004,
      "reward": 0.7709183096885681,
      "reward_std": 0.032701339572668076,
      "reward_std_all": 0.034078831784427166,
      "rewards-max/forensics_sinkhorn_reward": 0.9712007939815521,
      "rewards-max/forensics_strict_edge_reward": 0.49997714161872864,
      "rewards-max/forensics_tiou_reward": 0.9523380696773529,
      "rewards/forensics_sinkhorn_reward": 0.772654265165329,
      "rewards/forensics_strict_edge_reward": 0.05834335833787918,
      "rewards/forensics_tiou_reward": 0.606402724981308,
      "step": 778
    },
    {
      "completion_length": 22.953125,
      "epoch": 2.3901689708141323,
      "grad_norm": 6.1731767654418945,
      "kl": 0.3408203125,
      "learning_rate": 4.015384615384615e-07,
      "loss": 0.072,
      "reward": 0.7628608047962189,
      "reward_std": 0.05756998807191849,
      "reward_std_all": 0.05423630028963089,
      "rewards-max/forensics_sinkhorn_reward": 0.9783333539962769,
      "rewards-max/forensics_strict_edge_reward": 0.3482862673699856,
      "rewards-max/forensics_tiou_reward": 0.9654828011989594,
      "rewards/forensics_sinkhorn_reward": 0.7358412146568298,
      "rewards/forensics_strict_edge_reward": 0.018373942002654076,
      "rewards/forensics_tiou_reward": 0.534670501947403,
      "step": 779
    },
    {
      "completion_length": 24.703125,
      "epoch": 2.3932411674347156,
      "grad_norm": 6.467382907867432,
      "kl": 0.4228515625,
      "learning_rate": 4.0076923076923077e-07,
      "loss": 0.0016,
      "reward": 0.7755519449710846,
      "reward_std": 0.01691279001533985,
      "reward_std_all": 0.026217976585030556,
      "rewards-max/forensics_sinkhorn_reward": 0.975931853055954,
      "rewards-max/forensics_strict_edge_reward": 0.41118453443050385,
      "rewards-max/forensics_tiou_reward": 0.9637545347213745,
      "rewards/forensics_sinkhorn_reward": 0.6811948716640472,
      "rewards/forensics_strict_edge_reward": 0.046211812645196915,
      "rewards/forensics_tiou_reward": 0.5217367857694626,
      "step": 780
    },
    {
      "completion_length": 25.890625,
      "epoch": 2.3963133640552994,
      "grad_norm": 7.514714241027832,
      "kl": 0.716796875,
      "learning_rate": 4e-07,
      "loss": 0.0643,
      "reward": 0.7408812940120697,
      "reward_std": 0.06052900291979313,
      "reward_std_all": 0.06353462114930153,
      "rewards-max/forensics_sinkhorn_reward": 0.9810847043991089,
      "rewards-max/forensics_strict_edge_reward": 0.1619311049580574,
      "rewards-max/forensics_tiou_reward": 0.979879230260849,
      "rewards/forensics_sinkhorn_reward": 0.6796942055225372,
      "rewards/forensics_strict_edge_reward": 0.02223648689687252,
      "rewards/forensics_tiou_reward": 0.5666987299919128,
      "step": 781
    },
    {
      "completion_length": 23.78125,
      "epoch": 2.399385560675883,
      "grad_norm": 5.461182594299316,
      "kl": 0.373046875,
      "learning_rate": 3.992307692307692e-07,
      "loss": 0.082,
      "reward": 0.7295321226119995,
      "reward_std": 0.056205710396170616,
      "reward_std_all": 0.07038667052984238,
      "rewards-max/forensics_sinkhorn_reward": 0.9809097945690155,
      "rewards-max/forensics_strict_edge_reward": 0.5755647718906403,
      "rewards-max/forensics_tiou_reward": 0.9803724586963654,
      "rewards/forensics_sinkhorn_reward": 0.6805444657802582,
      "rewards/forensics_strict_edge_reward": 0.05719384737312794,
      "rewards/forensics_tiou_reward": 0.5087615847587585,
      "step": 782
    },
    {
      "completion_length": 24.53125,
      "epoch": 2.402457757296467,
      "grad_norm": 5.203139781951904,
      "kl": 0.3681640625,
      "learning_rate": 3.9846153846153846e-07,
      "loss": -0.0373,
      "reward": 0.794217050075531,
      "reward_std": 0.02001475973520428,
      "reward_std_all": 0.019737162394449115,
      "rewards-max/forensics_sinkhorn_reward": 0.9808190166950226,
      "rewards-max/forensics_strict_edge_reward": 0.6703200340270996,
      "rewards-max/forensics_tiou_reward": 0.979800820350647,
      "rewards/forensics_sinkhorn_reward": 0.7889376282691956,
      "rewards/forensics_strict_edge_reward": 0.06330468133091927,
      "rewards/forensics_tiou_reward": 0.6088988184928894,
      "step": 783
    },
    {
      "completion_length": 26.0625,
      "epoch": 2.4055299539170507,
      "grad_norm": 8.550252914428711,
      "kl": 0.7080078125,
      "learning_rate": 3.976923076923077e-07,
      "loss": 0.049,
      "reward": 0.7523815035820007,
      "reward_std": 0.022314386442303658,
      "reward_std_all": 0.04067572485655546,
      "rewards-max/forensics_sinkhorn_reward": 0.9800293147563934,
      "rewards-max/forensics_strict_edge_reward": 0.5543488264083862,
      "rewards-max/forensics_tiou_reward": 0.9781858921051025,
      "rewards/forensics_sinkhorn_reward": 0.7517675459384918,
      "rewards/forensics_strict_edge_reward": 0.06510726921260357,
      "rewards/forensics_tiou_reward": 0.6315990388393402,
      "step": 784
    },
    {
      "completion_length": 26.46875,
      "epoch": 2.4086021505376345,
      "grad_norm": 6.828033447265625,
      "kl": 0.5986328125,
      "learning_rate": 3.969230769230769e-07,
      "loss": 0.0513,
      "reward": 0.7966414988040924,
      "reward_std": 0.016070806741481647,
      "reward_std_all": 0.016141781583428383,
      "rewards-max/forensics_sinkhorn_reward": 0.9736221730709076,
      "rewards-max/forensics_strict_edge_reward": 0.4439363479614258,
      "rewards-max/forensics_tiou_reward": 0.9593269526958466,
      "rewards/forensics_sinkhorn_reward": 0.7902680337429047,
      "rewards/forensics_strict_edge_reward": 0.043748676776885986,
      "rewards/forensics_tiou_reward": 0.6291409730911255,
      "step": 785
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.4116743471582183,
      "grad_norm": 6.455838203430176,
      "kl": 0.439453125,
      "learning_rate": 3.961538461538461e-07,
      "loss": -0.0636,
      "reward": 0.7818875014781952,
      "reward_std": 0.005461482796818018,
      "reward_std_all": 0.01679652463644743,
      "rewards-max/forensics_sinkhorn_reward": 0.9793233573436737,
      "rewards-max/forensics_strict_edge_reward": 0.4809299036860466,
      "rewards-max/forensics_tiou_reward": 0.9693501591682434,
      "rewards/forensics_sinkhorn_reward": 0.7399799525737762,
      "rewards/forensics_strict_edge_reward": 0.06050163973122835,
      "rewards/forensics_tiou_reward": 0.5831087231636047,
      "step": 786
    },
    {
      "completion_length": 24.546875,
      "epoch": 2.4147465437788016,
      "grad_norm": 6.022796630859375,
      "kl": 0.4501953125,
      "learning_rate": 3.9538461538461537e-07,
      "loss": 0.0565,
      "reward": 0.7745746970176697,
      "reward_std": 0.049408361315727234,
      "reward_std_all": 0.045105837285518646,
      "rewards-max/forensics_sinkhorn_reward": 0.9803611636161804,
      "rewards-max/forensics_strict_edge_reward": 0.6650453805923462,
      "rewards-max/forensics_tiou_reward": 0.97615846991539,
      "rewards/forensics_sinkhorn_reward": 0.7903018295764923,
      "rewards/forensics_strict_edge_reward": 0.0453534871339798,
      "rewards/forensics_tiou_reward": 0.5750612020492554,
      "step": 787
    },
    {
      "completion_length": 23.4375,
      "epoch": 2.4178187403993854,
      "grad_norm": 7.890564918518066,
      "kl": 0.4248046875,
      "learning_rate": 3.946153846153846e-07,
      "loss": -0.0426,
      "reward": 0.798454612493515,
      "reward_std": 0.0123209897428751,
      "reward_std_all": 0.013079363852739334,
      "rewards-max/forensics_sinkhorn_reward": 0.9703890979290009,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.950672835111618,
      "rewards/forensics_sinkhorn_reward": 0.7921273410320282,
      "rewards/forensics_strict_edge_reward": 0.03178036864846945,
      "rewards/forensics_tiou_reward": 0.6051205992698669,
      "step": 788
    },
    {
      "completion_length": 24.5,
      "epoch": 2.420890937019969,
      "grad_norm": 17.2929744720459,
      "kl": 1.05859375,
      "learning_rate": 3.9384615384615385e-07,
      "loss": 0.0514,
      "reward": 0.7901206314563751,
      "reward_std": 0.02510606893338263,
      "reward_std_all": 0.024008476990275085,
      "rewards-max/forensics_sinkhorn_reward": 0.9751625657081604,
      "rewards-max/forensics_strict_edge_reward": 0.44648270308971405,
      "rewards-max/forensics_tiou_reward": 0.9572595059871674,
      "rewards/forensics_sinkhorn_reward": 0.7870802283287048,
      "rewards/forensics_strict_edge_reward": 0.04222934693098068,
      "rewards/forensics_tiou_reward": 0.6363759338855743,
      "step": 789
    },
    {
      "completion_length": 27.109375,
      "epoch": 2.423963133640553,
      "grad_norm": 6.443227291107178,
      "kl": 0.609375,
      "learning_rate": 3.9307692307692306e-07,
      "loss": 0.02,
      "reward": 0.7285880446434021,
      "reward_std": 0.035065919160842896,
      "reward_std_all": 0.057381751015782356,
      "rewards-max/forensics_sinkhorn_reward": 0.9824122488498688,
      "rewards-max/forensics_strict_edge_reward": 0.4450170248746872,
      "rewards-max/forensics_tiou_reward": 0.9856995940208435,
      "rewards/forensics_sinkhorn_reward": 0.6582183837890625,
      "rewards/forensics_strict_edge_reward": 0.044429246336221695,
      "rewards/forensics_tiou_reward": 0.5173318982124329,
      "step": 790
    },
    {
      "completion_length": 21.34375,
      "epoch": 2.4270353302611367,
      "grad_norm": 6.878365993499756,
      "kl": 0.4765625,
      "learning_rate": 3.923076923076923e-07,
      "loss": -0.0254,
      "reward": 0.7389088869094849,
      "reward_std": 0.042641203850507736,
      "reward_std_all": 0.055841317400336266,
      "rewards-max/forensics_sinkhorn_reward": 0.9750976264476776,
      "rewards-max/forensics_strict_edge_reward": 0.3663579970598221,
      "rewards-max/forensics_tiou_reward": 0.9573134481906891,
      "rewards/forensics_sinkhorn_reward": 0.7015232145786285,
      "rewards/forensics_strict_edge_reward": 0.031197549775242805,
      "rewards/forensics_tiou_reward": 0.48108527064323425,
      "step": 791
    },
    {
      "completion_length": 28.625,
      "epoch": 2.4301075268817205,
      "grad_norm": 8.256376266479492,
      "kl": 0.509765625,
      "learning_rate": 3.9153846153846154e-07,
      "loss": 0.0293,
      "reward": 0.7329299449920654,
      "reward_std": 0.04113407898694277,
      "reward_std_all": 0.059879439882934093,
      "rewards-max/forensics_sinkhorn_reward": 0.9804126620292664,
      "rewards-max/forensics_strict_edge_reward": 0.18141059315530583,
      "rewards-max/forensics_tiou_reward": 0.9703086912631989,
      "rewards/forensics_sinkhorn_reward": 0.6394668519496918,
      "rewards/forensics_strict_edge_reward": 0.028711345039482694,
      "rewards/forensics_tiou_reward": 0.5332444906234741,
      "step": 792
    },
    {
      "completion_length": 24.71875,
      "epoch": 2.4331797235023043,
      "grad_norm": 7.536142826080322,
      "kl": 0.3798828125,
      "learning_rate": 3.9076923076923076e-07,
      "loss": 0.0189,
      "reward": 0.7762034237384796,
      "reward_std": 0.02248046500608325,
      "reward_std_all": 0.02937858458608389,
      "rewards-max/forensics_sinkhorn_reward": 0.9733494520187378,
      "rewards-max/forensics_strict_edge_reward": 0.611439049243927,
      "rewards-max/forensics_tiou_reward": 0.9580375254154205,
      "rewards/forensics_sinkhorn_reward": 0.7851764559745789,
      "rewards/forensics_strict_edge_reward": 0.08584416285157204,
      "rewards/forensics_tiou_reward": 0.6370685994625092,
      "step": 793
    },
    {
      "completion_length": 23.75,
      "epoch": 2.436251920122888,
      "grad_norm": 7.061822414398193,
      "kl": 0.373046875,
      "learning_rate": 3.8999999999999997e-07,
      "loss": 0.1435,
      "reward": 0.718271791934967,
      "reward_std": 0.07680114731192589,
      "reward_std_all": 0.08229995146393776,
      "rewards-max/forensics_sinkhorn_reward": 0.968230813741684,
      "rewards-max/forensics_strict_edge_reward": 0.297693882137537,
      "rewards-max/forensics_tiou_reward": 0.9514982402324677,
      "rewards/forensics_sinkhorn_reward": 0.6630187630653381,
      "rewards/forensics_strict_edge_reward": 0.02985548786818981,
      "rewards/forensics_tiou_reward": 0.4649924635887146,
      "step": 794
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.439324116743472,
      "grad_norm": 6.680801868438721,
      "kl": 0.400390625,
      "learning_rate": 3.892307692307692e-07,
      "loss": -0.002,
      "reward": 0.7570345997810364,
      "reward_std": 0.047948017716407776,
      "reward_std_all": 0.04842264950275421,
      "rewards-max/forensics_sinkhorn_reward": 0.9688136279582977,
      "rewards-max/forensics_strict_edge_reward": 0.43426865339279175,
      "rewards-max/forensics_tiou_reward": 0.9476158916950226,
      "rewards/forensics_sinkhorn_reward": 0.7207259833812714,
      "rewards/forensics_strict_edge_reward": 0.052136498503386974,
      "rewards/forensics_tiou_reward": 0.5448647737503052,
      "step": 795
    },
    {
      "completion_length": 24.515625,
      "epoch": 2.442396313364055,
      "grad_norm": 6.125454902648926,
      "kl": 0.4521484375,
      "learning_rate": 3.8846153846153845e-07,
      "loss": 0.082,
      "reward": 0.7798048853874207,
      "reward_std": 0.03906075842678547,
      "reward_std_all": 0.0366881899535656,
      "rewards-max/forensics_sinkhorn_reward": 0.9764613807201385,
      "rewards-max/forensics_strict_edge_reward": 0.5368446409702301,
      "rewards-max/forensics_tiou_reward": 0.9709765017032623,
      "rewards/forensics_sinkhorn_reward": 0.7818411290645599,
      "rewards/forensics_strict_edge_reward": 0.047675407491624355,
      "rewards/forensics_tiou_reward": 0.5522487759590149,
      "step": 796
    },
    {
      "completion_length": 28.0625,
      "epoch": 2.445468509984639,
      "grad_norm": 6.537867546081543,
      "kl": 0.509765625,
      "learning_rate": 3.8769230769230767e-07,
      "loss": 0.066,
      "reward": 0.7326102554798126,
      "reward_std": 0.06640099920332432,
      "reward_std_all": 0.07297881692647934,
      "rewards-max/forensics_sinkhorn_reward": 0.9687842726707458,
      "rewards-max/forensics_strict_edge_reward": 0.5674136877059937,
      "rewards-max/forensics_tiou_reward": 0.9565220475196838,
      "rewards/forensics_sinkhorn_reward": 0.6978190541267395,
      "rewards/forensics_strict_edge_reward": 0.03712647035717964,
      "rewards/forensics_tiou_reward": 0.544646605849266,
      "step": 797
    },
    {
      "completion_length": 27.3125,
      "epoch": 2.4485407066052227,
      "grad_norm": 6.3780059814453125,
      "kl": 0.359375,
      "learning_rate": 3.8692307692307693e-07,
      "loss": -0.0169,
      "reward": 0.7660839855670929,
      "reward_std": 0.05149790830910206,
      "reward_std_all": 0.04898243956267834,
      "rewards-max/forensics_sinkhorn_reward": 0.9635316729545593,
      "rewards-max/forensics_strict_edge_reward": 0.24298831820487976,
      "rewards-max/forensics_tiou_reward": 0.9307839572429657,
      "rewards/forensics_sinkhorn_reward": 0.7519309818744659,
      "rewards/forensics_strict_edge_reward": 0.034842345863580704,
      "rewards/forensics_tiou_reward": 0.5665420144796371,
      "step": 798
    },
    {
      "completion_length": 22.421875,
      "epoch": 2.4516129032258065,
      "grad_norm": 7.240115642547607,
      "kl": 0.3056640625,
      "learning_rate": 3.8615384615384615e-07,
      "loss": -0.0991,
      "reward": 0.746983677148819,
      "reward_std": 0.027230439241975546,
      "reward_std_all": 0.03990176133811474,
      "rewards-max/forensics_sinkhorn_reward": 0.9655338823795319,
      "rewards-max/forensics_strict_edge_reward": 0.4415254667401314,
      "rewards-max/forensics_tiou_reward": 0.9391043484210968,
      "rewards/forensics_sinkhorn_reward": 0.7123233079910278,
      "rewards/forensics_strict_edge_reward": 0.058852363377809525,
      "rewards/forensics_tiou_reward": 0.5145593583583832,
      "step": 799
    },
    {
      "completion_length": 23.21875,
      "epoch": 2.4546850998463903,
      "grad_norm": 6.695939064025879,
      "kl": 0.435546875,
      "learning_rate": 3.8538461538461536e-07,
      "loss": 0.0667,
      "reward": 0.7429887652397156,
      "reward_std": 0.041397085413336754,
      "reward_std_all": 0.055790881626307964,
      "rewards-max/forensics_sinkhorn_reward": 0.9609082341194153,
      "rewards-max/forensics_strict_edge_reward": 0.217692282050848,
      "rewards-max/forensics_tiou_reward": 0.9286926984786987,
      "rewards/forensics_sinkhorn_reward": 0.7052525579929352,
      "rewards/forensics_strict_edge_reward": 0.013827256858348846,
      "rewards/forensics_tiou_reward": 0.5005229264497757,
      "step": 800
    },
    {
      "completion_length": 24.640625,
      "epoch": 2.457757296466974,
      "grad_norm": 5.643989086151123,
      "kl": 0.435546875,
      "learning_rate": 3.8461538461538463e-07,
      "loss": -0.0006,
      "reward": 0.7864490449428558,
      "reward_std": 0.02757717203348875,
      "reward_std_all": 0.026779944077134132,
      "rewards-max/forensics_sinkhorn_reward": 0.9760318398475647,
      "rewards-max/forensics_strict_edge_reward": 0.3983181156218052,
      "rewards-max/forensics_tiou_reward": 0.9660834968090057,
      "rewards/forensics_sinkhorn_reward": 0.7995275259017944,
      "rewards/forensics_strict_edge_reward": 0.03602513810619712,
      "rewards/forensics_tiou_reward": 0.6266466677188873,
      "step": 801
    },
    {
      "completion_length": 22.109375,
      "epoch": 2.460829493087558,
      "grad_norm": 8.067039489746094,
      "kl": 0.4033203125,
      "learning_rate": 3.8384615384615384e-07,
      "loss": -0.078,
      "reward": 0.8009158968925476,
      "reward_std": 0.009288029512390494,
      "reward_std_all": 0.009976844070479274,
      "rewards-max/forensics_sinkhorn_reward": 0.9800974130630493,
      "rewards-max/forensics_strict_edge_reward": 0.610392838716507,
      "rewards-max/forensics_tiou_reward": 0.9782612919807434,
      "rewards/forensics_sinkhorn_reward": 0.819951593875885,
      "rewards/forensics_strict_edge_reward": 0.06914505548775196,
      "rewards/forensics_tiou_reward": 0.653372049331665,
      "step": 802
    },
    {
      "completion_length": 23.9375,
      "epoch": 2.463901689708141,
      "grad_norm": 9.664855003356934,
      "kl": 0.4111328125,
      "learning_rate": 3.8307692307692305e-07,
      "loss": -0.0479,
      "reward": 0.7944272756576538,
      "reward_std": 0.02101552952080965,
      "reward_std_all": 0.020007622428238392,
      "rewards-max/forensics_sinkhorn_reward": 0.9726088643074036,
      "rewards-max/forensics_strict_edge_reward": 0.5150517970323563,
      "rewards-max/forensics_tiou_reward": 0.9614275097846985,
      "rewards/forensics_sinkhorn_reward": 0.800573080778122,
      "rewards/forensics_strict_edge_reward": 0.0846688374876976,
      "rewards/forensics_tiou_reward": 0.6339612901210785,
      "step": 803
    },
    {
      "completion_length": 29.21875,
      "epoch": 2.466973886328725,
      "grad_norm": 6.1493120193481445,
      "kl": 0.4482421875,
      "learning_rate": 3.8230769230769227e-07,
      "loss": -0.0031,
      "reward": 0.7685520350933075,
      "reward_std": 0.0521476324647665,
      "reward_std_all": 0.048700351268053055,
      "rewards-max/forensics_sinkhorn_reward": 0.9560860991477966,
      "rewards-max/forensics_strict_edge_reward": 0.37371348589658737,
      "rewards-max/forensics_tiou_reward": 0.9280765652656555,
      "rewards/forensics_sinkhorn_reward": 0.7394578456878662,
      "rewards/forensics_strict_edge_reward": 0.027523923199623823,
      "rewards/forensics_tiou_reward": 0.551457092165947,
      "step": 804
    },
    {
      "completion_length": 28.234375,
      "epoch": 2.4700460829493087,
      "grad_norm": 5.336205959320068,
      "kl": 0.48046875,
      "learning_rate": 3.8153846153846153e-07,
      "loss": 0.02,
      "reward": 0.7685888111591339,
      "reward_std": 0.03964877314865589,
      "reward_std_all": 0.039560604840517044,
      "rewards-max/forensics_sinkhorn_reward": 0.9715499877929688,
      "rewards-max/forensics_strict_edge_reward": 0.38352395594120026,
      "rewards-max/forensics_tiou_reward": 0.9550268650054932,
      "rewards/forensics_sinkhorn_reward": 0.7535265386104584,
      "rewards/forensics_strict_edge_reward": 0.04730609245598316,
      "rewards/forensics_tiou_reward": 0.5976895987987518,
      "step": 805
    },
    {
      "completion_length": 24.578125,
      "epoch": 2.4731182795698925,
      "grad_norm": 7.276841163635254,
      "kl": 0.439453125,
      "learning_rate": 3.8076923076923075e-07,
      "loss": 0.0624,
      "reward": 0.7841724157333374,
      "reward_std": 0.02172014769166708,
      "reward_std_all": 0.022845714818686247,
      "rewards-max/forensics_sinkhorn_reward": 0.9799271821975708,
      "rewards-max/forensics_strict_edge_reward": 0.5081118941307068,
      "rewards-max/forensics_tiou_reward": 0.9750998020172119,
      "rewards/forensics_sinkhorn_reward": 0.7347804605960846,
      "rewards/forensics_strict_edge_reward": 0.051063284277915955,
      "rewards/forensics_tiou_reward": 0.5782826542854309,
      "step": 806
    },
    {
      "completion_length": 22.046875,
      "epoch": 2.4761904761904763,
      "grad_norm": 8.184629440307617,
      "kl": 0.4453125,
      "learning_rate": 3.7999999999999996e-07,
      "loss": 0.0003,
      "reward": 0.7361490428447723,
      "reward_std": 0.047765668481588364,
      "reward_std_all": 0.05957828834652901,
      "rewards-max/forensics_sinkhorn_reward": 0.9644052982330322,
      "rewards-max/forensics_strict_edge_reward": 0.42620211839675903,
      "rewards-max/forensics_tiou_reward": 0.9309097528457642,
      "rewards/forensics_sinkhorn_reward": 0.7031010389328003,
      "rewards/forensics_strict_edge_reward": 0.044778984040021896,
      "rewards/forensics_tiou_reward": 0.5468145906925201,
      "step": 807
    },
    {
      "completion_length": 26.09375,
      "epoch": 2.47926267281106,
      "grad_norm": 5.530717849731445,
      "kl": 0.505859375,
      "learning_rate": 3.792307692307692e-07,
      "loss": 0.0402,
      "reward": 0.7692598104476929,
      "reward_std": 0.03282414190471172,
      "reward_std_all": 0.037301208823919296,
      "rewards-max/forensics_sinkhorn_reward": 0.9711673855781555,
      "rewards-max/forensics_strict_edge_reward": 0.46365442872047424,
      "rewards-max/forensics_tiou_reward": 0.9538992047309875,
      "rewards/forensics_sinkhorn_reward": 0.7506484389305115,
      "rewards/forensics_strict_edge_reward": 0.04683945141732693,
      "rewards/forensics_tiou_reward": 0.5290432274341583,
      "step": 808
    },
    {
      "completion_length": 22.4375,
      "epoch": 2.482334869431644,
      "grad_norm": 7.163198471069336,
      "kl": 0.4287109375,
      "learning_rate": 3.7846153846153844e-07,
      "loss": 0.0086,
      "reward": 0.7658103704452515,
      "reward_std": 0.041942136362195015,
      "reward_std_all": 0.04240199504420161,
      "rewards-max/forensics_sinkhorn_reward": 0.9752622246742249,
      "rewards-max/forensics_strict_edge_reward": 0.5136472433805466,
      "rewards-max/forensics_tiou_reward": 0.9657600224018097,
      "rewards/forensics_sinkhorn_reward": 0.7411050200462341,
      "rewards/forensics_strict_edge_reward": 0.04014834575355053,
      "rewards/forensics_tiou_reward": 0.6372168958187103,
      "step": 809
    },
    {
      "completion_length": 23.578125,
      "epoch": 2.485407066052227,
      "grad_norm": 6.545090198516846,
      "kl": 0.4052734375,
      "learning_rate": 3.776923076923077e-07,
      "loss": 0.0246,
      "reward": 0.7755844295024872,
      "reward_std": 0.035714052617549896,
      "reward_std_all": 0.03537882678210735,
      "rewards-max/forensics_sinkhorn_reward": 0.9816451072692871,
      "rewards-max/forensics_strict_edge_reward": 0.5182250887155533,
      "rewards-max/forensics_tiou_reward": 0.9833538830280304,
      "rewards/forensics_sinkhorn_reward": 0.7658020257949829,
      "rewards/forensics_strict_edge_reward": 0.062298040837049484,
      "rewards/forensics_tiou_reward": 0.5788566768169403,
      "step": 810
    },
    {
      "completion_length": 26.078125,
      "epoch": 2.488479262672811,
      "grad_norm": 5.695175647735596,
      "kl": 0.384765625,
      "learning_rate": 3.769230769230769e-07,
      "loss": 0.0696,
      "reward": 0.7689500153064728,
      "reward_std": 0.054762765765190125,
      "reward_std_all": 0.05035436525940895,
      "rewards-max/forensics_sinkhorn_reward": 0.9811549186706543,
      "rewards-max/forensics_strict_edge_reward": 0.43443287909030914,
      "rewards-max/forensics_tiou_reward": 0.9838602244853973,
      "rewards/forensics_sinkhorn_reward": 0.7608823776245117,
      "rewards/forensics_strict_edge_reward": 0.0469246581196785,
      "rewards/forensics_tiou_reward": 0.5968925803899765,
      "step": 811
    },
    {
      "completion_length": 25.171875,
      "epoch": 2.4915514592933947,
      "grad_norm": 8.152897834777832,
      "kl": 0.4814453125,
      "learning_rate": 3.7615384615384614e-07,
      "loss": 0.0406,
      "reward": 0.7602109909057617,
      "reward_std": 0.032314042560756207,
      "reward_std_all": 0.04319041036069393,
      "rewards-max/forensics_sinkhorn_reward": 0.969071626663208,
      "rewards-max/forensics_strict_edge_reward": 0.5338919758796692,
      "rewards-max/forensics_tiou_reward": 0.9422836899757385,
      "rewards/forensics_sinkhorn_reward": 0.7146058678627014,
      "rewards/forensics_strict_edge_reward": 0.04252919368445873,
      "rewards/forensics_tiou_reward": 0.5651288628578186,
      "step": 812
    },
    {
      "completion_length": 21.53125,
      "epoch": 2.4946236559139785,
      "grad_norm": 8.996001243591309,
      "kl": 0.5654296875,
      "learning_rate": 3.7538461538461535e-07,
      "loss": 0.0766,
      "reward": 0.726635754108429,
      "reward_std": 0.09438022971153259,
      "reward_std_all": 0.08953701332211494,
      "rewards-max/forensics_sinkhorn_reward": 0.9702315032482147,
      "rewards-max/forensics_strict_edge_reward": 0.4322323799133301,
      "rewards-max/forensics_tiou_reward": 0.9449051022529602,
      "rewards/forensics_sinkhorn_reward": 0.6878076493740082,
      "rewards/forensics_strict_edge_reward": 0.030578547157347202,
      "rewards/forensics_tiou_reward": 0.5006276071071625,
      "step": 813
    },
    {
      "completion_length": 24.359375,
      "epoch": 2.4976958525345623,
      "grad_norm": 5.823667526245117,
      "kl": 0.3447265625,
      "learning_rate": 3.746153846153846e-07,
      "loss": 0.0292,
      "reward": 0.77423956990242,
      "reward_std": 0.03534059040248394,
      "reward_std_all": 0.03571422025561333,
      "rewards-max/forensics_sinkhorn_reward": 0.9816868305206299,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9784800112247467,
      "rewards/forensics_sinkhorn_reward": 0.7884427011013031,
      "rewards/forensics_strict_edge_reward": 0.044325729832053185,
      "rewards/forensics_tiou_reward": 0.6137241721153259,
      "step": 814
    },
    {
      "completion_length": 24.734375,
      "epoch": 2.500768049155146,
      "grad_norm": 6.902800559997559,
      "kl": 0.373046875,
      "learning_rate": 3.7384615384615383e-07,
      "loss": -0.1012,
      "reward": 0.7677480578422546,
      "reward_std": 0.038594845682382584,
      "reward_std_all": 0.038783714175224304,
      "rewards-max/forensics_sinkhorn_reward": 0.9815033674240112,
      "rewards-max/forensics_strict_edge_reward": 0.6497856378555298,
      "rewards-max/forensics_tiou_reward": 0.9784395396709442,
      "rewards/forensics_sinkhorn_reward": 0.7567847669124603,
      "rewards/forensics_strict_edge_reward": 0.08069471642374992,
      "rewards/forensics_tiou_reward": 0.6400593519210815,
      "step": 815
    },
    {
      "completion_length": 23.4375,
      "epoch": 2.50384024577573,
      "grad_norm": 7.0461297035217285,
      "kl": 0.6767578125,
      "learning_rate": 3.7307692307692304e-07,
      "loss": 0.036,
      "reward": 0.7832101881504059,
      "reward_std": 0.03930920735001564,
      "reward_std_all": 0.0360188540071249,
      "rewards-max/forensics_sinkhorn_reward": 0.9746695756912231,
      "rewards-max/forensics_strict_edge_reward": 0.38085488975048065,
      "rewards-max/forensics_tiou_reward": 0.9610243439674377,
      "rewards/forensics_sinkhorn_reward": 0.80288165807724,
      "rewards/forensics_strict_edge_reward": 0.03647105302661657,
      "rewards/forensics_tiou_reward": 0.6239672303199768,
      "step": 816
    },
    {
      "completion_length": 22.921875,
      "epoch": 2.506912442396313,
      "grad_norm": 5.7387003898620605,
      "kl": 0.4345703125,
      "learning_rate": 3.7230769230769226e-07,
      "loss": 0.0783,
      "reward": 0.7808813452720642,
      "reward_std": 0.031604460906237364,
      "reward_std_all": 0.030940623488277197,
      "rewards-max/forensics_sinkhorn_reward": 0.9728605449199677,
      "rewards-max/forensics_strict_edge_reward": 0.34690839797258377,
      "rewards-max/forensics_tiou_reward": 0.9601346254348755,
      "rewards/forensics_sinkhorn_reward": 0.7829806208610535,
      "rewards/forensics_strict_edge_reward": 0.038947814144194126,
      "rewards/forensics_tiou_reward": 0.6252005696296692,
      "step": 817
    },
    {
      "completion_length": 24.65625,
      "epoch": 2.509984639016897,
      "grad_norm": 7.982183456420898,
      "kl": 0.453125,
      "learning_rate": 3.715384615384615e-07,
      "loss": -0.0224,
      "reward": 0.744920015335083,
      "reward_std": 0.06022843159735203,
      "reward_std_all": 0.060013994574546814,
      "rewards-max/forensics_sinkhorn_reward": 0.9838713705539703,
      "rewards-max/forensics_strict_edge_reward": 0.40730567276477814,
      "rewards-max/forensics_tiou_reward": 0.9837249219417572,
      "rewards/forensics_sinkhorn_reward": 0.7298568189144135,
      "rewards/forensics_strict_edge_reward": 0.04030741564929485,
      "rewards/forensics_tiou_reward": 0.554964154958725,
      "step": 818
    },
    {
      "completion_length": 26.484375,
      "epoch": 2.5130568356374807,
      "grad_norm": 7.044643402099609,
      "kl": 0.4130859375,
      "learning_rate": 3.707692307692308e-07,
      "loss": 0.0354,
      "reward": 0.7880862951278687,
      "reward_std": 0.025936144404113293,
      "reward_std_all": 0.02517931954935193,
      "rewards-max/forensics_sinkhorn_reward": 0.9787719547748566,
      "rewards-max/forensics_strict_edge_reward": 0.545335203409195,
      "rewards-max/forensics_tiou_reward": 0.973312497138977,
      "rewards/forensics_sinkhorn_reward": 0.7827937304973602,
      "rewards/forensics_strict_edge_reward": 0.0776303056627512,
      "rewards/forensics_tiou_reward": 0.63780677318573,
      "step": 819
    },
    {
      "completion_length": 26.953125,
      "epoch": 2.5161290322580645,
      "grad_norm": 6.002485752105713,
      "kl": 0.357421875,
      "learning_rate": 3.7e-07,
      "loss": 0.0216,
      "reward": 0.7836267352104187,
      "reward_std": 0.02646884275600314,
      "reward_std_all": 0.026713935425505042,
      "rewards-max/forensics_sinkhorn_reward": 0.9808662533760071,
      "rewards-max/forensics_strict_edge_reward": 0.7024016231298447,
      "rewards-max/forensics_tiou_reward": 0.9787032902240753,
      "rewards/forensics_sinkhorn_reward": 0.7612941861152649,
      "rewards/forensics_strict_edge_reward": 0.060273317620158195,
      "rewards/forensics_tiou_reward": 0.5803772658109665,
      "step": 820
    },
    {
      "completion_length": 24.921875,
      "epoch": 2.5192012288786483,
      "grad_norm": 6.659425258636475,
      "kl": 0.3935546875,
      "learning_rate": 3.692307692307692e-07,
      "loss": -0.0423,
      "reward": 0.762675940990448,
      "reward_std": 0.02883820328861475,
      "reward_std_all": 0.03780536912381649,
      "rewards-max/forensics_sinkhorn_reward": 0.9712316393852234,
      "rewards-max/forensics_strict_edge_reward": 0.5641956627368927,
      "rewards-max/forensics_tiou_reward": 0.9587167799472809,
      "rewards/forensics_sinkhorn_reward": 0.6636323630809784,
      "rewards/forensics_strict_edge_reward": 0.05758802406489849,
      "rewards/forensics_tiou_reward": 0.490509033203125,
      "step": 821
    },
    {
      "completion_length": 26.484375,
      "epoch": 2.522273425499232,
      "grad_norm": 6.552952289581299,
      "kl": 0.3095703125,
      "learning_rate": 3.6846153846153843e-07,
      "loss": -0.0357,
      "reward": 0.7869359850883484,
      "reward_std": 0.03357419278472662,
      "reward_std_all": 0.0307298693805933,
      "rewards-max/forensics_sinkhorn_reward": 0.981089174747467,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9760348796844482,
      "rewards/forensics_sinkhorn_reward": 0.7797903120517731,
      "rewards/forensics_strict_edge_reward": 0.09093379229307175,
      "rewards/forensics_tiou_reward": 0.6062157452106476,
      "step": 822
    },
    {
      "completion_length": 27.5625,
      "epoch": 2.525345622119816,
      "grad_norm": 7.467958927154541,
      "kl": 0.78515625,
      "learning_rate": 3.676923076923077e-07,
      "loss": 0.085,
      "reward": 0.7619181871414185,
      "reward_std": 0.04946198966354132,
      "reward_std_all": 0.05205788090825081,
      "rewards-max/forensics_sinkhorn_reward": 0.9742065370082855,
      "rewards-max/forensics_strict_edge_reward": 0.32841892540454865,
      "rewards-max/forensics_tiou_reward": 0.963228166103363,
      "rewards/forensics_sinkhorn_reward": 0.7031374871730804,
      "rewards/forensics_strict_edge_reward": 0.034230551682412624,
      "rewards/forensics_tiou_reward": 0.5146307349205017,
      "step": 823
    },
    {
      "completion_length": 23.609375,
      "epoch": 2.528417818740399,
      "grad_norm": 6.944554328918457,
      "kl": 0.3896484375,
      "learning_rate": 3.669230769230769e-07,
      "loss": 0.0233,
      "reward": 0.7956259250640869,
      "reward_std": 0.018913740059360862,
      "reward_std_all": 0.018102563451975584,
      "rewards-max/forensics_sinkhorn_reward": 0.9687866568565369,
      "rewards-max/forensics_strict_edge_reward": 0.3031960055232048,
      "rewards-max/forensics_tiou_reward": 0.9496646821498871,
      "rewards/forensics_sinkhorn_reward": 0.8209105134010315,
      "rewards/forensics_strict_edge_reward": 0.038944835774600506,
      "rewards/forensics_tiou_reward": 0.6706737577915192,
      "step": 824
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.531490015360983,
      "grad_norm": 5.000720024108887,
      "kl": 0.4326171875,
      "learning_rate": 3.661538461538461e-07,
      "loss": 0.064,
      "reward": 0.7750047147274017,
      "reward_std": 0.03483832720667124,
      "reward_std_all": 0.034753344021737576,
      "rewards-max/forensics_sinkhorn_reward": 0.9664384424686432,
      "rewards-max/forensics_strict_edge_reward": 0.36570020020008087,
      "rewards-max/forensics_tiou_reward": 0.9382423758506775,
      "rewards/forensics_sinkhorn_reward": 0.7145609855651855,
      "rewards/forensics_strict_edge_reward": 0.02810710482299328,
      "rewards/forensics_tiou_reward": 0.5236670076847076,
      "step": 825
    },
    {
      "completion_length": 23.84375,
      "epoch": 2.5345622119815667,
      "grad_norm": 7.4446492195129395,
      "kl": 0.4169921875,
      "learning_rate": 3.6538461538461534e-07,
      "loss": -0.0324,
      "reward": 0.7327803671360016,
      "reward_std": 0.030034526251256466,
      "reward_std_all": 0.04897845070809126,
      "rewards-max/forensics_sinkhorn_reward": 0.9798649847507477,
      "rewards-max/forensics_strict_edge_reward": 0.43260666728019714,
      "rewards-max/forensics_tiou_reward": 0.9728585183620453,
      "rewards/forensics_sinkhorn_reward": 0.7051509916782379,
      "rewards/forensics_strict_edge_reward": 0.053047118708491325,
      "rewards/forensics_tiou_reward": 0.49886880815029144,
      "step": 826
    },
    {
      "completion_length": 24.796875,
      "epoch": 2.5376344086021505,
      "grad_norm": 5.823850154876709,
      "kl": 0.513671875,
      "learning_rate": 3.646153846153846e-07,
      "loss": -0.0465,
      "reward": 0.79350146651268,
      "reward_std": 0.02255242969840765,
      "reward_std_all": 0.02102587977424264,
      "rewards-max/forensics_sinkhorn_reward": 0.982158899307251,
      "rewards-max/forensics_strict_edge_reward": 0.6188668608665466,
      "rewards-max/forensics_tiou_reward": 0.9808251559734344,
      "rewards/forensics_sinkhorn_reward": 0.7822153270244598,
      "rewards/forensics_strict_edge_reward": 0.09992687404155731,
      "rewards/forensics_tiou_reward": 0.6293988227844238,
      "step": 827
    },
    {
      "completion_length": 25.03125,
      "epoch": 2.5407066052227343,
      "grad_norm": 5.7146897315979,
      "kl": 0.5439453125,
      "learning_rate": 3.638461538461538e-07,
      "loss": -0.0501,
      "reward": 0.7982992827892303,
      "reward_std": 0.009307227563112974,
      "reward_std_all": 0.010752857197076082,
      "rewards-max/forensics_sinkhorn_reward": 0.9772972464561462,
      "rewards-max/forensics_strict_edge_reward": 0.4462435245513916,
      "rewards-max/forensics_tiou_reward": 0.9692386090755463,
      "rewards/forensics_sinkhorn_reward": 0.8179176449775696,
      "rewards/forensics_strict_edge_reward": 0.045708198100328445,
      "rewards/forensics_tiou_reward": 0.6584866046905518,
      "step": 828
    },
    {
      "completion_length": 24.203125,
      "epoch": 2.543778801843318,
      "grad_norm": 8.162965774536133,
      "kl": 0.58984375,
      "learning_rate": 3.6307692307692303e-07,
      "loss": 0.0256,
      "reward": 0.761536180973053,
      "reward_std": 0.04634995851665735,
      "reward_std_all": 0.04633969208225608,
      "rewards-max/forensics_sinkhorn_reward": 0.9686580896377563,
      "rewards-max/forensics_strict_edge_reward": 0.4602070748806,
      "rewards-max/forensics_tiou_reward": 0.948428750038147,
      "rewards/forensics_sinkhorn_reward": 0.7241172194480896,
      "rewards/forensics_strict_edge_reward": 0.04312197305262089,
      "rewards/forensics_tiou_reward": 0.5207806825637817,
      "step": 829
    },
    {
      "completion_length": 26.75,
      "epoch": 2.546850998463902,
      "grad_norm": 6.480062961578369,
      "kl": 0.5,
      "learning_rate": 3.623076923076923e-07,
      "loss": -0.0146,
      "reward": 0.7652693092823029,
      "reward_std": 0.03601199388504028,
      "reward_std_all": 0.03820193000137806,
      "rewards-max/forensics_sinkhorn_reward": 0.9627272486686707,
      "rewards-max/forensics_strict_edge_reward": 0.3250425159931183,
      "rewards-max/forensics_tiou_reward": 0.9352085292339325,
      "rewards/forensics_sinkhorn_reward": 0.6996555924415588,
      "rewards/forensics_strict_edge_reward": 0.0357260899618268,
      "rewards/forensics_tiou_reward": 0.4842202514410019,
      "step": 830
    },
    {
      "completion_length": 25.734375,
      "epoch": 2.549923195084485,
      "grad_norm": 7.159539222717285,
      "kl": 0.3447265625,
      "learning_rate": 3.615384615384615e-07,
      "loss": 0.0218,
      "reward": 0.7724495828151703,
      "reward_std": 0.022107000462710857,
      "reward_std_all": 0.02887740544974804,
      "rewards-max/forensics_sinkhorn_reward": 0.9714807868003845,
      "rewards-max/forensics_strict_edge_reward": 0.6642571836709976,
      "rewards-max/forensics_tiou_reward": 0.9574495553970337,
      "rewards/forensics_sinkhorn_reward": 0.7079287469387054,
      "rewards/forensics_strict_edge_reward": 0.061724722385406494,
      "rewards/forensics_tiou_reward": 0.5649285912513733,
      "step": 831
    },
    {
      "completion_length": 25.734375,
      "epoch": 2.5529953917050694,
      "grad_norm": 5.78025484085083,
      "kl": 0.4140625,
      "learning_rate": 3.607692307692308e-07,
      "loss": -0.0632,
      "reward": 0.8009428679943085,
      "reward_std": 0.009844640269875526,
      "reward_std_all": 0.01033797487616539,
      "rewards-max/forensics_sinkhorn_reward": 0.9833230078220367,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9867922067642212,
      "rewards/forensics_sinkhorn_reward": 0.813687652349472,
      "rewards/forensics_strict_edge_reward": 0.10038382932543755,
      "rewards/forensics_tiou_reward": 0.6429755985736847,
      "step": 832
    },
    {
      "completion_length": 27.21875,
      "epoch": 2.5560675883256527,
      "grad_norm": 6.240879058837891,
      "kl": 0.404296875,
      "learning_rate": 3.6e-07,
      "loss": -0.0767,
      "reward": 0.7801215350627899,
      "reward_std": 0.02652745321393013,
      "reward_std_all": 0.027298646979033947,
      "rewards-max/forensics_sinkhorn_reward": 0.9788046479225159,
      "rewards-max/forensics_strict_edge_reward": 0.7149653136730194,
      "rewards-max/forensics_tiou_reward": 0.9750836789608002,
      "rewards/forensics_sinkhorn_reward": 0.7800545692443848,
      "rewards/forensics_strict_edge_reward": 0.09308995306491852,
      "rewards/forensics_tiou_reward": 0.6111434102058411,
      "step": 833
    },
    {
      "completion_length": 26.390625,
      "epoch": 2.5591397849462365,
      "grad_norm": 7.029072284698486,
      "kl": 0.361328125,
      "learning_rate": 3.592307692307692e-07,
      "loss": -0.0547,
      "reward": 0.7378956973552704,
      "reward_std": 0.03174940124154091,
      "reward_std_all": 0.05309857614338398,
      "rewards-max/forensics_sinkhorn_reward": 0.9699995517730713,
      "rewards-max/forensics_strict_edge_reward": 0.21593499183654785,
      "rewards-max/forensics_tiou_reward": 0.9449818134307861,
      "rewards/forensics_sinkhorn_reward": 0.7287596166133881,
      "rewards/forensics_strict_edge_reward": 0.015540463849902153,
      "rewards/forensics_tiou_reward": 0.588735044002533,
      "step": 834
    },
    {
      "completion_length": 28.453125,
      "epoch": 2.5622119815668203,
      "grad_norm": 5.400612831115723,
      "kl": 0.3681640625,
      "learning_rate": 3.584615384615384e-07,
      "loss": -0.0226,
      "reward": 0.7697452008724213,
      "reward_std": 0.0069609241327270865,
      "reward_std_all": 0.024013196118175983,
      "rewards-max/forensics_sinkhorn_reward": 0.9752312898635864,
      "rewards-max/forensics_strict_edge_reward": 0.5106222033500671,
      "rewards-max/forensics_tiou_reward": 0.9684425890445709,
      "rewards/forensics_sinkhorn_reward": 0.7359969019889832,
      "rewards/forensics_strict_edge_reward": 0.07896332815289497,
      "rewards/forensics_tiou_reward": 0.6389298439025879,
      "step": 835
    },
    {
      "completion_length": 23.375,
      "epoch": 2.565284178187404,
      "grad_norm": 10.46802806854248,
      "kl": 0.498046875,
      "learning_rate": 3.576923076923077e-07,
      "loss": 0.0702,
      "reward": 0.6789011657238007,
      "reward_std": 0.1004459597170353,
      "reward_std_all": 0.10752253606915474,
      "rewards-max/forensics_sinkhorn_reward": 0.9712836444377899,
      "rewards-max/forensics_strict_edge_reward": 0.3850630521774292,
      "rewards-max/forensics_tiou_reward": 0.9504100680351257,
      "rewards/forensics_sinkhorn_reward": 0.5842867493629456,
      "rewards/forensics_strict_edge_reward": 0.024403300136327744,
      "rewards/forensics_tiou_reward": 0.37575168907642365,
      "step": 836
    },
    {
      "completion_length": 26.75,
      "epoch": 2.568356374807988,
      "grad_norm": 6.118475914001465,
      "kl": 0.3671875,
      "learning_rate": 3.569230769230769e-07,
      "loss": 0.059,
      "reward": 0.7586475908756256,
      "reward_std": 0.022838707081973553,
      "reward_std_all": 0.03663126099854708,
      "rewards-max/forensics_sinkhorn_reward": 0.9756968915462494,
      "rewards-max/forensics_strict_edge_reward": 0.4100942313671112,
      "rewards-max/forensics_tiou_reward": 0.9595807790756226,
      "rewards/forensics_sinkhorn_reward": 0.7179143130779266,
      "rewards/forensics_strict_edge_reward": 0.05096758063882589,
      "rewards/forensics_tiou_reward": 0.5556458234786987,
      "step": 837
    },
    {
      "completion_length": 26.34375,
      "epoch": 2.571428571428571,
      "grad_norm": 6.709192276000977,
      "kl": 0.40625,
      "learning_rate": 3.561538461538461e-07,
      "loss": 0.034,
      "reward": 0.7539643943309784,
      "reward_std": 0.05733291432261467,
      "reward_std_all": 0.055841682478785515,
      "rewards-max/forensics_sinkhorn_reward": 0.9674429893493652,
      "rewards-max/forensics_strict_edge_reward": 0.41214974224567413,
      "rewards-max/forensics_tiou_reward": 0.9433720409870148,
      "rewards/forensics_sinkhorn_reward": 0.6555047333240509,
      "rewards/forensics_strict_edge_reward": 0.044981760904192924,
      "rewards/forensics_tiou_reward": 0.4953923672437668,
      "step": 838
    },
    {
      "completion_length": 22.671875,
      "epoch": 2.5745007680491554,
      "grad_norm": 6.218178749084473,
      "kl": 0.3271484375,
      "learning_rate": 3.553846153846154e-07,
      "loss": 0.023,
      "reward": 0.7943322062492371,
      "reward_std": 0.017492675338871777,
      "reward_std_all": 0.01754956168588251,
      "rewards-max/forensics_sinkhorn_reward": 0.9769434630870819,
      "rewards-max/forensics_strict_edge_reward": 0.5180972218513489,
      "rewards-max/forensics_tiou_reward": 0.9618965685367584,
      "rewards/forensics_sinkhorn_reward": 0.7683298587799072,
      "rewards/forensics_strict_edge_reward": 0.03022100031375885,
      "rewards/forensics_tiou_reward": 0.5984969735145569,
      "step": 839
    },
    {
      "completion_length": 23.140625,
      "epoch": 2.5775729646697387,
      "grad_norm": 8.696381568908691,
      "kl": 0.3896484375,
      "learning_rate": 3.546153846153846e-07,
      "loss": 0.0391,
      "reward": 0.7825334966182709,
      "reward_std": 0.026044090511277318,
      "reward_std_all": 0.026556806173175573,
      "rewards-max/forensics_sinkhorn_reward": 0.9735026955604553,
      "rewards-max/forensics_strict_edge_reward": 0.5550394356250763,
      "rewards-max/forensics_tiou_reward": 0.9624049663543701,
      "rewards/forensics_sinkhorn_reward": 0.7676086723804474,
      "rewards/forensics_strict_edge_reward": 0.06960730953142047,
      "rewards/forensics_tiou_reward": 0.5777238607406616,
      "step": 840
    },
    {
      "completion_length": 23.359375,
      "epoch": 2.5806451612903225,
      "grad_norm": 7.205254077911377,
      "kl": 0.3828125,
      "learning_rate": 3.5384615384615386e-07,
      "loss": -0.0619,
      "reward": 0.798693060874939,
      "reward_std": 0.008784370846115053,
      "reward_std_all": 0.010526914848014712,
      "rewards-max/forensics_sinkhorn_reward": 0.9589394330978394,
      "rewards-max/forensics_strict_edge_reward": 0.5556192100048065,
      "rewards-max/forensics_tiou_reward": 0.9105022847652435,
      "rewards/forensics_sinkhorn_reward": 0.7489767670631409,
      "rewards/forensics_strict_edge_reward": 0.03442367259413004,
      "rewards/forensics_tiou_reward": 0.5373772382736206,
      "step": 841
    },
    {
      "completion_length": 26.671875,
      "epoch": 2.5837173579109063,
      "grad_norm": 7.8087921142578125,
      "kl": 0.4736328125,
      "learning_rate": 3.530769230769231e-07,
      "loss": 0.0782,
      "reward": 0.7697588801383972,
      "reward_std": 0.03502904251217842,
      "reward_std_all": 0.03619568608701229,
      "rewards-max/forensics_sinkhorn_reward": 0.9665721952915192,
      "rewards-max/forensics_strict_edge_reward": 0.18208679929375648,
      "rewards-max/forensics_tiou_reward": 0.9475252628326416,
      "rewards/forensics_sinkhorn_reward": 0.7143911719322205,
      "rewards/forensics_strict_edge_reward": 0.021277020685374737,
      "rewards/forensics_tiou_reward": 0.5383630990982056,
      "step": 842
    },
    {
      "completion_length": 26.515625,
      "epoch": 2.58678955453149,
      "grad_norm": 8.95711898803711,
      "kl": 0.3271484375,
      "learning_rate": 3.523076923076923e-07,
      "loss": 0.0027,
      "reward": 0.7596006095409393,
      "reward_std": 0.043809883296489716,
      "reward_std_all": 0.0447835810482502,
      "rewards-max/forensics_sinkhorn_reward": 0.9660608172416687,
      "rewards-max/forensics_strict_edge_reward": 0.3816026896238327,
      "rewards-max/forensics_tiou_reward": 0.9437192380428314,
      "rewards/forensics_sinkhorn_reward": 0.6992762088775635,
      "rewards/forensics_strict_edge_reward": 0.0367299597710371,
      "rewards/forensics_tiou_reward": 0.5227875113487244,
      "step": 843
    },
    {
      "completion_length": 28.015625,
      "epoch": 2.589861751152074,
      "grad_norm": 6.578871726989746,
      "kl": 0.3427734375,
      "learning_rate": 3.515384615384615e-07,
      "loss": 0.0903,
      "reward": 0.7659287750720978,
      "reward_std": 0.023452683351933956,
      "reward_std_all": 0.03290449269115925,
      "rewards-max/forensics_sinkhorn_reward": 0.969709038734436,
      "rewards-max/forensics_strict_edge_reward": 0.3031543120741844,
      "rewards-max/forensics_tiou_reward": 0.9447314739227295,
      "rewards/forensics_sinkhorn_reward": 0.715493232011795,
      "rewards/forensics_strict_edge_reward": 0.032781170681118965,
      "rewards/forensics_tiou_reward": 0.5718809962272644,
      "step": 844
    },
    {
      "completion_length": 22.859375,
      "epoch": 2.592933947772657,
      "grad_norm": 6.628844738006592,
      "kl": 0.3095703125,
      "learning_rate": 3.5076923076923077e-07,
      "loss": -0.0267,
      "reward": 0.8024424910545349,
      "reward_std": 0.006019152235239744,
      "reward_std_all": 0.007194468751549721,
      "rewards-max/forensics_sinkhorn_reward": 0.985887885093689,
      "rewards-max/forensics_strict_edge_reward": 0.7079876959323883,
      "rewards-max/forensics_tiou_reward": 0.9913246035575867,
      "rewards/forensics_sinkhorn_reward": 0.7873571813106537,
      "rewards/forensics_strict_edge_reward": 0.11130447313189507,
      "rewards/forensics_tiou_reward": 0.6853983402252197,
      "step": 845
    },
    {
      "completion_length": 27.0625,
      "epoch": 2.5960061443932414,
      "grad_norm": 6.098621368408203,
      "kl": 0.3828125,
      "learning_rate": 3.5e-07,
      "loss": 0.0336,
      "reward": 0.7811104953289032,
      "reward_std": 0.03250595089048147,
      "reward_std_all": 0.031409887596964836,
      "rewards-max/forensics_sinkhorn_reward": 0.9762377440929413,
      "rewards-max/forensics_strict_edge_reward": 0.3277631551027298,
      "rewards-max/forensics_tiou_reward": 0.9666021764278412,
      "rewards/forensics_sinkhorn_reward": 0.8292959034442902,
      "rewards/forensics_strict_edge_reward": 0.03509234730154276,
      "rewards/forensics_tiou_reward": 0.705353707075119,
      "step": 846
    },
    {
      "completion_length": 22.5,
      "epoch": 2.5990783410138247,
      "grad_norm": 7.072641849517822,
      "kl": 0.490234375,
      "learning_rate": 3.492307692307692e-07,
      "loss": 0.0513,
      "reward": 0.7754418253898621,
      "reward_std": 0.0495359655469656,
      "reward_std_all": 0.0449034608900547,
      "rewards-max/forensics_sinkhorn_reward": 0.9815656244754791,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9776550531387329,
      "rewards/forensics_sinkhorn_reward": 0.7434287965297699,
      "rewards/forensics_strict_edge_reward": 0.10167280584573746,
      "rewards/forensics_tiou_reward": 0.5550550222396851,
      "step": 847
    },
    {
      "completion_length": 28.375,
      "epoch": 2.6021505376344085,
      "grad_norm": 6.0685648918151855,
      "kl": 0.3505859375,
      "learning_rate": 3.4846153846153846e-07,
      "loss": -0.0492,
      "reward": 0.7713539898395538,
      "reward_std": 0.022928372025489807,
      "reward_std_all": 0.027851849794387817,
      "rewards-max/forensics_sinkhorn_reward": 0.9784337282180786,
      "rewards-max/forensics_strict_edge_reward": 0.5772995054721832,
      "rewards-max/forensics_tiou_reward": 0.9733076095581055,
      "rewards/forensics_sinkhorn_reward": 0.7771579921245575,
      "rewards/forensics_strict_edge_reward": 0.07326511666178703,
      "rewards/forensics_tiou_reward": 0.6361951231956482,
      "step": 848
    },
    {
      "completion_length": 21.34375,
      "epoch": 2.6052227342549923,
      "grad_norm": 6.334676265716553,
      "kl": 0.3388671875,
      "learning_rate": 3.476923076923077e-07,
      "loss": 0.0131,
      "reward": 0.783098578453064,
      "reward_std": 0.03318763669813052,
      "reward_std_all": 0.0318101414013654,
      "rewards-max/forensics_sinkhorn_reward": 0.9774705767631531,
      "rewards-max/forensics_strict_edge_reward": 0.45131057500839233,
      "rewards-max/forensics_tiou_reward": 0.97104611992836,
      "rewards/forensics_sinkhorn_reward": 0.7690629959106445,
      "rewards/forensics_strict_edge_reward": 0.05449436977505684,
      "rewards/forensics_tiou_reward": 0.5566140711307526,
      "step": 849
    },
    {
      "completion_length": 26.8125,
      "epoch": 2.608294930875576,
      "grad_norm": 6.9710211753845215,
      "kl": 0.3720703125,
      "learning_rate": 3.4692307692307694e-07,
      "loss": 0.0559,
      "reward": 0.770164430141449,
      "reward_std": 0.03695664647966623,
      "reward_std_all": 0.038189563900232315,
      "rewards-max/forensics_sinkhorn_reward": 0.9696973264217377,
      "rewards-max/forensics_strict_edge_reward": 0.35963234305381775,
      "rewards-max/forensics_tiou_reward": 0.9524188935756683,
      "rewards/forensics_sinkhorn_reward": 0.7625154256820679,
      "rewards/forensics_strict_edge_reward": 0.036447688937187195,
      "rewards/forensics_tiou_reward": 0.600161224603653,
      "step": 850
    },
    {
      "completion_length": 22.265625,
      "epoch": 2.61136712749616,
      "grad_norm": 7.391836166381836,
      "kl": 0.5458984375,
      "learning_rate": 3.461538461538461e-07,
      "loss": -0.093,
      "reward": 0.7949300408363342,
      "reward_std": 0.018572487635537982,
      "reward_std_all": 0.018357835244387388,
      "rewards-max/forensics_sinkhorn_reward": 0.9704853594303131,
      "rewards-max/forensics_strict_edge_reward": 0.5755647718906403,
      "rewards-max/forensics_tiou_reward": 0.9537580013275146,
      "rewards/forensics_sinkhorn_reward": 0.77029949426651,
      "rewards/forensics_strict_edge_reward": 0.06721517816185951,
      "rewards/forensics_tiou_reward": 0.5712245106697083,
      "step": 851
    },
    {
      "completion_length": 25.5,
      "epoch": 2.6144393241167436,
      "grad_norm": 6.851655960083008,
      "kl": 0.431640625,
      "learning_rate": 3.4538461538461537e-07,
      "loss": 0.0694,
      "reward": 0.7745218575000763,
      "reward_std": 0.04891432262957096,
      "reward_std_all": 0.04504808969795704,
      "rewards-max/forensics_sinkhorn_reward": 0.9710507690906525,
      "rewards-max/forensics_strict_edge_reward": 0.35583512485027313,
      "rewards-max/forensics_tiou_reward": 0.952382355928421,
      "rewards/forensics_sinkhorn_reward": 0.7188180685043335,
      "rewards/forensics_strict_edge_reward": 0.03706084564328194,
      "rewards/forensics_tiou_reward": 0.5666094720363617,
      "step": 852
    },
    {
      "completion_length": 26.140625,
      "epoch": 2.6175115207373274,
      "grad_norm": 5.715908527374268,
      "kl": 0.4814453125,
      "learning_rate": 3.446153846153846e-07,
      "loss": 0.0129,
      "reward": 0.7810347378253937,
      "reward_std": 0.007664534961804748,
      "reward_std_all": 0.018055451102554798,
      "rewards-max/forensics_sinkhorn_reward": 0.977533757686615,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9764356315135956,
      "rewards/forensics_sinkhorn_reward": 0.8036974370479584,
      "rewards/forensics_strict_edge_reward": 0.06971276924014091,
      "rewards/forensics_tiou_reward": 0.6891526579856873,
      "step": 853
    },
    {
      "completion_length": 28.40625,
      "epoch": 2.6205837173579107,
      "grad_norm": 6.492609977722168,
      "kl": 0.4296875,
      "learning_rate": 3.4384615384615385e-07,
      "loss": 0.066,
      "reward": 0.7661904394626617,
      "reward_std": 0.04235505126416683,
      "reward_std_all": 0.04210215248167515,
      "rewards-max/forensics_sinkhorn_reward": 0.9806798994541168,
      "rewards-max/forensics_strict_edge_reward": 0.5072144269943237,
      "rewards-max/forensics_tiou_reward": 0.9722208678722382,
      "rewards/forensics_sinkhorn_reward": 0.725837379693985,
      "rewards/forensics_strict_edge_reward": 0.07120615616440773,
      "rewards/forensics_tiou_reward": 0.5614742636680603,
      "step": 854
    },
    {
      "completion_length": 26.84375,
      "epoch": 2.6236559139784945,
      "grad_norm": 6.485671043395996,
      "kl": 0.3515625,
      "learning_rate": 3.4307692307692307e-07,
      "loss": 0.0132,
      "reward": 0.7593265175819397,
      "reward_std": 0.048219675198197365,
      "reward_std_all": 0.048309044912457466,
      "rewards-max/forensics_sinkhorn_reward": 0.9812074601650238,
      "rewards-max/forensics_strict_edge_reward": 0.5009046979248524,
      "rewards-max/forensics_tiou_reward": 0.9783868193626404,
      "rewards/forensics_sinkhorn_reward": 0.687039852142334,
      "rewards/forensics_strict_edge_reward": 0.03890147851780057,
      "rewards/forensics_tiou_reward": 0.5140741169452667,
      "step": 855
    },
    {
      "completion_length": 27.125,
      "epoch": 2.6267281105990783,
      "grad_norm": 6.7252960205078125,
      "kl": 0.423828125,
      "learning_rate": 3.423076923076923e-07,
      "loss": 0.0578,
      "reward": 0.8009144961833954,
      "reward_std": 0.00815256917849183,
      "reward_std_all": 0.00943104550242424,
      "rewards-max/forensics_sinkhorn_reward": 0.9746412336826324,
      "rewards-max/forensics_strict_edge_reward": 0.4658895283937454,
      "rewards-max/forensics_tiou_reward": 0.9623723924160004,
      "rewards/forensics_sinkhorn_reward": 0.7948645651340485,
      "rewards/forensics_strict_edge_reward": 0.04237115755677223,
      "rewards/forensics_tiou_reward": 0.6105165481567383,
      "step": 856
    },
    {
      "completion_length": 29.640625,
      "epoch": 2.629800307219662,
      "grad_norm": 5.547269344329834,
      "kl": 0.3896484375,
      "learning_rate": 3.415384615384615e-07,
      "loss": 0.0038,
      "reward": 0.7593252062797546,
      "reward_std": 0.047577159479260445,
      "reward_std_all": 0.048084791749715805,
      "rewards-max/forensics_sinkhorn_reward": 0.9765348732471466,
      "rewards-max/forensics_strict_edge_reward": 0.3975096046924591,
      "rewards-max/forensics_tiou_reward": 0.96563321352005,
      "rewards/forensics_sinkhorn_reward": 0.7286836206912994,
      "rewards/forensics_strict_edge_reward": 0.029137314297258854,
      "rewards/forensics_tiou_reward": 0.617317259311676,
      "step": 857
    },
    {
      "completion_length": 24.375,
      "epoch": 2.632872503840246,
      "grad_norm": 7.136158466339111,
      "kl": 0.380859375,
      "learning_rate": 3.4076923076923076e-07,
      "loss": -0.0502,
      "reward": 0.8005831837654114,
      "reward_std": 0.009478362393565476,
      "reward_std_all": 0.010426819673739374,
      "rewards-max/forensics_sinkhorn_reward": 0.9771153628826141,
      "rewards-max/forensics_strict_edge_reward": 0.589797779917717,
      "rewards-max/forensics_tiou_reward": 0.9662573933601379,
      "rewards/forensics_sinkhorn_reward": 0.7977124750614166,
      "rewards/forensics_strict_edge_reward": 0.06449764408171177,
      "rewards/forensics_tiou_reward": 0.6565912365913391,
      "step": 858
    },
    {
      "completion_length": 25.390625,
      "epoch": 2.6359447004608296,
      "grad_norm": 8.077125549316406,
      "kl": 0.5380859375,
      "learning_rate": 3.4000000000000003e-07,
      "loss": 0.0564,
      "reward": 0.7776792943477631,
      "reward_std": 0.03402771148830652,
      "reward_std_all": 0.033630541525781155,
      "rewards-max/forensics_sinkhorn_reward": 0.9806347787380219,
      "rewards-max/forensics_strict_edge_reward": 0.46688374876976013,
      "rewards-max/forensics_tiou_reward": 0.9724855124950409,
      "rewards/forensics_sinkhorn_reward": 0.745956301689148,
      "rewards/forensics_strict_edge_reward": 0.04939423128962517,
      "rewards/forensics_tiou_reward": 0.533210501074791,
      "step": 859
    },
    {
      "completion_length": 26.625,
      "epoch": 2.6390168970814134,
      "grad_norm": 7.56339168548584,
      "kl": 0.3759765625,
      "learning_rate": 3.392307692307692e-07,
      "loss": 0.0328,
      "reward": 0.7567961513996124,
      "reward_std": 0.014927607960999012,
      "reward_std_all": 0.03430978488177061,
      "rewards-max/forensics_sinkhorn_reward": 0.9630439877510071,
      "rewards-max/forensics_strict_edge_reward": 0.4086703062057495,
      "rewards-max/forensics_tiou_reward": 0.9361987709999084,
      "rewards/forensics_sinkhorn_reward": 0.7104624211788177,
      "rewards/forensics_strict_edge_reward": 0.04800635762512684,
      "rewards/forensics_tiou_reward": 0.5369768738746643,
      "step": 860
    },
    {
      "completion_length": 23.765625,
      "epoch": 2.6420890937019967,
      "grad_norm": 6.882630825042725,
      "kl": 0.6337890625,
      "learning_rate": 3.3846153846153845e-07,
      "loss": -0.0506,
      "reward": 0.773601621389389,
      "reward_std": 0.03424524620641023,
      "reward_std_all": 0.03732738457620144,
      "rewards-max/forensics_sinkhorn_reward": 0.9870184063911438,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9938859343528748,
      "rewards/forensics_sinkhorn_reward": 0.7765021622180939,
      "rewards/forensics_strict_edge_reward": 0.10097365826368332,
      "rewards/forensics_tiou_reward": 0.6132642030715942,
      "step": 861
    },
    {
      "completion_length": 23.390625,
      "epoch": 2.6451612903225805,
      "grad_norm": 7.320603370666504,
      "kl": 0.40234375,
      "learning_rate": 3.3769230769230767e-07,
      "loss": 0.0332,
      "reward": 0.7851105034351349,
      "reward_std": 0.022209536284208298,
      "reward_std_all": 0.023150648456066847,
      "rewards-max/forensics_sinkhorn_reward": 0.9819738566875458,
      "rewards-max/forensics_strict_edge_reward": 0.44819067418575287,
      "rewards-max/forensics_tiou_reward": 0.9811996817588806,
      "rewards/forensics_sinkhorn_reward": 0.7480071187019348,
      "rewards/forensics_strict_edge_reward": 0.04352950118482113,
      "rewards/forensics_tiou_reward": 0.5766371190547943,
      "step": 862
    },
    {
      "completion_length": 24.90625,
      "epoch": 2.6482334869431643,
      "grad_norm": 8.631448745727539,
      "kl": 0.3603515625,
      "learning_rate": 3.3692307692307693e-07,
      "loss": 0.0747,
      "reward": 0.7614580988883972,
      "reward_std": 0.052041955292224884,
      "reward_std_all": 0.05037449859082699,
      "rewards-max/forensics_sinkhorn_reward": 0.9802218675613403,
      "rewards-max/forensics_strict_edge_reward": 0.4709221124649048,
      "rewards-max/forensics_tiou_reward": 0.9764159917831421,
      "rewards/forensics_sinkhorn_reward": 0.7534906268119812,
      "rewards/forensics_strict_edge_reward": 0.05140429548919201,
      "rewards/forensics_tiou_reward": 0.5758537948131561,
      "step": 863
    },
    {
      "completion_length": 24.296875,
      "epoch": 2.651305683563748,
      "grad_norm": 5.976758003234863,
      "kl": 0.484375,
      "learning_rate": 3.361538461538461e-07,
      "loss": 0.1438,
      "reward": 0.7576067447662354,
      "reward_std": 0.03556547313928604,
      "reward_std_all": 0.04388141445815563,
      "rewards-max/forensics_sinkhorn_reward": 0.9853132665157318,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.988389253616333,
      "rewards/forensics_sinkhorn_reward": 0.6779365539550781,
      "rewards/forensics_strict_edge_reward": 0.053706939332187176,
      "rewards/forensics_tiou_reward": 0.4765654504299164,
      "step": 864
    },
    {
      "completion_length": 24.203125,
      "epoch": 2.654377880184332,
      "grad_norm": 7.945371627807617,
      "kl": 0.5185546875,
      "learning_rate": 3.3538461538461536e-07,
      "loss": 0.0568,
      "reward": 0.7656260132789612,
      "reward_std": 0.02904272172600031,
      "reward_std_all": 0.037207008339464664,
      "rewards-max/forensics_sinkhorn_reward": 0.9739159345626831,
      "rewards-max/forensics_strict_edge_reward": 0.43436920642852783,
      "rewards-max/forensics_tiou_reward": 0.9567350447177887,
      "rewards/forensics_sinkhorn_reward": 0.733266294002533,
      "rewards/forensics_strict_edge_reward": 0.040281789377331734,
      "rewards/forensics_tiou_reward": 0.5680558085441589,
      "step": 865
    },
    {
      "completion_length": 23.90625,
      "epoch": 2.6574500768049156,
      "grad_norm": 7.331233024597168,
      "kl": 0.353515625,
      "learning_rate": 3.346153846153846e-07,
      "loss": 0.098,
      "reward": 0.7859025597572327,
      "reward_std": 0.01198681362438947,
      "reward_std_all": 0.01833949238061905,
      "rewards-max/forensics_sinkhorn_reward": 0.9438060820102692,
      "rewards-max/forensics_strict_edge_reward": 0.2502056583762169,
      "rewards-max/forensics_tiou_reward": 0.8869885504245758,
      "rewards/forensics_sinkhorn_reward": 0.6848159432411194,
      "rewards/forensics_strict_edge_reward": 0.0314021403901279,
      "rewards/forensics_tiou_reward": 0.46765780448913574,
      "step": 866
    },
    {
      "completion_length": 25.90625,
      "epoch": 2.6605222734254994,
      "grad_norm": 8.003960609436035,
      "kl": 0.4013671875,
      "learning_rate": 3.3384615384615384e-07,
      "loss": 0.0098,
      "reward": 0.7739655673503876,
      "reward_std": 0.032875911332666874,
      "reward_std_all": 0.033791971392929554,
      "rewards-max/forensics_sinkhorn_reward": 0.969976395368576,
      "rewards-max/forensics_strict_edge_reward": 0.3409062083810568,
      "rewards-max/forensics_tiou_reward": 0.9520120918750763,
      "rewards/forensics_sinkhorn_reward": 0.7691605389118195,
      "rewards/forensics_strict_edge_reward": 0.02913345070555806,
      "rewards/forensics_tiou_reward": 0.6000836491584778,
      "step": 867
    },
    {
      "completion_length": 27.8125,
      "epoch": 2.6635944700460827,
      "grad_norm": 5.123772144317627,
      "kl": 0.4521484375,
      "learning_rate": 3.330769230769231e-07,
      "loss": 0.0651,
      "reward": 0.7847822606563568,
      "reward_std": 0.03287061327137053,
      "reward_std_all": 0.03096646349877119,
      "rewards-max/forensics_sinkhorn_reward": 0.9806443750858307,
      "rewards-max/forensics_strict_edge_reward": 0.3979259729385376,
      "rewards-max/forensics_tiou_reward": 0.9798247516155243,
      "rewards/forensics_sinkhorn_reward": 0.7681035101413727,
      "rewards/forensics_strict_edge_reward": 0.04085130803287029,
      "rewards/forensics_tiou_reward": 0.5970896780490875,
      "step": 868
    },
    {
      "completion_length": 24.75,
      "epoch": 2.6666666666666665,
      "grad_norm": 6.576493263244629,
      "kl": 0.4921875,
      "learning_rate": 3.3230769230769227e-07,
      "loss": 0.0461,
      "reward": 0.7866532802581787,
      "reward_std": 0.02790371049195528,
      "reward_std_all": 0.026956720277667046,
      "rewards-max/forensics_sinkhorn_reward": 0.9795994758605957,
      "rewards-max/forensics_strict_edge_reward": 0.3017394542694092,
      "rewards-max/forensics_tiou_reward": 0.9731035828590393,
      "rewards/forensics_sinkhorn_reward": 0.8042738139629364,
      "rewards/forensics_strict_edge_reward": 0.038672901690006256,
      "rewards/forensics_tiou_reward": 0.6221809089183807,
      "step": 869
    },
    {
      "completion_length": 23.421875,
      "epoch": 2.6697388632872503,
      "grad_norm": 7.078146457672119,
      "kl": 0.36328125,
      "learning_rate": 3.3153846153846154e-07,
      "loss": 0.1048,
      "reward": 0.7879143953323364,
      "reward_std": 0.03151775151491165,
      "reward_std_all": 0.0291249705478549,
      "rewards-max/forensics_sinkhorn_reward": 0.9752669632434845,
      "rewards-max/forensics_strict_edge_reward": 0.3613831289112568,
      "rewards-max/forensics_tiou_reward": 0.9633481502532959,
      "rewards/forensics_sinkhorn_reward": 0.7794161438941956,
      "rewards/forensics_strict_edge_reward": 0.030842178501188755,
      "rewards/forensics_tiou_reward": 0.6213583648204803,
      "step": 870
    },
    {
      "completion_length": 24.390625,
      "epoch": 2.672811059907834,
      "grad_norm": 6.476468563079834,
      "kl": 0.3876953125,
      "learning_rate": 3.3076923076923075e-07,
      "loss": 0.0771,
      "reward": 0.7819820940494537,
      "reward_std": 0.041447589173913,
      "reward_std_all": 0.038038556231185794,
      "rewards-max/forensics_sinkhorn_reward": 0.980660617351532,
      "rewards-max/forensics_strict_edge_reward": 0.40621142089366913,
      "rewards-max/forensics_tiou_reward": 0.9798448085784912,
      "rewards/forensics_sinkhorn_reward": 0.8017632365226746,
      "rewards/forensics_strict_edge_reward": 0.04641946218907833,
      "rewards/forensics_tiou_reward": 0.6274144053459167,
      "step": 871
    },
    {
      "completion_length": 21.984375,
      "epoch": 2.675883256528418,
      "grad_norm": 6.965875625610352,
      "kl": 0.6806640625,
      "learning_rate": 3.3e-07,
      "loss": -0.0166,
      "reward": 0.778942883014679,
      "reward_std": 0.03457558713853359,
      "reward_std_all": 0.033460176549851894,
      "rewards-max/forensics_sinkhorn_reward": 0.9775469303131104,
      "rewards-max/forensics_strict_edge_reward": 0.5677289068698883,
      "rewards-max/forensics_tiou_reward": 0.9740689992904663,
      "rewards/forensics_sinkhorn_reward": 0.7621929049491882,
      "rewards/forensics_strict_edge_reward": 0.04410046711564064,
      "rewards/forensics_tiou_reward": 0.5665038824081421,
      "step": 872
    },
    {
      "completion_length": 25.390625,
      "epoch": 2.6789554531490016,
      "grad_norm": 5.497211933135986,
      "kl": 0.458984375,
      "learning_rate": 3.292307692307692e-07,
      "loss": -0.0322,
      "reward": 0.7828190922737122,
      "reward_std": 0.023700293619185686,
      "reward_std_all": 0.02463339827954769,
      "rewards-max/forensics_sinkhorn_reward": 0.97964146733284,
      "rewards-max/forensics_strict_edge_reward": 0.4625151753425598,
      "rewards-max/forensics_tiou_reward": 0.9753009080886841,
      "rewards/forensics_sinkhorn_reward": 0.7698780000209808,
      "rewards/forensics_strict_edge_reward": 0.048558713868260384,
      "rewards/forensics_tiou_reward": 0.5946592688560486,
      "step": 873
    },
    {
      "completion_length": 25.25,
      "epoch": 2.6820276497695854,
      "grad_norm": 6.597065448760986,
      "kl": 0.44140625,
      "learning_rate": 3.2846153846153844e-07,
      "loss": -0.011,
      "reward": 0.7752099335193634,
      "reward_std": 0.032667411025613546,
      "reward_std_all": 0.03331703133881092,
      "rewards-max/forensics_sinkhorn_reward": 0.9657025933265686,
      "rewards-max/forensics_strict_edge_reward": 0.4365037977695465,
      "rewards-max/forensics_tiou_reward": 0.935239315032959,
      "rewards/forensics_sinkhorn_reward": 0.7020754218101501,
      "rewards/forensics_strict_edge_reward": 0.06294703856110573,
      "rewards/forensics_tiou_reward": 0.5676369667053223,
      "step": 874
    },
    {
      "completion_length": 24.53125,
      "epoch": 2.6850998463901687,
      "grad_norm": 6.741431713104248,
      "kl": 0.4208984375,
      "learning_rate": 3.2769230769230766e-07,
      "loss": 0.074,
      "reward": 0.7442548871040344,
      "reward_std": 0.042622510343790054,
      "reward_std_all": 0.05395118519663811,
      "rewards-max/forensics_sinkhorn_reward": 0.9597696363925934,
      "rewards-max/forensics_strict_edge_reward": 0.28908829391002655,
      "rewards-max/forensics_tiou_reward": 0.9321906268596649,
      "rewards/forensics_sinkhorn_reward": 0.6757753193378448,
      "rewards/forensics_strict_edge_reward": 0.015029619913548231,
      "rewards/forensics_tiou_reward": 0.43416234850883484,
      "step": 875
    },
    {
      "completion_length": 25.875,
      "epoch": 2.688172043010753,
      "grad_norm": 5.43801212310791,
      "kl": 0.404296875,
      "learning_rate": 3.269230769230769e-07,
      "loss": 0.0921,
      "reward": 0.7625550925731659,
      "reward_std": 0.051054059993475676,
      "reward_std_all": 0.04903370328247547,
      "rewards-max/forensics_sinkhorn_reward": 0.9782402217388153,
      "rewards-max/forensics_strict_edge_reward": 0.5079155564308167,
      "rewards-max/forensics_tiou_reward": 0.9718825817108154,
      "rewards/forensics_sinkhorn_reward": 0.7580789923667908,
      "rewards/forensics_strict_edge_reward": 0.06340498477220535,
      "rewards/forensics_tiou_reward": 0.5975939035415649,
      "step": 876
    },
    {
      "completion_length": 21.078125,
      "epoch": 2.6912442396313363,
      "grad_norm": 8.42210578918457,
      "kl": 0.537109375,
      "learning_rate": 3.2615384615384614e-07,
      "loss": 0.0409,
      "reward": 0.7910671830177307,
      "reward_std": 0.026095330715179443,
      "reward_std_all": 0.024382092989981174,
      "rewards-max/forensics_sinkhorn_reward": 0.9795141220092773,
      "rewards-max/forensics_strict_edge_reward": 0.6089114248752594,
      "rewards-max/forensics_tiou_reward": 0.9750579297542572,
      "rewards/forensics_sinkhorn_reward": 0.7909414768218994,
      "rewards/forensics_strict_edge_reward": 0.07646562159061432,
      "rewards/forensics_tiou_reward": 0.6087453067302704,
      "step": 877
    },
    {
      "completion_length": 23.046875,
      "epoch": 2.69431643625192,
      "grad_norm": 8.746633529663086,
      "kl": 0.41796875,
      "learning_rate": 3.2538461538461535e-07,
      "loss": -0.05,
      "reward": 0.7502448558807373,
      "reward_std": 0.041549570858478546,
      "reward_std_all": 0.04764724709093571,
      "rewards-max/forensics_sinkhorn_reward": 0.9598841667175293,
      "rewards-max/forensics_strict_edge_reward": 0.2799844369292259,
      "rewards-max/forensics_tiou_reward": 0.914097011089325,
      "rewards/forensics_sinkhorn_reward": 0.6958881914615631,
      "rewards/forensics_strict_edge_reward": 0.030589445494115353,
      "rewards/forensics_tiou_reward": 0.4557827115058899,
      "step": 878
    },
    {
      "completion_length": 24.015625,
      "epoch": 2.697388632872504,
      "grad_norm": 7.766796112060547,
      "kl": 0.322265625,
      "learning_rate": 3.246153846153846e-07,
      "loss": -0.0236,
      "reward": 0.7907491624355316,
      "reward_std": 0.014267993858084083,
      "reward_std_all": 0.016191008500754833,
      "rewards-max/forensics_sinkhorn_reward": 0.9699321687221527,
      "rewards-max/forensics_strict_edge_reward": 0.395659938454628,
      "rewards-max/forensics_tiou_reward": 0.9422529935836792,
      "rewards/forensics_sinkhorn_reward": 0.7945507168769836,
      "rewards/forensics_strict_edge_reward": 0.02563648484647274,
      "rewards/forensics_tiou_reward": 0.5886345505714417,
      "step": 879
    },
    {
      "completion_length": 23.65625,
      "epoch": 2.7004608294930876,
      "grad_norm": 7.014945983886719,
      "kl": 0.328125,
      "learning_rate": 3.2384615384615383e-07,
      "loss": 0.0406,
      "reward": 0.7355389893054962,
      "reward_std": 0.016631951555609703,
      "reward_std_all": 0.044751500710844994,
      "rewards-max/forensics_sinkhorn_reward": 0.9689073860645294,
      "rewards-max/forensics_strict_edge_reward": 0.5409056097269058,
      "rewards-max/forensics_tiou_reward": 0.9531961679458618,
      "rewards/forensics_sinkhorn_reward": 0.7022032141685486,
      "rewards/forensics_strict_edge_reward": 0.05139649286866188,
      "rewards/forensics_tiou_reward": 0.5159423351287842,
      "step": 880
    },
    {
      "completion_length": 25.5,
      "epoch": 2.7035330261136714,
      "grad_norm": 6.072692394256592,
      "kl": 0.3310546875,
      "learning_rate": 3.230769230769231e-07,
      "loss": 0.0062,
      "reward": 0.7354509234428406,
      "reward_std": 0.06188024766743183,
      "reward_std_all": 0.06381972692906857,
      "rewards-max/forensics_sinkhorn_reward": 0.9820116758346558,
      "rewards-max/forensics_strict_edge_reward": 0.6660739481449127,
      "rewards-max/forensics_tiou_reward": 0.9831415712833405,
      "rewards/forensics_sinkhorn_reward": 0.6853375434875488,
      "rewards/forensics_strict_edge_reward": 0.08211547508835793,
      "rewards/forensics_tiou_reward": 0.5030774474143982,
      "step": 881
    },
    {
      "completion_length": 23.59375,
      "epoch": 2.7066052227342547,
      "grad_norm": 7.028433322906494,
      "kl": 0.5595703125,
      "learning_rate": 3.2230769230769226e-07,
      "loss": 0.0181,
      "reward": 0.7433847784996033,
      "reward_std": 0.022243929095566273,
      "reward_std_all": 0.043988005723804235,
      "rewards-max/forensics_sinkhorn_reward": 0.974721223115921,
      "rewards-max/forensics_strict_edge_reward": 0.2735706493258476,
      "rewards-max/forensics_tiou_reward": 0.9526520073413849,
      "rewards/forensics_sinkhorn_reward": 0.717790424823761,
      "rewards/forensics_strict_edge_reward": 0.015662362333387136,
      "rewards/forensics_tiou_reward": 0.5121937096118927,
      "step": 882
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.709677419354839,
      "grad_norm": 6.091968536376953,
      "kl": 0.353515625,
      "learning_rate": 3.215384615384615e-07,
      "loss": -0.0441,
      "reward": 0.763269454240799,
      "reward_std": 0.02644934505224228,
      "reward_std_all": 0.033726188354194164,
      "rewards-max/forensics_sinkhorn_reward": 0.9591072499752045,
      "rewards-max/forensics_strict_edge_reward": 0.4755018949508667,
      "rewards-max/forensics_tiou_reward": 0.8998135924339294,
      "rewards/forensics_sinkhorn_reward": 0.6978059411048889,
      "rewards/forensics_strict_edge_reward": 0.03748997300863266,
      "rewards/forensics_tiou_reward": 0.5095913261175156,
      "step": 883
    },
    {
      "completion_length": 22.703125,
      "epoch": 2.7127496159754223,
      "grad_norm": 6.6974897384643555,
      "kl": 0.48046875,
      "learning_rate": 3.2076923076923074e-07,
      "loss": 0.0353,
      "reward": 0.7272904217243195,
      "reward_std": 0.054483357816934586,
      "reward_std_all": 0.06717763468623161,
      "rewards-max/forensics_sinkhorn_reward": 0.9842516481876373,
      "rewards-max/forensics_strict_edge_reward": 0.5641956627368927,
      "rewards-max/forensics_tiou_reward": 0.9870635867118835,
      "rewards/forensics_sinkhorn_reward": 0.6770074367523193,
      "rewards/forensics_strict_edge_reward": 0.07712662033736706,
      "rewards/forensics_tiou_reward": 0.5355311930179596,
      "step": 884
    },
    {
      "completion_length": 24.03125,
      "epoch": 2.715821812596006,
      "grad_norm": 9.831250190734863,
      "kl": 0.3203125,
      "learning_rate": 3.2e-07,
      "loss": -0.133,
      "reward": 0.7568110227584839,
      "reward_std": 0.03770480491220951,
      "reward_std_all": 0.04390639625489712,
      "rewards-max/forensics_sinkhorn_reward": 0.9880782961845398,
      "rewards-max/forensics_strict_edge_reward": 0.7032848298549652,
      "rewards-max/forensics_tiou_reward": 0.9976378679275513,
      "rewards/forensics_sinkhorn_reward": 0.7352791726589203,
      "rewards/forensics_strict_edge_reward": 0.06085056625306606,
      "rewards/forensics_tiou_reward": 0.5396204590797424,
      "step": 885
    },
    {
      "completion_length": 24.328125,
      "epoch": 2.71889400921659,
      "grad_norm": 6.376682281494141,
      "kl": 0.5341796875,
      "learning_rate": 3.1923076923076917e-07,
      "loss": 0.057,
      "reward": 0.789494663476944,
      "reward_std": 0.028508832678198814,
      "reward_std_all": 0.026704078540205956,
      "rewards-max/forensics_sinkhorn_reward": 0.9782539904117584,
      "rewards-max/forensics_strict_edge_reward": 0.5928346812725067,
      "rewards-max/forensics_tiou_reward": 0.9672174453735352,
      "rewards/forensics_sinkhorn_reward": 0.7617781758308411,
      "rewards/forensics_strict_edge_reward": 0.07468359172344208,
      "rewards/forensics_tiou_reward": 0.6197552382946014,
      "step": 886
    },
    {
      "completion_length": 27.53125,
      "epoch": 2.7219662058371736,
      "grad_norm": 6.296478271484375,
      "kl": 0.3984375,
      "learning_rate": 3.1846153846153843e-07,
      "loss": 0.1155,
      "reward": 0.7246982455253601,
      "reward_std": 0.05483421869575977,
      "reward_std_all": 0.06889109499752522,
      "rewards-max/forensics_sinkhorn_reward": 0.9743309020996094,
      "rewards-max/forensics_strict_edge_reward": 0.5106222033500671,
      "rewards-max/forensics_tiou_reward": 0.9612350165843964,
      "rewards/forensics_sinkhorn_reward": 0.6660466194152832,
      "rewards/forensics_strict_edge_reward": 0.03257205709815025,
      "rewards/forensics_tiou_reward": 0.49309150874614716,
      "step": 887
    },
    {
      "completion_length": 28.40625,
      "epoch": 2.7250384024577574,
      "grad_norm": 7.544881343841553,
      "kl": 0.4462890625,
      "learning_rate": 3.176923076923077e-07,
      "loss": 0.0142,
      "reward": 0.7929440140724182,
      "reward_std": 0.022569983964785933,
      "reward_std_all": 0.021536318818107247,
      "rewards-max/forensics_sinkhorn_reward": 0.9800297617912292,
      "rewards-max/forensics_strict_edge_reward": 0.6462232172489166,
      "rewards-max/forensics_tiou_reward": 0.9761813580989838,
      "rewards/forensics_sinkhorn_reward": 0.7631358206272125,
      "rewards/forensics_strict_edge_reward": 0.07859235256910324,
      "rewards/forensics_tiou_reward": 0.6183327734470367,
      "step": 888
    },
    {
      "completion_length": 24.21875,
      "epoch": 2.7281105990783407,
      "grad_norm": 6.1285200119018555,
      "kl": 0.361328125,
      "learning_rate": 3.169230769230769e-07,
      "loss": -0.0415,
      "reward": 0.7854051291942596,
      "reward_std": 0.018973970785737038,
      "reward_std_all": 0.020699525251984596,
      "rewards-max/forensics_sinkhorn_reward": 0.9865277707576752,
      "rewards-max/forensics_strict_edge_reward": 0.740818202495575,
      "rewards-max/forensics_tiou_reward": 0.9950718581676483,
      "rewards/forensics_sinkhorn_reward": 0.7743553221225739,
      "rewards/forensics_strict_edge_reward": 0.1093767061829567,
      "rewards/forensics_tiou_reward": 0.592162549495697,
      "step": 889
    },
    {
      "completion_length": 28.84375,
      "epoch": 2.731182795698925,
      "grad_norm": 6.169064044952393,
      "kl": 0.4755859375,
      "learning_rate": 3.161538461538462e-07,
      "loss": 0.1096,
      "reward": 0.759487122297287,
      "reward_std": 0.05229371739551425,
      "reward_std_all": 0.05094345472753048,
      "rewards-max/forensics_sinkhorn_reward": 0.9460888504981995,
      "rewards-max/forensics_strict_edge_reward": 0.13046828098595142,
      "rewards-max/forensics_tiou_reward": 0.886269211769104,
      "rewards/forensics_sinkhorn_reward": 0.7039983570575714,
      "rewards/forensics_strict_edge_reward": 0.016776454111095518,
      "rewards/forensics_tiou_reward": 0.5225991904735565,
      "step": 890
    },
    {
      "completion_length": 22.984375,
      "epoch": 2.7342549923195083,
      "grad_norm": 7.557509899139404,
      "kl": 0.4287109375,
      "learning_rate": 3.1538461538461534e-07,
      "loss": 0.0599,
      "reward": 0.7233087420463562,
      "reward_std": 0.05783488880842924,
      "reward_std_all": 0.07157101295888424,
      "rewards-max/forensics_sinkhorn_reward": 0.9794314205646515,
      "rewards-max/forensics_strict_edge_reward": 0.513702392578125,
      "rewards-max/forensics_tiou_reward": 0.9697105884552002,
      "rewards/forensics_sinkhorn_reward": 0.7331129908561707,
      "rewards/forensics_strict_edge_reward": 0.038604557514190674,
      "rewards/forensics_tiou_reward": 0.5922600626945496,
      "step": 891
    },
    {
      "completion_length": 25.609375,
      "epoch": 2.737327188940092,
      "grad_norm": 6.303308010101318,
      "kl": 0.3671875,
      "learning_rate": 3.146153846153846e-07,
      "loss": 0.0118,
      "reward": 0.7755441963672638,
      "reward_std": 0.03237912431359291,
      "reward_std_all": 0.032713793218135834,
      "rewards-max/forensics_sinkhorn_reward": 0.9826041162014008,
      "rewards-max/forensics_strict_edge_reward": 0.6082689762115479,
      "rewards-max/forensics_tiou_reward": 0.9823521375656128,
      "rewards/forensics_sinkhorn_reward": 0.7373369932174683,
      "rewards/forensics_strict_edge_reward": 0.0750274695456028,
      "rewards/forensics_tiou_reward": 0.603648841381073,
      "step": 892
    },
    {
      "completion_length": 28.78125,
      "epoch": 2.740399385560676,
      "grad_norm": 7.071726322174072,
      "kl": 0.3349609375,
      "learning_rate": 3.138461538461538e-07,
      "loss": 0.0537,
      "reward": 0.7808607518672943,
      "reward_std": 0.03796014003455639,
      "reward_std_all": 0.03547706454992294,
      "rewards-max/forensics_sinkhorn_reward": 0.9695606529712677,
      "rewards-max/forensics_strict_edge_reward": 0.1922900229692459,
      "rewards-max/forensics_tiou_reward": 0.948619544506073,
      "rewards/forensics_sinkhorn_reward": 0.7388911247253418,
      "rewards/forensics_strict_edge_reward": 0.0209337892010808,
      "rewards/forensics_tiou_reward": 0.5478960871696472,
      "step": 893
    },
    {
      "completion_length": 24.515625,
      "epoch": 2.7434715821812596,
      "grad_norm": 7.154181480407715,
      "kl": 0.41796875,
      "learning_rate": 3.130769230769231e-07,
      "loss": 0.0446,
      "reward": 0.7802463173866272,
      "reward_std": 0.01966150081716478,
      "reward_std_all": 0.025625449139624834,
      "rewards-max/forensics_sinkhorn_reward": 0.9764305651187897,
      "rewards-max/forensics_strict_edge_reward": 0.5805044323205948,
      "rewards-max/forensics_tiou_reward": 0.9683592617511749,
      "rewards/forensics_sinkhorn_reward": 0.7766991853713989,
      "rewards/forensics_strict_edge_reward": 0.03843802958726883,
      "rewards/forensics_tiou_reward": 0.5610609948635101,
      "step": 894
    },
    {
      "completion_length": 23.28125,
      "epoch": 2.7465437788018434,
      "grad_norm": 6.259949207305908,
      "kl": 0.4013671875,
      "learning_rate": 3.1230769230769225e-07,
      "loss": 0.0221,
      "reward": 0.7917965054512024,
      "reward_std": 0.01987306633964181,
      "reward_std_all": 0.01968832570128143,
      "rewards-max/forensics_sinkhorn_reward": 0.9782572984695435,
      "rewards-max/forensics_strict_edge_reward": 0.6697178184986115,
      "rewards-max/forensics_tiou_reward": 0.9722922146320343,
      "rewards/forensics_sinkhorn_reward": 0.8117140829563141,
      "rewards/forensics_strict_edge_reward": 0.10706542804837227,
      "rewards/forensics_tiou_reward": 0.6331513822078705,
      "step": 895
    },
    {
      "completion_length": 23.265625,
      "epoch": 2.749615975422427,
      "grad_norm": 7.234539985656738,
      "kl": 0.373046875,
      "learning_rate": 3.115384615384615e-07,
      "loss": 0.0009,
      "reward": 0.78554967045784,
      "reward_std": 0.0241275355219841,
      "reward_std_all": 0.02427137177437544,
      "rewards-max/forensics_sinkhorn_reward": 0.9738987684249878,
      "rewards-max/forensics_strict_edge_reward": 0.5304600298404694,
      "rewards-max/forensics_tiou_reward": 0.957859456539154,
      "rewards/forensics_sinkhorn_reward": 0.8048319518566132,
      "rewards/forensics_strict_edge_reward": 0.035548653453588486,
      "rewards/forensics_tiou_reward": 0.6434448659420013,
      "step": 896
    },
    {
      "completion_length": 24.09375,
      "epoch": 2.752688172043011,
      "grad_norm": 7.665812969207764,
      "kl": 0.33203125,
      "learning_rate": 3.1076923076923073e-07,
      "loss": 0.0575,
      "reward": 0.7500982880592346,
      "reward_std": 0.019594736048020422,
      "reward_std_all": 0.040620594983920455,
      "rewards-max/forensics_sinkhorn_reward": 0.9745773375034332,
      "rewards-max/forensics_strict_edge_reward": 0.5334757566452026,
      "rewards-max/forensics_tiou_reward": 0.9637032747268677,
      "rewards/forensics_sinkhorn_reward": 0.7317178845405579,
      "rewards/forensics_strict_edge_reward": 0.03189040161669254,
      "rewards/forensics_tiou_reward": 0.5382594168186188,
      "step": 897
    },
    {
      "completion_length": 24.765625,
      "epoch": 2.7557603686635943,
      "grad_norm": 6.286808013916016,
      "kl": 0.4677734375,
      "learning_rate": 3.1e-07,
      "loss": 0.0275,
      "reward": 0.7586717903614044,
      "reward_std": 0.03851949516683817,
      "reward_std_all": 0.04244000278413296,
      "rewards-max/forensics_sinkhorn_reward": 0.982568770647049,
      "rewards-max/forensics_strict_edge_reward": 0.7919256687164307,
      "rewards-max/forensics_tiou_reward": 0.9833351969718933,
      "rewards/forensics_sinkhorn_reward": 0.7378112971782684,
      "rewards/forensics_strict_edge_reward": 0.06397556886076927,
      "rewards/forensics_tiou_reward": 0.5730501413345337,
      "step": 898
    },
    {
      "completion_length": 22.15625,
      "epoch": 2.758832565284178,
      "grad_norm": 6.014642715454102,
      "kl": 0.3193359375,
      "learning_rate": 3.0923076923076926e-07,
      "loss": -0.1096,
      "reward": 0.7790084481239319,
      "reward_std": 0.0038165163132362068,
      "reward_std_all": 0.017298184684477746,
      "rewards-max/forensics_sinkhorn_reward": 0.9764589965343475,
      "rewards-max/forensics_strict_edge_reward": 0.6666710078716278,
      "rewards-max/forensics_tiou_reward": 0.9692865610122681,
      "rewards/forensics_sinkhorn_reward": 0.7975846230983734,
      "rewards/forensics_strict_edge_reward": 0.0757320374250412,
      "rewards/forensics_tiou_reward": 0.6271298825740814,
      "step": 899
    },
    {
      "completion_length": 24.78125,
      "epoch": 2.761904761904762,
      "grad_norm": 6.572226524353027,
      "kl": 0.5673828125,
      "learning_rate": 3.084615384615384e-07,
      "loss": 0.0782,
      "reward": 0.7697769403457642,
      "reward_std": 0.051949769258499146,
      "reward_std_all": 0.04839233122766018,
      "rewards-max/forensics_sinkhorn_reward": 0.9765283465385437,
      "rewards-max/forensics_strict_edge_reward": 0.3677594065666199,
      "rewards-max/forensics_tiou_reward": 0.9712276458740234,
      "rewards/forensics_sinkhorn_reward": 0.7461222112178802,
      "rewards/forensics_strict_edge_reward": 0.03510338068008423,
      "rewards/forensics_tiou_reward": 0.5937781929969788,
      "step": 900
    },
    {
      "completion_length": 27.15625,
      "epoch": 2.7649769585253456,
      "grad_norm": 5.090052127838135,
      "kl": 0.3544921875,
      "learning_rate": 3.076923076923077e-07,
      "loss": -0.0315,
      "reward": 0.7396655678749084,
      "reward_std": 0.05852459347806871,
      "reward_std_all": 0.0629497196059674,
      "rewards-max/forensics_sinkhorn_reward": 0.9443351328372955,
      "rewards-max/forensics_strict_edge_reward": 0.5042535066604614,
      "rewards-max/forensics_tiou_reward": 0.8991941213607788,
      "rewards/forensics_sinkhorn_reward": 0.6952015161514282,
      "rewards/forensics_strict_edge_reward": 0.05487687699496746,
      "rewards/forensics_tiou_reward": 0.5911811888217926,
      "step": 901
    },
    {
      "completion_length": 24.265625,
      "epoch": 2.7680491551459294,
      "grad_norm": 6.973277568817139,
      "kl": 0.4033203125,
      "learning_rate": 3.069230769230769e-07,
      "loss": -0.1225,
      "reward": 0.7742849886417389,
      "reward_std": 0.03580767195671797,
      "reward_std_all": 0.03538953512907028,
      "rewards-max/forensics_sinkhorn_reward": 0.9876491725444794,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9958431720733643,
      "rewards/forensics_sinkhorn_reward": 0.7501144409179688,
      "rewards/forensics_strict_edge_reward": 0.08609053492546082,
      "rewards/forensics_tiou_reward": 0.5872697234153748,
      "step": 902
    },
    {
      "completion_length": 24.0625,
      "epoch": 2.771121351766513,
      "grad_norm": 5.603482723236084,
      "kl": 0.408203125,
      "learning_rate": 3.0615384615384617e-07,
      "loss": 0.0373,
      "reward": 0.7526363432407379,
      "reward_std": 0.01982464548200369,
      "reward_std_all": 0.038427100982517004,
      "rewards-max/forensics_sinkhorn_reward": 0.9706578552722931,
      "rewards-max/forensics_strict_edge_reward": 0.4364918917417526,
      "rewards-max/forensics_tiou_reward": 0.9509667456150055,
      "rewards/forensics_sinkhorn_reward": 0.7713809013366699,
      "rewards/forensics_strict_edge_reward": 0.054315462708473206,
      "rewards/forensics_tiou_reward": 0.6082891821861267,
      "step": 903
    },
    {
      "completion_length": 22.78125,
      "epoch": 2.774193548387097,
      "grad_norm": 5.9598708152771,
      "kl": 0.3828125,
      "learning_rate": 3.0538461538461533e-07,
      "loss": 0.007,
      "reward": 0.7905296385288239,
      "reward_std": 0.02283031865954399,
      "reward_std_all": 0.022248444613069296,
      "rewards-max/forensics_sinkhorn_reward": 0.9780167043209076,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9667685925960541,
      "rewards/forensics_sinkhorn_reward": 0.7922157645225525,
      "rewards/forensics_strict_edge_reward": 0.03989399038255215,
      "rewards/forensics_tiou_reward": 0.6914780735969543,
      "step": 904
    },
    {
      "completion_length": 26.46875,
      "epoch": 2.7772657450076803,
      "grad_norm": 7.4730048179626465,
      "kl": 0.35546875,
      "learning_rate": 3.046153846153846e-07,
      "loss": 0.0229,
      "reward": 0.7759870886802673,
      "reward_std": 0.04140360653400421,
      "reward_std_all": 0.03928742930293083,
      "rewards-max/forensics_sinkhorn_reward": 0.9813063144683838,
      "rewards-max/forensics_strict_edge_reward": 0.44126901030540466,
      "rewards-max/forensics_tiou_reward": 0.9787807762622833,
      "rewards/forensics_sinkhorn_reward": 0.7573422193527222,
      "rewards/forensics_strict_edge_reward": 0.05573381669819355,
      "rewards/forensics_tiou_reward": 0.6231882870197296,
      "step": 905
    },
    {
      "completion_length": 27.828125,
      "epoch": 2.780337941628264,
      "grad_norm": 10.943822860717773,
      "kl": 0.759765625,
      "learning_rate": 3.038461538461538e-07,
      "loss": 0.0238,
      "reward": 0.7420623004436493,
      "reward_std": 0.04311944358050823,
      "reward_std_all": 0.05500216409564018,
      "rewards-max/forensics_sinkhorn_reward": 0.9784555435180664,
      "rewards-max/forensics_strict_edge_reward": 0.5369860082864761,
      "rewards-max/forensics_tiou_reward": 0.9711928367614746,
      "rewards/forensics_sinkhorn_reward": 0.6715787649154663,
      "rewards/forensics_strict_edge_reward": 0.048827461898326874,
      "rewards/forensics_tiou_reward": 0.46424640715122223,
      "step": 906
    },
    {
      "completion_length": 22.140625,
      "epoch": 2.783410138248848,
      "grad_norm": 7.63977575302124,
      "kl": 0.4111328125,
      "learning_rate": 3.030769230769231e-07,
      "loss": -0.0851,
      "reward": 0.7965104579925537,
      "reward_std": 0.008766102488152683,
      "reward_std_all": 0.010924478061497211,
      "rewards-max/forensics_sinkhorn_reward": 0.9632592499256134,
      "rewards-max/forensics_strict_edge_reward": 0.36295904219150543,
      "rewards-max/forensics_tiou_reward": 0.9343819618225098,
      "rewards/forensics_sinkhorn_reward": 0.7981624007225037,
      "rewards/forensics_strict_edge_reward": 0.05395803041756153,
      "rewards/forensics_tiou_reward": 0.6360012888908386,
      "step": 907
    },
    {
      "completion_length": 25.0,
      "epoch": 2.7864823348694316,
      "grad_norm": 6.5564351081848145,
      "kl": 0.4013671875,
      "learning_rate": 3.0230769230769235e-07,
      "loss": 0.0683,
      "reward": 0.7615338563919067,
      "reward_std": 0.02800253638997674,
      "reward_std_all": 0.0384199321269989,
      "rewards-max/forensics_sinkhorn_reward": 0.9818038940429688,
      "rewards-max/forensics_strict_edge_reward": 0.6703200340270996,
      "rewards-max/forensics_tiou_reward": 0.973992109298706,
      "rewards/forensics_sinkhorn_reward": 0.7088770270347595,
      "rewards/forensics_strict_edge_reward": 0.06580499559640884,
      "rewards/forensics_tiou_reward": 0.5230540037155151,
      "step": 908
    },
    {
      "completion_length": 23.40625,
      "epoch": 2.7895545314900154,
      "grad_norm": 7.747780799865723,
      "kl": 0.37890625,
      "learning_rate": 3.015384615384615e-07,
      "loss": -0.0232,
      "reward": 0.7458509206771851,
      "reward_std": 0.06228780560195446,
      "reward_std_all": 0.06214812397956848,
      "rewards-max/forensics_sinkhorn_reward": 0.9784605205059052,
      "rewards-max/forensics_strict_edge_reward": 0.39672182500362396,
      "rewards-max/forensics_tiou_reward": 0.9760079681873322,
      "rewards/forensics_sinkhorn_reward": 0.7120535969734192,
      "rewards/forensics_strict_edge_reward": 0.06142611429095268,
      "rewards/forensics_tiou_reward": 0.5592007040977478,
      "step": 909
    },
    {
      "completion_length": 26.265625,
      "epoch": 2.792626728110599,
      "grad_norm": 6.38631010055542,
      "kl": 0.32421875,
      "learning_rate": 3.0076923076923077e-07,
      "loss": 0.0342,
      "reward": 0.7681140005588531,
      "reward_std": 0.05181721691042185,
      "reward_std_all": 0.0487508038058877,
      "rewards-max/forensics_sinkhorn_reward": 0.9793756306171417,
      "rewards-max/forensics_strict_edge_reward": 0.42168185114860535,
      "rewards-max/forensics_tiou_reward": 0.9728914797306061,
      "rewards/forensics_sinkhorn_reward": 0.7662889361381531,
      "rewards/forensics_strict_edge_reward": 0.025316869374364614,
      "rewards/forensics_tiou_reward": 0.5750271081924438,
      "step": 910
    },
    {
      "completion_length": 23.46875,
      "epoch": 2.795698924731183,
      "grad_norm": 7.410383701324463,
      "kl": 0.43359375,
      "learning_rate": 3e-07,
      "loss": 0.0886,
      "reward": 0.7392358779907227,
      "reward_std": 0.04668779671192169,
      "reward_std_all": 0.06156472489237785,
      "rewards-max/forensics_sinkhorn_reward": 0.981817215681076,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9832992553710938,
      "rewards/forensics_sinkhorn_reward": 0.701763391494751,
      "rewards/forensics_strict_edge_reward": 0.06407932937145233,
      "rewards/forensics_tiou_reward": 0.507690355181694,
      "step": 911
    },
    {
      "completion_length": 30.328125,
      "epoch": 2.7987711213517663,
      "grad_norm": 7.189874649047852,
      "kl": 0.44921875,
      "learning_rate": 2.9923076923076925e-07,
      "loss": 0.0346,
      "reward": 0.7700128555297852,
      "reward_std": 0.024975024163722992,
      "reward_std_all": 0.030186928808689117,
      "rewards-max/forensics_sinkhorn_reward": 0.9671170711517334,
      "rewards-max/forensics_strict_edge_reward": 0.2800605297088623,
      "rewards-max/forensics_tiou_reward": 0.950371503829956,
      "rewards/forensics_sinkhorn_reward": 0.6998220682144165,
      "rewards/forensics_strict_edge_reward": 0.01959676295518875,
      "rewards/forensics_tiou_reward": 0.55573371052742,
      "step": 912
    },
    {
      "completion_length": 26.53125,
      "epoch": 2.80184331797235,
      "grad_norm": 5.628510475158691,
      "kl": 0.4775390625,
      "learning_rate": 2.984615384615384e-07,
      "loss": -0.0435,
      "reward": 0.7791748642921448,
      "reward_std": 0.0343575831502676,
      "reward_std_all": 0.03324039001017809,
      "rewards-max/forensics_sinkhorn_reward": 0.9752766788005829,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9665505588054657,
      "rewards/forensics_sinkhorn_reward": 0.7837148308753967,
      "rewards/forensics_strict_edge_reward": 0.055988335981965065,
      "rewards/forensics_tiou_reward": 0.677174985408783,
      "step": 913
    },
    {
      "completion_length": 24.046875,
      "epoch": 2.804915514592934,
      "grad_norm": 8.908194541931152,
      "kl": 0.544921875,
      "learning_rate": 2.976923076923077e-07,
      "loss": 0.1081,
      "reward": 0.745013564825058,
      "reward_std": 0.07328765466809273,
      "reward_std_all": 0.07317396253347397,
      "rewards-max/forensics_sinkhorn_reward": 0.975455105304718,
      "rewards-max/forensics_strict_edge_reward": 0.3751096725463867,
      "rewards-max/forensics_tiou_reward": 0.9592891037464142,
      "rewards/forensics_sinkhorn_reward": 0.7103055715560913,
      "rewards/forensics_strict_edge_reward": 0.055866438895463943,
      "rewards/forensics_tiou_reward": 0.5602203607559204,
      "step": 914
    },
    {
      "completion_length": 24.796875,
      "epoch": 2.8079877112135176,
      "grad_norm": 7.86931848526001,
      "kl": 0.3798828125,
      "learning_rate": 2.969230769230769e-07,
      "loss": -0.0048,
      "reward": 0.7293969988822937,
      "reward_std": 0.06366960890591145,
      "reward_std_all": 0.07384347449988127,
      "rewards-max/forensics_sinkhorn_reward": 0.9783998429775238,
      "rewards-max/forensics_strict_edge_reward": 0.6516932249069214,
      "rewards-max/forensics_tiou_reward": 0.9714029133319855,
      "rewards/forensics_sinkhorn_reward": 0.721648633480072,
      "rewards/forensics_strict_edge_reward": 0.06707318499684334,
      "rewards/forensics_tiou_reward": 0.5482030808925629,
      "step": 915
    },
    {
      "completion_length": 23.0,
      "epoch": 2.8110599078341014,
      "grad_norm": 8.615691184997559,
      "kl": 0.333984375,
      "learning_rate": 2.9615384615384616e-07,
      "loss": 0.0394,
      "reward": 0.7975637018680573,
      "reward_std": 0.014812732115387917,
      "reward_std_all": 0.014827048173174262,
      "rewards-max/forensics_sinkhorn_reward": 0.9758424460887909,
      "rewards-max/forensics_strict_edge_reward": 0.3504088521003723,
      "rewards-max/forensics_tiou_reward": 0.9639734327793121,
      "rewards/forensics_sinkhorn_reward": 0.7906990945339203,
      "rewards/forensics_strict_edge_reward": 0.05354415066540241,
      "rewards/forensics_tiou_reward": 0.6212975680828094,
      "step": 916
    },
    {
      "completion_length": 31.046875,
      "epoch": 2.814132104454685,
      "grad_norm": 6.08396053314209,
      "kl": 0.4091796875,
      "learning_rate": 2.953846153846154e-07,
      "loss": 0.045,
      "reward": 0.775327742099762,
      "reward_std": 0.035288991406559944,
      "reward_std_all": 0.03504521492868662,
      "rewards-max/forensics_sinkhorn_reward": 0.9643989503383636,
      "rewards-max/forensics_strict_edge_reward": 0.24221647530794144,
      "rewards-max/forensics_tiou_reward": 0.9361734390258789,
      "rewards/forensics_sinkhorn_reward": 0.7590149343013763,
      "rewards/forensics_strict_edge_reward": 0.017918251920491457,
      "rewards/forensics_tiou_reward": 0.6163346469402313,
      "step": 917
    },
    {
      "completion_length": 24.640625,
      "epoch": 2.817204301075269,
      "grad_norm": 11.19936466217041,
      "kl": 0.5146484375,
      "learning_rate": 2.946153846153846e-07,
      "loss": 0.0735,
      "reward": 0.7731654047966003,
      "reward_std": 0.03079155646264553,
      "reward_std_all": 0.033738793805241585,
      "rewards-max/forensics_sinkhorn_reward": 0.9631089866161346,
      "rewards-max/forensics_strict_edge_reward": 0.6567722707986832,
      "rewards-max/forensics_tiou_reward": 0.9325627982616425,
      "rewards/forensics_sinkhorn_reward": 0.7363363802433014,
      "rewards/forensics_strict_edge_reward": 0.052322021685540676,
      "rewards/forensics_tiou_reward": 0.5574631989002228,
      "step": 918
    },
    {
      "completion_length": 26.875,
      "epoch": 2.8202764976958523,
      "grad_norm": 7.840920925140381,
      "kl": 0.5,
      "learning_rate": 2.9384615384615385e-07,
      "loss": 0.1418,
      "reward": 0.6827466189861298,
      "reward_std": 0.06862051598727703,
      "reward_std_all": 0.09641432389616966,
      "rewards-max/forensics_sinkhorn_reward": 0.9669873118400574,
      "rewards-max/forensics_strict_edge_reward": 0.470477819442749,
      "rewards-max/forensics_tiou_reward": 0.9408615827560425,
      "rewards/forensics_sinkhorn_reward": 0.6222117841243744,
      "rewards/forensics_strict_edge_reward": 0.04583904333412647,
      "rewards/forensics_tiou_reward": 0.44087323546409607,
      "step": 919
    },
    {
      "completion_length": 23.859375,
      "epoch": 2.823348694316436,
      "grad_norm": 7.66564416885376,
      "kl": 0.4404296875,
      "learning_rate": 2.9307692307692307e-07,
      "loss": 0.0453,
      "reward": 0.7615919709205627,
      "reward_std": 0.05230454821139574,
      "reward_std_all": 0.05055728368461132,
      "rewards-max/forensics_sinkhorn_reward": 0.9702597856521606,
      "rewards-max/forensics_strict_edge_reward": 0.5172730684280396,
      "rewards-max/forensics_tiou_reward": 0.950862318277359,
      "rewards/forensics_sinkhorn_reward": 0.7277268767356873,
      "rewards/forensics_strict_edge_reward": 0.04653790034353733,
      "rewards/forensics_tiou_reward": 0.5229520052671432,
      "step": 920
    },
    {
      "completion_length": 24.640625,
      "epoch": 2.82642089093702,
      "grad_norm": 6.400607585906982,
      "kl": 0.4140625,
      "learning_rate": 2.9230769230769234e-07,
      "loss": 0.0336,
      "reward": 0.7808120846748352,
      "reward_std": 0.03393091633915901,
      "reward_std_all": 0.032697465270757675,
      "rewards-max/forensics_sinkhorn_reward": 0.9762752950191498,
      "rewards-max/forensics_strict_edge_reward": 0.5256026089191437,
      "rewards-max/forensics_tiou_reward": 0.9645155370235443,
      "rewards/forensics_sinkhorn_reward": 0.7619127035140991,
      "rewards/forensics_strict_edge_reward": 0.06864850595593452,
      "rewards/forensics_tiou_reward": 0.6198344230651855,
      "step": 921
    },
    {
      "completion_length": 21.8125,
      "epoch": 2.8294930875576036,
      "grad_norm": 11.173480987548828,
      "kl": 0.3896484375,
      "learning_rate": 2.915384615384615e-07,
      "loss": -0.092,
      "reward": 0.7371495366096497,
      "reward_std": 0.06454205140471458,
      "reward_std_all": 0.0679437518119812,
      "rewards-max/forensics_sinkhorn_reward": 0.9849076271057129,
      "rewards-max/forensics_strict_edge_reward": 0.6174952685832977,
      "rewards-max/forensics_tiou_reward": 0.988201916217804,
      "rewards/forensics_sinkhorn_reward": 0.7455820441246033,
      "rewards/forensics_strict_edge_reward": 0.05170605145394802,
      "rewards/forensics_tiou_reward": 0.5283704400062561,
      "step": 922
    },
    {
      "completion_length": 21.328125,
      "epoch": 2.8325652841781874,
      "grad_norm": 6.439224720001221,
      "kl": 0.341796875,
      "learning_rate": 2.9076923076923076e-07,
      "loss": -0.0122,
      "reward": 0.7959490120410919,
      "reward_std": 0.017664469894953072,
      "reward_std_all": 0.01718078728299588,
      "rewards-max/forensics_sinkhorn_reward": 0.9670165777206421,
      "rewards-max/forensics_strict_edge_reward": 0.35353096202015877,
      "rewards-max/forensics_tiou_reward": 0.9317723214626312,
      "rewards/forensics_sinkhorn_reward": 0.8091810643672943,
      "rewards/forensics_strict_edge_reward": 0.06917553674429655,
      "rewards/forensics_tiou_reward": 0.618087112903595,
      "step": 923
    },
    {
      "completion_length": 22.78125,
      "epoch": 2.835637480798771,
      "grad_norm": 6.1916913986206055,
      "kl": 0.361328125,
      "learning_rate": 2.9e-07,
      "loss": -0.0478,
      "reward": 0.7839078307151794,
      "reward_std": 0.0374955041334033,
      "reward_std_all": 0.03433807007968426,
      "rewards-max/forensics_sinkhorn_reward": 0.9746071398258209,
      "rewards-max/forensics_strict_edge_reward": 0.5253343880176544,
      "rewards-max/forensics_tiou_reward": 0.9638371765613556,
      "rewards/forensics_sinkhorn_reward": 0.7497994303703308,
      "rewards/forensics_strict_edge_reward": 0.06894606910645962,
      "rewards/forensics_tiou_reward": 0.5614491105079651,
      "step": 924
    },
    {
      "completion_length": 23.328125,
      "epoch": 2.838709677419355,
      "grad_norm": 5.938061714172363,
      "kl": 0.427734375,
      "learning_rate": 2.8923076923076924e-07,
      "loss": -0.014,
      "reward": 0.7656663954257965,
      "reward_std": 0.04006636654958129,
      "reward_std_all": 0.04068001173436642,
      "rewards-max/forensics_sinkhorn_reward": 0.9810240268707275,
      "rewards-max/forensics_strict_edge_reward": 0.7161147743463516,
      "rewards-max/forensics_tiou_reward": 0.9753335416316986,
      "rewards/forensics_sinkhorn_reward": 0.7767938673496246,
      "rewards/forensics_strict_edge_reward": 0.06860461831092834,
      "rewards/forensics_tiou_reward": 0.5963899940252304,
      "step": 925
    },
    {
      "completion_length": 26.84375,
      "epoch": 2.8417818740399383,
      "grad_norm": 6.057534217834473,
      "kl": 0.349609375,
      "learning_rate": 2.884615384615384e-07,
      "loss": 0.0365,
      "reward": 0.790094792842865,
      "reward_std": 0.027516272850334644,
      "reward_std_all": 0.025783244520425797,
      "rewards-max/forensics_sinkhorn_reward": 0.9779844284057617,
      "rewards-max/forensics_strict_edge_reward": 0.4554872363805771,
      "rewards-max/forensics_tiou_reward": 0.9721227288246155,
      "rewards/forensics_sinkhorn_reward": 0.7805264294147491,
      "rewards/forensics_strict_edge_reward": 0.04612823389470577,
      "rewards/forensics_tiou_reward": 0.6290244460105896,
      "step": 926
    },
    {
      "completion_length": 22.734375,
      "epoch": 2.8448540706605225,
      "grad_norm": 9.973872184753418,
      "kl": 0.6357421875,
      "learning_rate": 2.8769230769230767e-07,
      "loss": -0.0264,
      "reward": 0.80205038189888,
      "reward_std": 0.008304423419758677,
      "reward_std_all": 0.008960157632827759,
      "rewards-max/forensics_sinkhorn_reward": 0.9668379724025726,
      "rewards-max/forensics_strict_edge_reward": 0.42128483951091766,
      "rewards-max/forensics_tiou_reward": 0.944214791059494,
      "rewards/forensics_sinkhorn_reward": 0.8630923628807068,
      "rewards/forensics_strict_edge_reward": 0.06236064434051514,
      "rewards/forensics_tiou_reward": 0.703075498342514,
      "step": 927
    },
    {
      "completion_length": 24.34375,
      "epoch": 2.847926267281106,
      "grad_norm": 6.080465316772461,
      "kl": 0.4248046875,
      "learning_rate": 2.8692307692307694e-07,
      "loss": 0.0269,
      "reward": 0.7701799869537354,
      "reward_std": 0.03887626272626221,
      "reward_std_all": 0.041213905438780785,
      "rewards-max/forensics_sinkhorn_reward": 0.968756377696991,
      "rewards-max/forensics_strict_edge_reward": 0.5866461992263794,
      "rewards-max/forensics_tiou_reward": 0.952529102563858,
      "rewards/forensics_sinkhorn_reward": 0.7578920125961304,
      "rewards/forensics_strict_edge_reward": 0.05804048106074333,
      "rewards/forensics_tiou_reward": 0.6163869500160217,
      "step": 928
    },
    {
      "completion_length": 27.5,
      "epoch": 2.8509984639016897,
      "grad_norm": 5.504904270172119,
      "kl": 0.369140625,
      "learning_rate": 2.8615384615384615e-07,
      "loss": -0.006,
      "reward": 0.7863728106021881,
      "reward_std": 0.03314994461834431,
      "reward_std_all": 0.031009666621685028,
      "rewards-max/forensics_sinkhorn_reward": 0.9769280254840851,
      "rewards-max/forensics_strict_edge_reward": 0.36135725677013397,
      "rewards-max/forensics_tiou_reward": 0.9705786108970642,
      "rewards/forensics_sinkhorn_reward": 0.7699757218360901,
      "rewards/forensics_strict_edge_reward": 0.04224963113665581,
      "rewards/forensics_tiou_reward": 0.5932248830795288,
      "step": 929
    },
    {
      "completion_length": 28.6875,
      "epoch": 2.8540706605222734,
      "grad_norm": 8.53110408782959,
      "kl": 0.41015625,
      "learning_rate": 2.8538461538461536e-07,
      "loss": 0.0537,
      "reward": 0.7795921564102173,
      "reward_std": 0.037309663370251656,
      "reward_std_all": 0.03563052415847778,
      "rewards-max/forensics_sinkhorn_reward": 0.973157525062561,
      "rewards-max/forensics_strict_edge_reward": 0.35295960307121277,
      "rewards-max/forensics_tiou_reward": 0.9585298001766205,
      "rewards/forensics_sinkhorn_reward": 0.7385379076004028,
      "rewards/forensics_strict_edge_reward": 0.04204967990517616,
      "rewards/forensics_tiou_reward": 0.5839419662952423,
      "step": 930
    },
    {
      "completion_length": 24.40625,
      "epoch": 2.857142857142857,
      "grad_norm": 4.909006118774414,
      "kl": 0.380859375,
      "learning_rate": 2.846153846153846e-07,
      "loss": 0.0068,
      "reward": 0.7803338468074799,
      "reward_std": 0.029603841714560986,
      "reward_std_all": 0.029441243037581444,
      "rewards-max/forensics_sinkhorn_reward": 0.9856491386890411,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9919756650924683,
      "rewards/forensics_sinkhorn_reward": 0.7775397598743439,
      "rewards/forensics_strict_edge_reward": 0.08727695420384407,
      "rewards/forensics_tiou_reward": 0.6601123809814453,
      "step": 931
    },
    {
      "completion_length": 27.109375,
      "epoch": 2.860215053763441,
      "grad_norm": 6.258732318878174,
      "kl": 0.6064453125,
      "learning_rate": 2.8384615384615384e-07,
      "loss": -0.067,
      "reward": 0.790252685546875,
      "reward_std": 0.014053014107048512,
      "reward_std_all": 0.01608276995830238,
      "rewards-max/forensics_sinkhorn_reward": 0.9822683036327362,
      "rewards-max/forensics_strict_edge_reward": 0.7069477140903473,
      "rewards-max/forensics_tiou_reward": 0.9813664555549622,
      "rewards/forensics_sinkhorn_reward": 0.7901530861854553,
      "rewards/forensics_strict_edge_reward": 0.06834856048226357,
      "rewards/forensics_tiou_reward": 0.638959139585495,
      "step": 932
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.8632872503840243,
      "grad_norm": 8.263409614562988,
      "kl": 0.3583984375,
      "learning_rate": 2.8307692307692306e-07,
      "loss": -0.046,
      "reward": 0.7551682591438293,
      "reward_std": 0.025509873405098915,
      "reward_std_all": 0.04069489426910877,
      "rewards-max/forensics_sinkhorn_reward": 0.9802829027175903,
      "rewards-max/forensics_strict_edge_reward": 0.7047859728336334,
      "rewards-max/forensics_tiou_reward": 0.9739718735218048,
      "rewards/forensics_sinkhorn_reward": 0.7444662153720856,
      "rewards/forensics_strict_edge_reward": 0.07609634101390839,
      "rewards/forensics_tiou_reward": 0.5730991214513779,
      "step": 933
    },
    {
      "completion_length": 25.3125,
      "epoch": 2.8663594470046085,
      "grad_norm": 7.07401180267334,
      "kl": 0.3759765625,
      "learning_rate": 2.823076923076923e-07,
      "loss": -0.0257,
      "reward": 0.7836066484451294,
      "reward_std": 0.02263760194182396,
      "reward_std_all": 0.023908040951937437,
      "rewards-max/forensics_sinkhorn_reward": 0.9720157384872437,
      "rewards-max/forensics_strict_edge_reward": 0.6077826023101807,
      "rewards-max/forensics_tiou_reward": 0.9483928382396698,
      "rewards/forensics_sinkhorn_reward": 0.7679768800735474,
      "rewards/forensics_strict_edge_reward": 0.07643739134073257,
      "rewards/forensics_tiou_reward": 0.5672307312488556,
      "step": 934
    },
    {
      "completion_length": 27.34375,
      "epoch": 2.869431643625192,
      "grad_norm": 8.587096214294434,
      "kl": 0.41015625,
      "learning_rate": 2.815384615384615e-07,
      "loss": 0.0653,
      "reward": 0.7920748293399811,
      "reward_std": 0.01960367150604725,
      "reward_std_all": 0.01962930243462324,
      "rewards-max/forensics_sinkhorn_reward": 0.9609352946281433,
      "rewards-max/forensics_strict_edge_reward": 0.3480601981282234,
      "rewards-max/forensics_tiou_reward": 0.9303916394710541,
      "rewards/forensics_sinkhorn_reward": 0.7339214384555817,
      "rewards/forensics_strict_edge_reward": 0.031633807346224785,
      "rewards/forensics_tiou_reward": 0.5565015971660614,
      "step": 935
    },
    {
      "completion_length": 28.578125,
      "epoch": 2.8725038402457757,
      "grad_norm": 4.94484281539917,
      "kl": 0.4189453125,
      "learning_rate": 2.8076923076923075e-07,
      "loss": -0.0323,
      "reward": 0.7443617284297943,
      "reward_std": 0.028490211814641953,
      "reward_std_all": 0.04467407241463661,
      "rewards-max/forensics_sinkhorn_reward": 0.9524249732494354,
      "rewards-max/forensics_strict_edge_reward": 0.2466142512857914,
      "rewards-max/forensics_tiou_reward": 0.9131266176700592,
      "rewards/forensics_sinkhorn_reward": 0.7439242899417877,
      "rewards/forensics_strict_edge_reward": 0.04146614484488964,
      "rewards/forensics_tiou_reward": 0.5943723618984222,
      "step": 936
    },
    {
      "completion_length": 25.953125,
      "epoch": 2.8755760368663594,
      "grad_norm": 6.3796210289001465,
      "kl": 0.3916015625,
      "learning_rate": 2.8e-07,
      "loss": -0.1148,
      "reward": 0.7893423140048981,
      "reward_std": 0.001399729895638302,
      "reward_std_all": 0.011288093868643045,
      "rewards-max/forensics_sinkhorn_reward": 0.9639879763126373,
      "rewards-max/forensics_strict_edge_reward": 0.6092528700828552,
      "rewards-max/forensics_tiou_reward": 0.9370357692241669,
      "rewards/forensics_sinkhorn_reward": 0.782519519329071,
      "rewards/forensics_strict_edge_reward": 0.03113400097936392,
      "rewards/forensics_tiou_reward": 0.5669940710067749,
      "step": 937
    },
    {
      "completion_length": 22.4375,
      "epoch": 2.878648233486943,
      "grad_norm": 7.684858798980713,
      "kl": 0.4521484375,
      "learning_rate": 2.7923076923076923e-07,
      "loss": 0.0146,
      "reward": 0.7724635899066925,
      "reward_std": 0.03320863237604499,
      "reward_std_all": 0.03374871890991926,
      "rewards-max/forensics_sinkhorn_reward": 0.9859174191951752,
      "rewards-max/forensics_strict_edge_reward": 0.5798844993114471,
      "rewards-max/forensics_tiou_reward": 0.9906460344791412,
      "rewards/forensics_sinkhorn_reward": 0.7651906311511993,
      "rewards/forensics_strict_edge_reward": 0.10105899535119534,
      "rewards/forensics_tiou_reward": 0.6554410755634308,
      "step": 938
    },
    {
      "completion_length": 26.078125,
      "epoch": 2.881720430107527,
      "grad_norm": 6.581593036651611,
      "kl": 0.3955078125,
      "learning_rate": 2.7846153846153845e-07,
      "loss": -0.0262,
      "reward": 0.7853602766990662,
      "reward_std": 0.03128944570198655,
      "reward_std_all": 0.029665396315976977,
      "rewards-max/forensics_sinkhorn_reward": 0.9768143892288208,
      "rewards-max/forensics_strict_edge_reward": 0.5422564446926117,
      "rewards-max/forensics_tiou_reward": 0.9632565975189209,
      "rewards/forensics_sinkhorn_reward": 0.7376110255718231,
      "rewards/forensics_strict_edge_reward": 0.06100765988230705,
      "rewards/forensics_tiou_reward": 0.621362566947937,
      "step": 939
    },
    {
      "completion_length": 25.1875,
      "epoch": 2.8847926267281108,
      "grad_norm": 7.288611888885498,
      "kl": 0.4287109375,
      "learning_rate": 2.7769230769230766e-07,
      "loss": 0.0723,
      "reward": 0.7238216996192932,
      "reward_std": 0.04130434989929199,
      "reward_std_all": 0.059953272342681885,
      "rewards-max/forensics_sinkhorn_reward": 0.9643373191356659,
      "rewards-max/forensics_strict_edge_reward": 0.37933504581451416,
      "rewards-max/forensics_tiou_reward": 0.9360926747322083,
      "rewards/forensics_sinkhorn_reward": 0.6596233248710632,
      "rewards/forensics_strict_edge_reward": 0.03859169781208038,
      "rewards/forensics_tiou_reward": 0.5242761969566345,
      "step": 940
    },
    {
      "completion_length": 26.546875,
      "epoch": 2.8878648233486945,
      "grad_norm": 5.448503017425537,
      "kl": 0.51171875,
      "learning_rate": 2.7692307692307693e-07,
      "loss": -0.0211,
      "reward": 0.7757891416549683,
      "reward_std": 0.013998943846672773,
      "reward_std_all": 0.0243055853061378,
      "rewards-max/forensics_sinkhorn_reward": 0.97568279504776,
      "rewards-max/forensics_strict_edge_reward": 0.3040331304073334,
      "rewards-max/forensics_tiou_reward": 0.9629320204257965,
      "rewards/forensics_sinkhorn_reward": 0.738633394241333,
      "rewards/forensics_strict_edge_reward": 0.03800390101969242,
      "rewards/forensics_tiou_reward": 0.5649455487728119,
      "step": 941
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.890937019969278,
      "grad_norm": 7.131460666656494,
      "kl": 0.4052734375,
      "learning_rate": 2.7615384615384614e-07,
      "loss": 0.0273,
      "reward": 0.7710304856300354,
      "reward_std": 0.03528754040598869,
      "reward_std_all": 0.036341216415166855,
      "rewards-max/forensics_sinkhorn_reward": 0.9779406785964966,
      "rewards-max/forensics_strict_edge_reward": 0.7482926547527313,
      "rewards-max/forensics_tiou_reward": 0.9732391238212585,
      "rewards/forensics_sinkhorn_reward": 0.7371939420700073,
      "rewards/forensics_strict_edge_reward": 0.0687514953315258,
      "rewards/forensics_tiou_reward": 0.6061978936195374,
      "step": 942
    },
    {
      "completion_length": 31.875,
      "epoch": 2.8940092165898617,
      "grad_norm": 5.898984432220459,
      "kl": 0.4091796875,
      "learning_rate": 2.753846153846154e-07,
      "loss": 0.0716,
      "reward": 0.7667081952095032,
      "reward_std": 0.037890542298555374,
      "reward_std_all": 0.03938679397106171,
      "rewards-max/forensics_sinkhorn_reward": 0.9741824567317963,
      "rewards-max/forensics_strict_edge_reward": 0.4500439167022705,
      "rewards-max/forensics_tiou_reward": 0.9584319889545441,
      "rewards/forensics_sinkhorn_reward": 0.6797219216823578,
      "rewards/forensics_strict_edge_reward": 0.0662405900657177,
      "rewards/forensics_tiou_reward": 0.5084726214408875,
      "step": 943
    },
    {
      "completion_length": 26.171875,
      "epoch": 2.8970814132104454,
      "grad_norm": 6.350056171417236,
      "kl": 0.4267578125,
      "learning_rate": 2.7461538461538457e-07,
      "loss": -0.0044,
      "reward": 0.7880567908287048,
      "reward_std": 0.015854358673095703,
      "reward_std_all": 0.01877483446151018,
      "rewards-max/forensics_sinkhorn_reward": 0.9706635773181915,
      "rewards-max/forensics_strict_edge_reward": 0.34197375178337097,
      "rewards-max/forensics_tiou_reward": 0.9555906057357788,
      "rewards/forensics_sinkhorn_reward": 0.7958191633224487,
      "rewards/forensics_strict_edge_reward": 0.050419146195054054,
      "rewards/forensics_tiou_reward": 0.6326647698879242,
      "step": 944
    },
    {
      "completion_length": 24.65625,
      "epoch": 2.900153609831029,
      "grad_norm": 7.172354221343994,
      "kl": 0.431640625,
      "learning_rate": 2.7384615384615383e-07,
      "loss": -0.0118,
      "reward": 0.7838167548179626,
      "reward_std": 0.026312639005482197,
      "reward_std_all": 0.026401929557323456,
      "rewards-max/forensics_sinkhorn_reward": 0.9838268160820007,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.9839701652526855,
      "rewards/forensics_sinkhorn_reward": 0.7639928460121155,
      "rewards/forensics_strict_edge_reward": 0.06062094494700432,
      "rewards/forensics_tiou_reward": 0.6455574035644531,
      "step": 945
    },
    {
      "completion_length": 22.3125,
      "epoch": 2.903225806451613,
      "grad_norm": 6.024261474609375,
      "kl": 0.380859375,
      "learning_rate": 2.7307692307692305e-07,
      "loss": 0.011,
      "reward": 0.7829612493515015,
      "reward_std": 0.02275223098695278,
      "reward_std_all": 0.025701877661049366,
      "rewards-max/forensics_sinkhorn_reward": 0.9701918661594391,
      "rewards-max/forensics_strict_edge_reward": 0.45597073435783386,
      "rewards-max/forensics_tiou_reward": 0.94937863945961,
      "rewards/forensics_sinkhorn_reward": 0.773603618144989,
      "rewards/forensics_strict_edge_reward": 0.044136928394436836,
      "rewards/forensics_tiou_reward": 0.5891891121864319,
      "step": 946
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.9062980030721968,
      "grad_norm": 6.6628241539001465,
      "kl": 0.3671875,
      "learning_rate": 2.723076923076923e-07,
      "loss": -0.0627,
      "reward": 0.7651896774768829,
      "reward_std": 0.03747825790196657,
      "reward_std_all": 0.03903198800981045,
      "rewards-max/forensics_sinkhorn_reward": 0.9684665203094482,
      "rewards-max/forensics_strict_edge_reward": 0.5092215240001678,
      "rewards-max/forensics_tiou_reward": 0.9340141713619232,
      "rewards/forensics_sinkhorn_reward": 0.7432336509227753,
      "rewards/forensics_strict_edge_reward": 0.05775942653417587,
      "rewards/forensics_tiou_reward": 0.6005450189113617,
      "step": 947
    },
    {
      "completion_length": 24.90625,
      "epoch": 2.9093701996927805,
      "grad_norm": 6.78341817855835,
      "kl": 0.802734375,
      "learning_rate": 2.7153846153846153e-07,
      "loss": 0.0532,
      "reward": 0.7652941346168518,
      "reward_std": 0.06271198019385338,
      "reward_std_all": 0.057068852707743645,
      "rewards-max/forensics_sinkhorn_reward": 0.9864369928836823,
      "rewards-max/forensics_strict_edge_reward": 0.7429331690073013,
      "rewards-max/forensics_tiou_reward": 0.9915224313735962,
      "rewards/forensics_sinkhorn_reward": 0.8003172278404236,
      "rewards/forensics_strict_edge_reward": 0.0941195972263813,
      "rewards/forensics_tiou_reward": 0.6602658629417419,
      "step": 948
    },
    {
      "completion_length": 26.765625,
      "epoch": 2.912442396313364,
      "grad_norm": 5.341413497924805,
      "kl": 0.43359375,
      "learning_rate": 2.7076923076923074e-07,
      "loss": -0.059,
      "reward": 0.7733366787433624,
      "reward_std": 0.02853454847354442,
      "reward_std_all": 0.0314378987532109,
      "rewards-max/forensics_sinkhorn_reward": 0.9816275537014008,
      "rewards-max/forensics_strict_edge_reward": 0.6799494326114655,
      "rewards-max/forensics_tiou_reward": 0.9820107221603394,
      "rewards/forensics_sinkhorn_reward": 0.7837059795856476,
      "rewards/forensics_strict_edge_reward": 0.06590120121836662,
      "rewards/forensics_tiou_reward": 0.6574687659740448,
      "step": 949
    },
    {
      "completion_length": 26.015625,
      "epoch": 2.9155145929339477,
      "grad_norm": 5.958072662353516,
      "kl": 0.3623046875,
      "learning_rate": 2.7e-07,
      "loss": -0.0375,
      "reward": 0.7917834222316742,
      "reward_std": 0.013905839063227177,
      "reward_std_all": 0.015560295898467302,
      "rewards-max/forensics_sinkhorn_reward": 0.9864714741706848,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9921102821826935,
      "rewards/forensics_sinkhorn_reward": 0.8159392178058624,
      "rewards/forensics_strict_edge_reward": 0.06670898199081421,
      "rewards/forensics_tiou_reward": 0.6903903782367706,
      "step": 950
    },
    {
      "completion_length": 28.609375,
      "epoch": 2.9185867895545314,
      "grad_norm": 8.098945617675781,
      "kl": 0.4228515625,
      "learning_rate": 2.692307692307692e-07,
      "loss": 0.0695,
      "reward": 0.7893881499767303,
      "reward_std": 0.021332168951630592,
      "reward_std_all": 0.021329888608306646,
      "rewards-max/forensics_sinkhorn_reward": 0.9774013459682465,
      "rewards-max/forensics_strict_edge_reward": 0.43898989260196686,
      "rewards-max/forensics_tiou_reward": 0.9694072306156158,
      "rewards/forensics_sinkhorn_reward": 0.7843541502952576,
      "rewards/forensics_strict_edge_reward": 0.04696057550609112,
      "rewards/forensics_tiou_reward": 0.6363456845283508,
      "step": 951
    },
    {
      "completion_length": 26.25,
      "epoch": 2.921658986175115,
      "grad_norm": 6.533669471740723,
      "kl": 0.482421875,
      "learning_rate": 2.6846153846153844e-07,
      "loss": 0.0321,
      "reward": 0.7868378162384033,
      "reward_std": 0.03366871178150177,
      "reward_std_all": 0.030828552320599556,
      "rewards-max/forensics_sinkhorn_reward": 0.982306957244873,
      "rewards-max/forensics_strict_edge_reward": 0.7064213752746582,
      "rewards-max/forensics_tiou_reward": 0.9835764169692993,
      "rewards/forensics_sinkhorn_reward": 0.7765793204307556,
      "rewards/forensics_strict_edge_reward": 0.08464748412370682,
      "rewards/forensics_tiou_reward": 0.6327232718467712,
      "step": 952
    },
    {
      "completion_length": 25.296875,
      "epoch": 2.924731182795699,
      "grad_norm": 6.321256160736084,
      "kl": 0.390625,
      "learning_rate": 2.6769230769230765e-07,
      "loss": -0.0232,
      "reward": 0.7829561531543732,
      "reward_std": 0.032995907589793205,
      "reward_std_all": 0.031361899338662624,
      "rewards-max/forensics_sinkhorn_reward": 0.9802819192409515,
      "rewards-max/forensics_strict_edge_reward": 0.5311143696308136,
      "rewards-max/forensics_tiou_reward": 0.9784325659275055,
      "rewards/forensics_sinkhorn_reward": 0.7779882550239563,
      "rewards/forensics_strict_edge_reward": 0.0906941294670105,
      "rewards/forensics_tiou_reward": 0.5935683250427246,
      "step": 953
    },
    {
      "completion_length": 28.609375,
      "epoch": 2.9278033794162828,
      "grad_norm": 6.258967399597168,
      "kl": 0.453125,
      "learning_rate": 2.669230769230769e-07,
      "loss": -0.0796,
      "reward": 0.7709528505802155,
      "reward_std": 0.014890304300934076,
      "reward_std_all": 0.027037723921239376,
      "rewards-max/forensics_sinkhorn_reward": 0.9728893637657166,
      "rewards-max/forensics_strict_edge_reward": 0.49009793996810913,
      "rewards-max/forensics_tiou_reward": 0.957776665687561,
      "rewards/forensics_sinkhorn_reward": 0.7949880957603455,
      "rewards/forensics_strict_edge_reward": 0.04341870732605457,
      "rewards/forensics_tiou_reward": 0.6536642014980316,
      "step": 954
    },
    {
      "completion_length": 28.34375,
      "epoch": 2.9308755760368665,
      "grad_norm": 6.511508941650391,
      "kl": 0.396484375,
      "learning_rate": 2.6615384615384613e-07,
      "loss": 0.1341,
      "reward": 0.699570506811142,
      "reward_std": 0.06877321191132069,
      "reward_std_all": 0.08597761392593384,
      "rewards-max/forensics_sinkhorn_reward": 0.9741756916046143,
      "rewards-max/forensics_strict_edge_reward": 0.6750346571207047,
      "rewards-max/forensics_tiou_reward": 0.961827427148819,
      "rewards/forensics_sinkhorn_reward": 0.6520311236381531,
      "rewards/forensics_strict_edge_reward": 0.048967983573675156,
      "rewards/forensics_tiou_reward": 0.4776657670736313,
      "step": 955
    },
    {
      "completion_length": 24.796875,
      "epoch": 2.93394777265745,
      "grad_norm": 7.172764778137207,
      "kl": 0.365234375,
      "learning_rate": 2.653846153846154e-07,
      "loss": -0.0498,
      "reward": 0.7952808737754822,
      "reward_std": 0.015259269624948502,
      "reward_std_all": 0.015575459692627192,
      "rewards-max/forensics_sinkhorn_reward": 0.9855163395404816,
      "rewards-max/forensics_strict_edge_reward": 0.7324085682630539,
      "rewards-max/forensics_tiou_reward": 0.9900736510753632,
      "rewards/forensics_sinkhorn_reward": 0.8110010921955109,
      "rewards/forensics_strict_edge_reward": 0.07634090632200241,
      "rewards/forensics_tiou_reward": 0.6371872425079346,
      "step": 956
    },
    {
      "completion_length": 27.421875,
      "epoch": 2.9370199692780337,
      "grad_norm": 5.5717082023620605,
      "kl": 0.3486328125,
      "learning_rate": 2.646153846153846e-07,
      "loss": 0.0002,
      "reward": 0.7687290012836456,
      "reward_std": 0.028186632320284843,
      "reward_std_all": 0.03327697794884443,
      "rewards-max/forensics_sinkhorn_reward": 0.9769890606403351,
      "rewards-max/forensics_strict_edge_reward": 0.4164971336722374,
      "rewards-max/forensics_tiou_reward": 0.9732830226421356,
      "rewards/forensics_sinkhorn_reward": 0.7343754768371582,
      "rewards/forensics_strict_edge_reward": 0.030427265912294388,
      "rewards/forensics_tiou_reward": 0.5603879690170288,
      "step": 957
    },
    {
      "completion_length": 22.140625,
      "epoch": 2.9400921658986174,
      "grad_norm": 7.643734931945801,
      "kl": 0.3955078125,
      "learning_rate": 2.638461538461538e-07,
      "loss": -0.0031,
      "reward": 0.7894318997859955,
      "reward_std": 0.029634172096848488,
      "reward_std_all": 0.02749689668416977,
      "rewards-max/forensics_sinkhorn_reward": 0.9882835149765015,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9975020289421082,
      "rewards/forensics_sinkhorn_reward": 0.7811249196529388,
      "rewards/forensics_strict_edge_reward": 0.12665444612503052,
      "rewards/forensics_tiou_reward": 0.6187235414981842,
      "step": 958
    },
    {
      "completion_length": 22.796875,
      "epoch": 2.943164362519201,
      "grad_norm": 7.90554666519165,
      "kl": 0.4208984375,
      "learning_rate": 2.630769230769231e-07,
      "loss": -0.0417,
      "reward": 0.7603213489055634,
      "reward_std": 0.010249454993754625,
      "reward_std_all": 0.02980798133648932,
      "rewards-max/forensics_sinkhorn_reward": 0.9772339463233948,
      "rewards-max/forensics_strict_edge_reward": 0.6737824082374573,
      "rewards-max/forensics_tiou_reward": 0.9695750772953033,
      "rewards/forensics_sinkhorn_reward": 0.7719021439552307,
      "rewards/forensics_strict_edge_reward": 0.07266329508274794,
      "rewards/forensics_tiou_reward": 0.5957571566104889,
      "step": 959
    },
    {
      "completion_length": 26.6875,
      "epoch": 2.946236559139785,
      "grad_norm": 7.067156791687012,
      "kl": 0.4736328125,
      "learning_rate": 2.623076923076923e-07,
      "loss": -0.016,
      "reward": 0.7907907664775848,
      "reward_std": 0.0215682964771986,
      "reward_std_all": 0.02152908779680729,
      "rewards-max/forensics_sinkhorn_reward": 0.9641233682632446,
      "rewards-max/forensics_strict_edge_reward": 0.36246465146541595,
      "rewards-max/forensics_tiou_reward": 0.93509641289711,
      "rewards/forensics_sinkhorn_reward": 0.777759850025177,
      "rewards/forensics_strict_edge_reward": 0.032037774100899696,
      "rewards/forensics_tiou_reward": 0.594470888376236,
      "step": 960
    },
    {
      "completion_length": 25.65625,
      "epoch": 2.9493087557603688,
      "grad_norm": 6.897115230560303,
      "kl": 0.416015625,
      "learning_rate": 2.615384615384615e-07,
      "loss": 0.0042,
      "reward": 0.7540203630924225,
      "reward_std": 0.054224368184804916,
      "reward_std_all": 0.0536857433617115,
      "rewards-max/forensics_sinkhorn_reward": 0.9810830056667328,
      "rewards-max/forensics_strict_edge_reward": 0.4428497329354286,
      "rewards-max/forensics_tiou_reward": 0.9796963632106781,
      "rewards/forensics_sinkhorn_reward": 0.6832908987998962,
      "rewards/forensics_strict_edge_reward": 0.04928294010460377,
      "rewards/forensics_tiou_reward": 0.5286776721477509,
      "step": 961
    },
    {
      "completion_length": 25.375,
      "epoch": 2.9523809523809526,
      "grad_norm": 7.65044641494751,
      "kl": 0.4140625,
      "learning_rate": 2.6076923076923073e-07,
      "loss": -0.0263,
      "reward": 0.7926420271396637,
      "reward_std": 0.013199944165535271,
      "reward_std_all": 0.016109101474285126,
      "rewards-max/forensics_sinkhorn_reward": 0.9750235676765442,
      "rewards-max/forensics_strict_edge_reward": 0.4856214374303818,
      "rewards-max/forensics_tiou_reward": 0.9629953801631927,
      "rewards/forensics_sinkhorn_reward": 0.7769274711608887,
      "rewards/forensics_strict_edge_reward": 0.04998968914151192,
      "rewards/forensics_tiou_reward": 0.6036560237407684,
      "step": 962
    },
    {
      "completion_length": 25.578125,
      "epoch": 2.955453149001536,
      "grad_norm": 7.593857288360596,
      "kl": 0.591796875,
      "learning_rate": 2.6e-07,
      "loss": 0.0247,
      "reward": 0.7926934063434601,
      "reward_std": 0.019429447129368782,
      "reward_std_all": 0.018984443973749876,
      "rewards-max/forensics_sinkhorn_reward": 0.9686860740184784,
      "rewards-max/forensics_strict_edge_reward": 0.5340259969234467,
      "rewards-max/forensics_tiou_reward": 0.9500504434108734,
      "rewards/forensics_sinkhorn_reward": 0.7709415256977081,
      "rewards/forensics_strict_edge_reward": 0.09765008091926575,
      "rewards/forensics_tiou_reward": 0.6083968877792358,
      "step": 963
    },
    {
      "completion_length": 24.9375,
      "epoch": 2.9585253456221197,
      "grad_norm": 8.355953216552734,
      "kl": 0.673828125,
      "learning_rate": 2.592307692307692e-07,
      "loss": 0.0164,
      "reward": 0.7740331590175629,
      "reward_std": 0.03096943348646164,
      "reward_std_all": 0.03554066177457571,
      "rewards-max/forensics_sinkhorn_reward": 0.9794531762599945,
      "rewards-max/forensics_strict_edge_reward": 0.5626286268234253,
      "rewards-max/forensics_tiou_reward": 0.9756730198860168,
      "rewards/forensics_sinkhorn_reward": 0.7831502556800842,
      "rewards/forensics_strict_edge_reward": 0.06531623937189579,
      "rewards/forensics_tiou_reward": 0.6291376054286957,
      "step": 964
    },
    {
      "completion_length": 26.4375,
      "epoch": 2.9615975422427034,
      "grad_norm": 6.946958541870117,
      "kl": 0.4169921875,
      "learning_rate": 2.584615384615385e-07,
      "loss": 0.0693,
      "reward": 0.7616422772407532,
      "reward_std": 0.017632032744586468,
      "reward_std_all": 0.03212199313566089,
      "rewards-max/forensics_sinkhorn_reward": 0.9661418199539185,
      "rewards-max/forensics_strict_edge_reward": 0.27562522888183594,
      "rewards-max/forensics_tiou_reward": 0.9407171607017517,
      "rewards/forensics_sinkhorn_reward": 0.7228060364723206,
      "rewards/forensics_strict_edge_reward": 0.016473293770104647,
      "rewards/forensics_tiou_reward": 0.5420316010713577,
      "step": 965
    },
    {
      "completion_length": 22.3125,
      "epoch": 2.964669738863287,
      "grad_norm": 8.032602310180664,
      "kl": 0.4072265625,
      "learning_rate": 2.5769230769230764e-07,
      "loss": -0.1939,
      "reward": 0.7683848440647125,
      "reward_std": 0.02473262115381658,
      "reward_std_all": 0.03069490659981966,
      "rewards-max/forensics_sinkhorn_reward": 0.9716497957706451,
      "rewards-max/forensics_strict_edge_reward": 0.3797953277826309,
      "rewards-max/forensics_tiou_reward": 0.9524769186973572,
      "rewards/forensics_sinkhorn_reward": 0.7723917663097382,
      "rewards/forensics_strict_edge_reward": 0.020787186920642853,
      "rewards/forensics_tiou_reward": 0.560745507478714,
      "step": 966
    },
    {
      "completion_length": 24.609375,
      "epoch": 2.967741935483871,
      "grad_norm": 8.025765419006348,
      "kl": 0.33203125,
      "learning_rate": 2.569230769230769e-07,
      "loss": 0.013,
      "reward": 0.7830418646335602,
      "reward_std": 0.021807745099067688,
      "reward_std_all": 0.026194868609309196,
      "rewards-max/forensics_sinkhorn_reward": 0.9831582903862,
      "rewards-max/forensics_strict_edge_reward": 0.5434601008892059,
      "rewards-max/forensics_tiou_reward": 0.9832358360290527,
      "rewards/forensics_sinkhorn_reward": 0.7656608521938324,
      "rewards/forensics_strict_edge_reward": 0.06763247773051262,
      "rewards/forensics_tiou_reward": 0.5992602407932281,
      "step": 967
    },
    {
      "completion_length": 23.296875,
      "epoch": 2.9708141321044548,
      "grad_norm": 7.894698143005371,
      "kl": 0.3984375,
      "learning_rate": 2.5615384615384617e-07,
      "loss": -0.1057,
      "reward": 0.7836722731590271,
      "reward_std": 0.015271040028892457,
      "reward_std_all": 0.01903004478663206,
      "rewards-max/forensics_sinkhorn_reward": 0.9866762161254883,
      "rewards-max/forensics_strict_edge_reward": 0.765928328037262,
      "rewards-max/forensics_tiou_reward": 0.9937331974506378,
      "rewards/forensics_sinkhorn_reward": 0.8119159638881683,
      "rewards/forensics_strict_edge_reward": 0.09621090814471245,
      "rewards/forensics_tiou_reward": 0.6576721966266632,
      "step": 968
    },
    {
      "completion_length": 27.8125,
      "epoch": 2.9738863287250386,
      "grad_norm": 8.255422592163086,
      "kl": 0.69140625,
      "learning_rate": 2.553846153846154e-07,
      "loss": 0.0335,
      "reward": 0.795542448759079,
      "reward_std": 0.018149579409509897,
      "reward_std_all": 0.01764894975349307,
      "rewards-max/forensics_sinkhorn_reward": 0.9790612757205963,
      "rewards-max/forensics_strict_edge_reward": 0.5434702187776566,
      "rewards-max/forensics_tiou_reward": 0.9714571535587311,
      "rewards/forensics_sinkhorn_reward": 0.7692901492118835,
      "rewards/forensics_strict_edge_reward": 0.06149275414645672,
      "rewards/forensics_tiou_reward": 0.6673773229122162,
      "step": 969
    },
    {
      "completion_length": 25.578125,
      "epoch": 2.976958525345622,
      "grad_norm": 8.330172538757324,
      "kl": 0.576171875,
      "learning_rate": 2.546153846153846e-07,
      "loss": 0.0956,
      "reward": 0.7594754099845886,
      "reward_std": 0.057717615738511086,
      "reward_std_all": 0.05509310029447079,
      "rewards-max/forensics_sinkhorn_reward": 0.9711673855781555,
      "rewards-max/forensics_strict_edge_reward": 0.2868872582912445,
      "rewards-max/forensics_tiou_reward": 0.9608927071094513,
      "rewards/forensics_sinkhorn_reward": 0.698628157377243,
      "rewards/forensics_strict_edge_reward": 0.026964273303747177,
      "rewards/forensics_tiou_reward": 0.49199673533439636,
      "step": 970
    },
    {
      "completion_length": 27.90625,
      "epoch": 2.980030721966206,
      "grad_norm": 7.289740085601807,
      "kl": 0.4365234375,
      "learning_rate": 2.538461538461538e-07,
      "loss": -0.0655,
      "reward": 0.7430791258811951,
      "reward_std": 0.07315962389111519,
      "reward_std_all": 0.0708240270614624,
      "rewards-max/forensics_sinkhorn_reward": 0.9829359352588654,
      "rewards-max/forensics_strict_edge_reward": 0.4154136925935745,
      "rewards-max/forensics_tiou_reward": 0.9849954545497894,
      "rewards/forensics_sinkhorn_reward": 0.6932599246501923,
      "rewards/forensics_strict_edge_reward": 0.04603692330420017,
      "rewards/forensics_tiou_reward": 0.5752106159925461,
      "step": 971
    },
    {
      "completion_length": 23.09375,
      "epoch": 2.9831029185867894,
      "grad_norm": 8.369033813476562,
      "kl": 0.45703125,
      "learning_rate": 2.530769230769231e-07,
      "loss": 0.0045,
      "reward": 0.7622338533401489,
      "reward_std": 0.05333455582149327,
      "reward_std_all": 0.05361155699938536,
      "rewards-max/forensics_sinkhorn_reward": 0.9765463173389435,
      "rewards-max/forensics_strict_edge_reward": 0.500888051756192,
      "rewards-max/forensics_tiou_reward": 0.969075083732605,
      "rewards/forensics_sinkhorn_reward": 0.7197300493717194,
      "rewards/forensics_strict_edge_reward": 0.04799827962415293,
      "rewards/forensics_tiou_reward": 0.545536071062088,
      "step": 972
    },
    {
      "completion_length": 28.6875,
      "epoch": 2.986175115207373,
      "grad_norm": 7.495294094085693,
      "kl": 0.501953125,
      "learning_rate": 2.523076923076923e-07,
      "loss": 0.0601,
      "reward": 0.7290346026420593,
      "reward_std": 0.07983504980802536,
      "reward_std_all": 0.08026107028126717,
      "rewards-max/forensics_sinkhorn_reward": 0.9679151475429535,
      "rewards-max/forensics_strict_edge_reward": 0.5337189733982086,
      "rewards-max/forensics_tiou_reward": 0.9547546207904816,
      "rewards/forensics_sinkhorn_reward": 0.6686730682849884,
      "rewards/forensics_strict_edge_reward": 0.04814748279750347,
      "rewards/forensics_tiou_reward": 0.5209978371858597,
      "step": 973
    },
    {
      "completion_length": 26.09375,
      "epoch": 2.989247311827957,
      "grad_norm": 6.683248996734619,
      "kl": 0.3603515625,
      "learning_rate": 2.515384615384615e-07,
      "loss": 0.0848,
      "reward": 0.7783689498901367,
      "reward_std": 0.031196131370961666,
      "reward_std_all": 0.030994687229394913,
      "rewards-max/forensics_sinkhorn_reward": 0.9770227372646332,
      "rewards-max/forensics_strict_edge_reward": 0.6254614591598511,
      "rewards-max/forensics_tiou_reward": 0.9663038551807404,
      "rewards/forensics_sinkhorn_reward": 0.7498216927051544,
      "rewards/forensics_strict_edge_reward": 0.08907464519143105,
      "rewards/forensics_tiou_reward": 0.6027576327323914,
      "step": 974
    },
    {
      "completion_length": 26.140625,
      "epoch": 2.9923195084485408,
      "grad_norm": 5.910890579223633,
      "kl": 0.4033203125,
      "learning_rate": 2.507692307692307e-07,
      "loss": 0.0107,
      "reward": 0.7884679436683655,
      "reward_std": 0.01977133611217141,
      "reward_std_all": 0.02073619212023914,
      "rewards-max/forensics_sinkhorn_reward": 0.9379538595676422,
      "rewards-max/forensics_strict_edge_reward": 0.28485091030597687,
      "rewards-max/forensics_tiou_reward": 0.8805476427078247,
      "rewards/forensics_sinkhorn_reward": 0.7613339424133301,
      "rewards/forensics_strict_edge_reward": 0.02250443398952484,
      "rewards/forensics_tiou_reward": 0.5938816666603088,
      "step": 975
    },
    {
      "completion_length": 26.90625,
      "epoch": 2.9953917050691246,
      "grad_norm": 14.797475814819336,
      "kl": 0.919921875,
      "learning_rate": 2.5e-07,
      "loss": 0.0468,
      "reward": 0.7526254653930664,
      "reward_std": 0.04192265495657921,
      "reward_std_all": 0.04984959214925766,
      "rewards-max/forensics_sinkhorn_reward": 0.9790851473808289,
      "rewards-max/forensics_strict_edge_reward": 0.732279509305954,
      "rewards-max/forensics_tiou_reward": 0.9749826192855835,
      "rewards/forensics_sinkhorn_reward": 0.7287679314613342,
      "rewards/forensics_strict_edge_reward": 0.10032219626009464,
      "rewards/forensics_tiou_reward": 0.5673039108514786,
      "step": 976
    },
    {
      "completion_length": 25.640625,
      "epoch": 2.998463901689708,
      "grad_norm": 6.769141674041748,
      "kl": 0.4326171875,
      "learning_rate": 2.492307692307692e-07,
      "loss": -0.0178,
      "reward": 0.78680419921875,
      "reward_std": 0.02899826830253005,
      "reward_std_all": 0.028097977861762047,
      "rewards-max/forensics_sinkhorn_reward": 0.9736434519290924,
      "rewards-max/forensics_strict_edge_reward": 0.43202342092990875,
      "rewards-max/forensics_tiou_reward": 0.9564531743526459,
      "rewards/forensics_sinkhorn_reward": 0.8092450201511383,
      "rewards/forensics_strict_edge_reward": 0.040807945653796196,
      "rewards/forensics_tiou_reward": 0.6201102435588837,
      "step": 977
    },
    {
      "completion_length": 45.400001525878906,
      "epoch": 3.0,
      "grad_norm": 6.789953708648682,
      "kl": 0.3359375,
      "learning_rate": 2.4846153846153847e-07,
      "loss": -0.0127,
      "reward": 0.6457938551902771,
      "reward_std": 0.1423122137784958,
      "reward_std_all": 0.1621311902999878,
      "rewards-max/forensics_sinkhorn_reward": 0.952204704284668,
      "rewards-max/forensics_strict_edge_reward": 0.15987974405288696,
      "rewards-max/forensics_tiou_reward": 0.9243769645690918,
      "rewards/forensics_sinkhorn_reward": 0.5259701013565063,
      "rewards/forensics_strict_edge_reward": 0.05272197723388672,
      "rewards/forensics_tiou_reward": 0.4872085154056549,
      "step": 978
    },
    {
      "completion_length": 26.703125,
      "epoch": 3.003072196620584,
      "grad_norm": 4.959333896636963,
      "kl": 0.474609375,
      "learning_rate": 2.476923076923077e-07,
      "loss": -0.0457,
      "reward": 0.7583388984203339,
      "reward_std": 0.06462671980261803,
      "reward_std_all": 0.060023700818419456,
      "rewards-max/forensics_sinkhorn_reward": 0.9866262674331665,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9923872947692871,
      "rewards/forensics_sinkhorn_reward": 0.7800576686859131,
      "rewards/forensics_strict_edge_reward": 0.1624867394566536,
      "rewards/forensics_tiou_reward": 0.6724132001399994,
      "step": 979
    },
    {
      "completion_length": 22.78125,
      "epoch": 3.0061443932411676,
      "grad_norm": 7.720793724060059,
      "kl": 0.45703125,
      "learning_rate": 2.469230769230769e-07,
      "loss": 0.0175,
      "reward": 0.7792612910270691,
      "reward_std": 0.029681597836315632,
      "reward_std_all": 0.031613211147487164,
      "rewards-max/forensics_sinkhorn_reward": 0.9769008159637451,
      "rewards-max/forensics_strict_edge_reward": 0.49170491099357605,
      "rewards-max/forensics_tiou_reward": 0.9657040536403656,
      "rewards/forensics_sinkhorn_reward": 0.7482587695121765,
      "rewards/forensics_strict_edge_reward": 0.0370264183729887,
      "rewards/forensics_tiou_reward": 0.4889727979898453,
      "step": 980
    },
    {
      "completion_length": 27.046875,
      "epoch": 3.0092165898617513,
      "grad_norm": 5.917934894561768,
      "kl": 0.4052734375,
      "learning_rate": 2.4615384615384616e-07,
      "loss": -0.0908,
      "reward": 0.7491249740123749,
      "reward_std": 0.02920160163193941,
      "reward_std_all": 0.04494953341782093,
      "rewards-max/forensics_sinkhorn_reward": 0.9670479893684387,
      "rewards-max/forensics_strict_edge_reward": 0.542974203824997,
      "rewards-max/forensics_tiou_reward": 0.9495201408863068,
      "rewards/forensics_sinkhorn_reward": 0.7244571447372437,
      "rewards/forensics_strict_edge_reward": 0.07542989403009415,
      "rewards/forensics_tiou_reward": 0.5893215239048004,
      "step": 981
    },
    {
      "completion_length": 27.34375,
      "epoch": 3.0122887864823347,
      "grad_norm": 7.1916937828063965,
      "kl": 0.361328125,
      "learning_rate": 2.453846153846154e-07,
      "loss": -0.029,
      "reward": 0.7591128647327423,
      "reward_std": 0.040023899637162685,
      "reward_std_all": 0.04359767632558942,
      "rewards-max/forensics_sinkhorn_reward": 0.9638025760650635,
      "rewards-max/forensics_strict_edge_reward": 0.4424971044063568,
      "rewards-max/forensics_tiou_reward": 0.936478853225708,
      "rewards/forensics_sinkhorn_reward": 0.7210189998149872,
      "rewards/forensics_strict_edge_reward": 0.0560453562065959,
      "rewards/forensics_tiou_reward": 0.5884034931659698,
      "step": 982
    },
    {
      "completion_length": 27.234375,
      "epoch": 3.0153609831029184,
      "grad_norm": 5.775172233581543,
      "kl": 0.46484375,
      "learning_rate": 2.446153846153846e-07,
      "loss": 0.0069,
      "reward": 0.7827566266059875,
      "reward_std": 0.016328588942997158,
      "reward_std_all": 0.020236500073224306,
      "rewards-max/forensics_sinkhorn_reward": 0.9837977886199951,
      "rewards-max/forensics_strict_edge_reward": 0.6084719300270081,
      "rewards-max/forensics_tiou_reward": 0.9856082499027252,
      "rewards/forensics_sinkhorn_reward": 0.7315722107887268,
      "rewards/forensics_strict_edge_reward": 0.06837034225463867,
      "rewards/forensics_tiou_reward": 0.5351631939411163,
      "step": 983
    },
    {
      "completion_length": 25.8125,
      "epoch": 3.0184331797235022,
      "grad_norm": 5.331113815307617,
      "kl": 0.3955078125,
      "learning_rate": 2.4384615384615386e-07,
      "loss": 0.0255,
      "reward": 0.7757098376750946,
      "reward_std": 0.01914827188011259,
      "reward_std_all": 0.02837231638841331,
      "rewards-max/forensics_sinkhorn_reward": 0.9695622324943542,
      "rewards-max/forensics_strict_edge_reward": 0.5450693517923355,
      "rewards-max/forensics_tiou_reward": 0.9519131183624268,
      "rewards/forensics_sinkhorn_reward": 0.7713768184185028,
      "rewards/forensics_strict_edge_reward": 0.04149833135306835,
      "rewards/forensics_tiou_reward": 0.5619720220565796,
      "step": 984
    },
    {
      "completion_length": 26.125,
      "epoch": 3.021505376344086,
      "grad_norm": 6.74477481842041,
      "kl": 0.4306640625,
      "learning_rate": 2.4307692307692307e-07,
      "loss": 0.0046,
      "reward": 0.7936764359474182,
      "reward_std": 0.021488572470843792,
      "reward_std_all": 0.020441648550331593,
      "rewards-max/forensics_sinkhorn_reward": 0.9734346866607666,
      "rewards-max/forensics_strict_edge_reward": 0.4649401903152466,
      "rewards-max/forensics_tiou_reward": 0.9590660631656647,
      "rewards/forensics_sinkhorn_reward": 0.8152184784412384,
      "rewards/forensics_strict_edge_reward": 0.05453978478908539,
      "rewards/forensics_tiou_reward": 0.6373121738433838,
      "step": 985
    },
    {
      "completion_length": 23.25,
      "epoch": 3.02457757296467,
      "grad_norm": 6.1330647468566895,
      "kl": 0.537109375,
      "learning_rate": 2.423076923076923e-07,
      "loss": -0.0313,
      "reward": 0.8042809665203094,
      "reward_std": 0.0017370268469676375,
      "reward_std_all": 0.004069990012794733,
      "rewards-max/forensics_sinkhorn_reward": 0.9845693707466125,
      "rewards-max/forensics_strict_edge_reward": 0.8205508291721344,
      "rewards-max/forensics_tiou_reward": 0.9880893528461456,
      "rewards/forensics_sinkhorn_reward": 0.7587930560112,
      "rewards/forensics_strict_edge_reward": 0.07652410492300987,
      "rewards/forensics_tiou_reward": 0.5600331127643585,
      "step": 986
    },
    {
      "completion_length": 24.6875,
      "epoch": 3.0276497695852536,
      "grad_norm": 5.716928482055664,
      "kl": 0.466796875,
      "learning_rate": 2.415384615384615e-07,
      "loss": 0.1341,
      "reward": 0.7715879380702972,
      "reward_std": 0.049929749220609665,
      "reward_std_all": 0.046334611251950264,
      "rewards-max/forensics_sinkhorn_reward": 0.9816415011882782,
      "rewards-max/forensics_strict_edge_reward": 0.6390409171581268,
      "rewards-max/forensics_tiou_reward": 0.979485809803009,
      "rewards/forensics_sinkhorn_reward": 0.7293767631053925,
      "rewards/forensics_strict_edge_reward": 0.0626205075532198,
      "rewards/forensics_tiou_reward": 0.5515097379684448,
      "step": 987
    },
    {
      "completion_length": 27.015625,
      "epoch": 3.0307219662058373,
      "grad_norm": 7.259721279144287,
      "kl": 0.5439453125,
      "learning_rate": 2.4076923076923076e-07,
      "loss": 0.0482,
      "reward": 0.7317168116569519,
      "reward_std": 0.03965841792523861,
      "reward_std_all": 0.05682964250445366,
      "rewards-max/forensics_sinkhorn_reward": 0.9532777369022369,
      "rewards-max/forensics_strict_edge_reward": 0.2507291752845049,
      "rewards-max/forensics_tiou_reward": 0.8968973457813263,
      "rewards/forensics_sinkhorn_reward": 0.7045575678348541,
      "rewards/forensics_strict_edge_reward": 0.04087254451587796,
      "rewards/forensics_tiou_reward": 0.5280151814222336,
      "step": 988
    },
    {
      "completion_length": 28.15625,
      "epoch": 3.033794162826421,
      "grad_norm": 5.311932563781738,
      "kl": 0.3623046875,
      "learning_rate": 2.4e-07,
      "loss": 0.0157,
      "reward": 0.7616330087184906,
      "reward_std": 0.05318225547671318,
      "reward_std_all": 0.05163053050637245,
      "rewards-max/forensics_sinkhorn_reward": 0.9824658036231995,
      "rewards-max/forensics_strict_edge_reward": 0.7709200084209442,
      "rewards-max/forensics_tiou_reward": 0.981179267168045,
      "rewards/forensics_sinkhorn_reward": 0.734519749879837,
      "rewards/forensics_strict_edge_reward": 0.06562213785946369,
      "rewards/forensics_tiou_reward": 0.6267839074134827,
      "step": 989
    },
    {
      "completion_length": 27.71875,
      "epoch": 3.0368663594470044,
      "grad_norm": 5.9007182121276855,
      "kl": 0.361328125,
      "learning_rate": 2.392307692307692e-07,
      "loss": 0.0203,
      "reward": 0.7876751124858856,
      "reward_std": 0.02279149554669857,
      "reward_std_all": 0.022793108597397804,
      "rewards-max/forensics_sinkhorn_reward": 0.9735144376754761,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.961983859539032,
      "rewards/forensics_sinkhorn_reward": 0.7518522143363953,
      "rewards/forensics_strict_edge_reward": 0.06898736208677292,
      "rewards/forensics_tiou_reward": 0.5634464621543884,
      "step": 990
    },
    {
      "completion_length": 24.375,
      "epoch": 3.0399385560675882,
      "grad_norm": 5.443921089172363,
      "kl": 0.416015625,
      "learning_rate": 2.3846153846153846e-07,
      "loss": -0.0622,
      "reward": 0.7881956100463867,
      "reward_std": 0.004593126475811005,
      "reward_std_all": 0.01376744220033288,
      "rewards-max/forensics_sinkhorn_reward": 0.9796591401100159,
      "rewards-max/forensics_strict_edge_reward": 0.5428982526063919,
      "rewards-max/forensics_tiou_reward": 0.9728708863258362,
      "rewards/forensics_sinkhorn_reward": 0.7974345982074738,
      "rewards/forensics_strict_edge_reward": 0.05237432010471821,
      "rewards/forensics_tiou_reward": 0.6353017389774323,
      "step": 991
    },
    {
      "completion_length": 22.375,
      "epoch": 3.043010752688172,
      "grad_norm": 7.135500907897949,
      "kl": 0.359375,
      "learning_rate": 2.376923076923077e-07,
      "loss": 0.0309,
      "reward": 0.7342794239521027,
      "reward_std": 0.07363515347242355,
      "reward_std_all": 0.07292628660798073,
      "rewards-max/forensics_sinkhorn_reward": 0.9836828410625458,
      "rewards-max/forensics_strict_edge_reward": 0.5798844993114471,
      "rewards-max/forensics_tiou_reward": 0.9840636253356934,
      "rewards/forensics_sinkhorn_reward": 0.6930269300937653,
      "rewards/forensics_strict_edge_reward": 0.068010114133358,
      "rewards/forensics_tiou_reward": 0.6013738214969635,
      "step": 992
    },
    {
      "completion_length": 27.84375,
      "epoch": 3.046082949308756,
      "grad_norm": 6.169590473175049,
      "kl": 0.4677734375,
      "learning_rate": 2.369230769230769e-07,
      "loss": 0.0033,
      "reward": 0.7931741774082184,
      "reward_std": 0.02350632380694151,
      "reward_std_all": 0.021934863179922104,
      "rewards-max/forensics_sinkhorn_reward": 0.9881683588027954,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9976514279842377,
      "rewards/forensics_sinkhorn_reward": 0.8018482029438019,
      "rewards/forensics_strict_edge_reward": 0.09957898780703545,
      "rewards/forensics_tiou_reward": 0.6848037242889404,
      "step": 993
    },
    {
      "completion_length": 23.71875,
      "epoch": 3.0491551459293396,
      "grad_norm": 7.383646011352539,
      "kl": 0.3974609375,
      "learning_rate": 2.3615384615384615e-07,
      "loss": -0.0246,
      "reward": 0.7882435619831085,
      "reward_std": 0.019959948491305113,
      "reward_std_all": 0.020694226026535034,
      "rewards-max/forensics_sinkhorn_reward": 0.9624886214733124,
      "rewards-max/forensics_strict_edge_reward": 0.47060726583004,
      "rewards-max/forensics_tiou_reward": 0.9383303225040436,
      "rewards/forensics_sinkhorn_reward": 0.7870935797691345,
      "rewards/forensics_strict_edge_reward": 0.06390727404505014,
      "rewards/forensics_tiou_reward": 0.6133965253829956,
      "step": 994
    },
    {
      "completion_length": 27.625,
      "epoch": 3.0522273425499233,
      "grad_norm": 7.402097702026367,
      "kl": 0.841796875,
      "learning_rate": 2.3538461538461537e-07,
      "loss": 0.1325,
      "reward": 0.777203381061554,
      "reward_std": 0.04489178769290447,
      "reward_std_all": 0.04120407812297344,
      "rewards-max/forensics_sinkhorn_reward": 0.9838225245475769,
      "rewards-max/forensics_strict_edge_reward": 0.711992472410202,
      "rewards-max/forensics_tiou_reward": 0.9861078262329102,
      "rewards/forensics_sinkhorn_reward": 0.7461260557174683,
      "rewards/forensics_strict_edge_reward": 0.06251674890518188,
      "rewards/forensics_tiou_reward": 0.578484982252121,
      "step": 995
    },
    {
      "completion_length": 23.8125,
      "epoch": 3.055299539170507,
      "grad_norm": 9.460627555847168,
      "kl": 0.3466796875,
      "learning_rate": 2.346153846153846e-07,
      "loss": 0.0729,
      "reward": 0.7703458368778229,
      "reward_std": 0.033436913043260574,
      "reward_std_all": 0.03661218471825123,
      "rewards-max/forensics_sinkhorn_reward": 0.9774574637413025,
      "rewards-max/forensics_strict_edge_reward": 0.6858793199062347,
      "rewards-max/forensics_tiou_reward": 0.9674629867076874,
      "rewards/forensics_sinkhorn_reward": 0.7489789128303528,
      "rewards/forensics_strict_edge_reward": 0.06656774878501892,
      "rewards/forensics_tiou_reward": 0.5939091145992279,
      "step": 996
    },
    {
      "completion_length": 23.4375,
      "epoch": 3.0583717357910905,
      "grad_norm": 7.666184425354004,
      "kl": 0.3427734375,
      "learning_rate": 2.3384615384615382e-07,
      "loss": 0.013,
      "reward": 0.7976861894130707,
      "reward_std": 0.01347716012969613,
      "reward_std_all": 0.013524901121854782,
      "rewards-max/forensics_sinkhorn_reward": 0.9863299131393433,
      "rewards-max/forensics_strict_edge_reward": 0.7727619707584381,
      "rewards-max/forensics_tiou_reward": 0.9932603538036346,
      "rewards/forensics_sinkhorn_reward": 0.8029391467571259,
      "rewards/forensics_strict_edge_reward": 0.09224163740873337,
      "rewards/forensics_tiou_reward": 0.7024266421794891,
      "step": 997
    },
    {
      "completion_length": 22.765625,
      "epoch": 3.0614439324116742,
      "grad_norm": 7.671378135681152,
      "kl": 0.3955078125,
      "learning_rate": 2.3307692307692306e-07,
      "loss": 0.021,
      "reward": 0.7388734221458435,
      "reward_std": 0.04700673930346966,
      "reward_std_all": 0.05533293914049864,
      "rewards-max/forensics_sinkhorn_reward": 0.9816805422306061,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9803777933120728,
      "rewards/forensics_sinkhorn_reward": 0.7590557932853699,
      "rewards/forensics_strict_edge_reward": 0.08677707985043526,
      "rewards/forensics_tiou_reward": 0.6043465435504913,
      "step": 998
    },
    {
      "completion_length": 25.03125,
      "epoch": 3.064516129032258,
      "grad_norm": 7.003360271453857,
      "kl": 0.36328125,
      "learning_rate": 2.323076923076923e-07,
      "loss": -0.0828,
      "reward": 0.7394166588783264,
      "reward_std": 0.03539008158259094,
      "reward_std_all": 0.05119227757677436,
      "rewards-max/forensics_sinkhorn_reward": 0.934757262468338,
      "rewards-max/forensics_strict_edge_reward": 0.3563908040523529,
      "rewards-max/forensics_tiou_reward": 0.8671800494194031,
      "rewards/forensics_sinkhorn_reward": 0.7166878283023834,
      "rewards/forensics_strict_edge_reward": 0.026496575213968754,
      "rewards/forensics_tiou_reward": 0.5384787768125534,
      "step": 999
    },
    {
      "completion_length": 27.03125,
      "epoch": 3.067588325652842,
      "grad_norm": 7.328857898712158,
      "kl": 0.3486328125,
      "learning_rate": 2.3153846153846151e-07,
      "loss": 0.0246,
      "reward": 0.7877543270587921,
      "reward_std": 0.018549500033259392,
      "reward_std_all": 0.020084666088223457,
      "rewards-max/forensics_sinkhorn_reward": 0.980575293302536,
      "rewards-max/forensics_strict_edge_reward": 0.5611837804317474,
      "rewards-max/forensics_tiou_reward": 0.9733076393604279,
      "rewards/forensics_sinkhorn_reward": 0.7969784140586853,
      "rewards/forensics_strict_edge_reward": 0.07388940826058388,
      "rewards/forensics_tiou_reward": 0.6340926289558411,
      "step": 1000
    },
    {
      "completion_length": 24.921875,
      "epoch": 3.0706605222734256,
      "grad_norm": 7.923988342285156,
      "kl": 0.4599609375,
      "learning_rate": 2.3076923076923078e-07,
      "loss": -0.1565,
      "reward": 0.7649178504943848,
      "reward_std": 0.022441738285124302,
      "reward_std_all": 0.03210772667080164,
      "rewards-max/forensics_sinkhorn_reward": 0.9858047962188721,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9913252890110016,
      "rewards/forensics_sinkhorn_reward": 0.811410665512085,
      "rewards/forensics_strict_edge_reward": 0.10716601088643074,
      "rewards/forensics_tiou_reward": 0.6656777262687683,
      "step": 1001
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.0737327188940093,
      "grad_norm": 7.003560543060303,
      "kl": 0.5078125,
      "learning_rate": 2.3e-07,
      "loss": 0.0791,
      "reward": 0.7710753977298737,
      "reward_std": 0.05215002782642841,
      "reward_std_all": 0.049032729119062424,
      "rewards-max/forensics_sinkhorn_reward": 0.9771219491958618,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9694568812847137,
      "rewards/forensics_sinkhorn_reward": 0.7442140579223633,
      "rewards/forensics_strict_edge_reward": 0.050317876040935516,
      "rewards/forensics_tiou_reward": 0.5827845931053162,
      "step": 1002
    },
    {
      "completion_length": 28.390625,
      "epoch": 3.076804915514593,
      "grad_norm": 7.061075210571289,
      "kl": 0.390625,
      "learning_rate": 2.2923076923076924e-07,
      "loss": 0.0295,
      "reward": 0.7898973524570465,
      "reward_std": 0.018348817713558674,
      "reward_std_all": 0.019082793965935707,
      "rewards-max/forensics_sinkhorn_reward": 0.9782528281211853,
      "rewards-max/forensics_strict_edge_reward": 0.648295670747757,
      "rewards-max/forensics_tiou_reward": 0.9729925096035004,
      "rewards/forensics_sinkhorn_reward": 0.7874289751052856,
      "rewards/forensics_strict_edge_reward": 0.053695086389780045,
      "rewards/forensics_tiou_reward": 0.6193864047527313,
      "step": 1003
    },
    {
      "completion_length": 25.40625,
      "epoch": 3.0798771121351765,
      "grad_norm": 6.717024803161621,
      "kl": 0.4111328125,
      "learning_rate": 2.2846153846153845e-07,
      "loss": 0.0176,
      "reward": 0.7751254141330719,
      "reward_std": 0.023421993479132652,
      "reward_std_all": 0.027735531330108643,
      "rewards-max/forensics_sinkhorn_reward": 0.9854898750782013,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9903962314128876,
      "rewards/forensics_sinkhorn_reward": 0.7805410623550415,
      "rewards/forensics_strict_edge_reward": 0.07166671380400658,
      "rewards/forensics_tiou_reward": 0.6478052735328674,
      "step": 1004
    },
    {
      "completion_length": 24.6875,
      "epoch": 3.0829493087557602,
      "grad_norm": 6.646816253662109,
      "kl": 0.4306640625,
      "learning_rate": 2.276923076923077e-07,
      "loss": 0.0074,
      "reward": 0.7948185205459595,
      "reward_std": 0.006585415219888091,
      "reward_std_all": 0.010670221876353025,
      "rewards-max/forensics_sinkhorn_reward": 0.9811517298221588,
      "rewards-max/forensics_strict_edge_reward": 0.7169294059276581,
      "rewards-max/forensics_tiou_reward": 0.9816878139972687,
      "rewards/forensics_sinkhorn_reward": 0.8136547207832336,
      "rewards/forensics_strict_edge_reward": 0.07384081371128559,
      "rewards/forensics_tiou_reward": 0.6509673893451691,
      "step": 1005
    },
    {
      "completion_length": 24.4375,
      "epoch": 3.086021505376344,
      "grad_norm": 7.82511043548584,
      "kl": 0.5361328125,
      "learning_rate": 2.269230769230769e-07,
      "loss": 0.0154,
      "reward": 0.751520425081253,
      "reward_std": 0.03427379298955202,
      "reward_std_all": 0.04467732645571232,
      "rewards-max/forensics_sinkhorn_reward": 0.9816468954086304,
      "rewards-max/forensics_strict_edge_reward": 0.5101321190595627,
      "rewards-max/forensics_tiou_reward": 0.9819508492946625,
      "rewards/forensics_sinkhorn_reward": 0.7288010716438293,
      "rewards/forensics_strict_edge_reward": 0.06439515668898821,
      "rewards/forensics_tiou_reward": 0.5648376494646072,
      "step": 1006
    },
    {
      "completion_length": 28.25,
      "epoch": 3.089093701996928,
      "grad_norm": 7.651692867279053,
      "kl": 0.423828125,
      "learning_rate": 2.2615384615384614e-07,
      "loss": 0.0278,
      "reward": 0.7939517498016357,
      "reward_std": 0.01635875552892685,
      "reward_std_all": 0.01652663154527545,
      "rewards-max/forensics_sinkhorn_reward": 0.9765446186065674,
      "rewards-max/forensics_strict_edge_reward": 0.5847881734371185,
      "rewards-max/forensics_tiou_reward": 0.9686045050621033,
      "rewards/forensics_sinkhorn_reward": 0.777296781539917,
      "rewards/forensics_strict_edge_reward": 0.06536553613841534,
      "rewards/forensics_tiou_reward": 0.6147065758705139,
      "step": 1007
    },
    {
      "completion_length": 24.15625,
      "epoch": 3.0921658986175116,
      "grad_norm": 6.68137788772583,
      "kl": 0.458984375,
      "learning_rate": 2.2538461538461536e-07,
      "loss": -0.0139,
      "reward": 0.8010082244873047,
      "reward_std": 0.009938985574990511,
      "reward_std_all": 0.010356903309002519,
      "rewards-max/forensics_sinkhorn_reward": 0.9835589826107025,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9868846833705902,
      "rewards/forensics_sinkhorn_reward": 0.8226138353347778,
      "rewards/forensics_strict_edge_reward": 0.12305283918976784,
      "rewards/forensics_tiou_reward": 0.6767447888851166,
      "step": 1008
    },
    {
      "completion_length": 24.296875,
      "epoch": 3.0952380952380953,
      "grad_norm": 6.800692081451416,
      "kl": 0.388671875,
      "learning_rate": 2.246153846153846e-07,
      "loss": -0.0781,
      "reward": 0.7979781627655029,
      "reward_std": 0.005417791486252099,
      "reward_std_all": 0.008354698540642858,
      "rewards-max/forensics_sinkhorn_reward": 0.9782651364803314,
      "rewards-max/forensics_strict_edge_reward": 0.5996775329113007,
      "rewards-max/forensics_tiou_reward": 0.9719991385936737,
      "rewards/forensics_sinkhorn_reward": 0.8156782686710358,
      "rewards/forensics_strict_edge_reward": 0.059946175664663315,
      "rewards/forensics_tiou_reward": 0.6396426856517792,
      "step": 1009
    },
    {
      "completion_length": 26.46875,
      "epoch": 3.098310291858679,
      "grad_norm": 6.90981912612915,
      "kl": 0.30859375,
      "learning_rate": 2.238461538461538e-07,
      "loss": -0.0168,
      "reward": 0.7820433080196381,
      "reward_std": 0.024989839643239975,
      "reward_std_all": 0.02579499501734972,
      "rewards-max/forensics_sinkhorn_reward": 0.9828394055366516,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9847594201564789,
      "rewards/forensics_sinkhorn_reward": 0.7349554598331451,
      "rewards/forensics_strict_edge_reward": 0.09728322923183441,
      "rewards/forensics_tiou_reward": 0.643566906452179,
      "step": 1010
    },
    {
      "completion_length": 26.046875,
      "epoch": 3.1013824884792625,
      "grad_norm": 10.07581901550293,
      "kl": 0.4794921875,
      "learning_rate": 2.2307692307692308e-07,
      "loss": -0.0007,
      "reward": 0.8011649250984192,
      "reward_std": 0.00943698303308338,
      "reward_std_all": 0.010170570574700832,
      "rewards-max/forensics_sinkhorn_reward": 0.9847974181175232,
      "rewards-max/forensics_strict_edge_reward": 0.8106843829154968,
      "rewards-max/forensics_tiou_reward": 0.9873700439929962,
      "rewards/forensics_sinkhorn_reward": 0.8386944532394409,
      "rewards/forensics_strict_edge_reward": 0.07515140622854233,
      "rewards/forensics_tiou_reward": 0.7244249880313873,
      "step": 1011
    },
    {
      "completion_length": 26.921875,
      "epoch": 3.1044546850998462,
      "grad_norm": 6.623039245605469,
      "kl": 0.3515625,
      "learning_rate": 2.2230769230769232e-07,
      "loss": -0.0697,
      "reward": 0.7927201986312866,
      "reward_std": 0.012995188008062541,
      "reward_std_all": 0.014578230679035187,
      "rewards-max/forensics_sinkhorn_reward": 0.9790248572826385,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9731108248233795,
      "rewards/forensics_sinkhorn_reward": 0.808045357465744,
      "rewards/forensics_strict_edge_reward": 0.07952362485229969,
      "rewards/forensics_tiou_reward": 0.6333538889884949,
      "step": 1012
    },
    {
      "completion_length": 23.09375,
      "epoch": 3.10752688172043,
      "grad_norm": 8.892803192138672,
      "kl": 0.59765625,
      "learning_rate": 2.2153846153846153e-07,
      "loss": 0.0679,
      "reward": 0.7902256548404694,
      "reward_std": 0.018521045334637165,
      "reward_std_all": 0.019365476444363594,
      "rewards-max/forensics_sinkhorn_reward": 0.9729085862636566,
      "rewards-max/forensics_strict_edge_reward": 0.4254278540611267,
      "rewards-max/forensics_tiou_reward": 0.9594883322715759,
      "rewards/forensics_sinkhorn_reward": 0.7539655566215515,
      "rewards/forensics_strict_edge_reward": 0.04137061722576618,
      "rewards/forensics_tiou_reward": 0.5788467228412628,
      "step": 1013
    },
    {
      "completion_length": 24.8125,
      "epoch": 3.110599078341014,
      "grad_norm": 6.978785514831543,
      "kl": 0.41015625,
      "learning_rate": 2.2076923076923077e-07,
      "loss": -0.0097,
      "reward": 0.7393431067466736,
      "reward_std": 0.061101729050278664,
      "reward_std_all": 0.06692987959831953,
      "rewards-max/forensics_sinkhorn_reward": 0.981665700674057,
      "rewards-max/forensics_strict_edge_reward": 0.5905451774597168,
      "rewards-max/forensics_tiou_reward": 0.9794435203075409,
      "rewards/forensics_sinkhorn_reward": 0.735967218875885,
      "rewards/forensics_strict_edge_reward": 0.055809155106544495,
      "rewards/forensics_tiou_reward": 0.5769125521183014,
      "step": 1014
    },
    {
      "completion_length": 24.0625,
      "epoch": 3.1136712749615976,
      "grad_norm": 9.499916076660156,
      "kl": 0.4453125,
      "learning_rate": 2.1999999999999998e-07,
      "loss": 0.0182,
      "reward": 0.7842219471931458,
      "reward_std": 0.02815741952508688,
      "reward_std_all": 0.028189199045300484,
      "rewards-max/forensics_sinkhorn_reward": 0.9752615690231323,
      "rewards-max/forensics_strict_edge_reward": 0.45014238357543945,
      "rewards-max/forensics_tiou_reward": 0.9649885594844818,
      "rewards/forensics_sinkhorn_reward": 0.7661046981811523,
      "rewards/forensics_strict_edge_reward": 0.05514029320329428,
      "rewards/forensics_tiou_reward": 0.6143540740013123,
      "step": 1015
    },
    {
      "completion_length": 26.265625,
      "epoch": 3.1167434715821813,
      "grad_norm": 8.420561790466309,
      "kl": 0.537109375,
      "learning_rate": 2.1923076923076922e-07,
      "loss": 0.0005,
      "reward": 0.7930474877357483,
      "reward_std": 0.016180745093151927,
      "reward_std_all": 0.016743574989959598,
      "rewards-max/forensics_sinkhorn_reward": 0.9781449139118195,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9711476862430573,
      "rewards/forensics_sinkhorn_reward": 0.8070593774318695,
      "rewards/forensics_strict_edge_reward": 0.04402648285031319,
      "rewards/forensics_tiou_reward": 0.6509575843811035,
      "step": 1016
    },
    {
      "completion_length": 22.8125,
      "epoch": 3.119815668202765,
      "grad_norm": 6.706992149353027,
      "kl": 0.4482421875,
      "learning_rate": 2.1846153846153844e-07,
      "loss": 0.0082,
      "reward": 0.789323627948761,
      "reward_std": 0.029145733453333378,
      "reward_std_all": 0.02728074137121439,
      "rewards-max/forensics_sinkhorn_reward": 0.9754561483860016,
      "rewards-max/forensics_strict_edge_reward": 0.44632039964199066,
      "rewards-max/forensics_tiou_reward": 0.9678931534290314,
      "rewards/forensics_sinkhorn_reward": 0.8108571171760559,
      "rewards/forensics_strict_edge_reward": 0.07083212211728096,
      "rewards/forensics_tiou_reward": 0.617788702249527,
      "step": 1017
    },
    {
      "completion_length": 27.125,
      "epoch": 3.122887864823349,
      "grad_norm": 6.364751815795898,
      "kl": 0.3115234375,
      "learning_rate": 2.1769230769230768e-07,
      "loss": -0.0916,
      "reward": 0.8025465905666351,
      "reward_std": 0.007537896861322224,
      "reward_std_all": 0.008358046761713922,
      "rewards-max/forensics_sinkhorn_reward": 0.985928863286972,
      "rewards-max/forensics_strict_edge_reward": 0.6284831166267395,
      "rewards-max/forensics_tiou_reward": 0.991470605134964,
      "rewards/forensics_sinkhorn_reward": 0.8566762208938599,
      "rewards/forensics_strict_edge_reward": 0.08063226565718651,
      "rewards/forensics_tiou_reward": 0.7208655476570129,
      "step": 1018
    },
    {
      "completion_length": 23.734375,
      "epoch": 3.1259600614439322,
      "grad_norm": 8.180124282836914,
      "kl": 0.544921875,
      "learning_rate": 2.169230769230769e-07,
      "loss": 0.1329,
      "reward": 0.7846002578735352,
      "reward_std": 0.028543259482830763,
      "reward_std_all": 0.027554437518119812,
      "rewards-max/forensics_sinkhorn_reward": 0.9824437201023102,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9836185276508331,
      "rewards/forensics_sinkhorn_reward": 0.769586592912674,
      "rewards/forensics_strict_edge_reward": 0.10542835295200348,
      "rewards/forensics_tiou_reward": 0.64137402176857,
      "step": 1019
    },
    {
      "completion_length": 27.359375,
      "epoch": 3.129032258064516,
      "grad_norm": 9.39953327178955,
      "kl": 0.642578125,
      "learning_rate": 2.1615384615384613e-07,
      "loss": 0.0667,
      "reward": 0.779457300901413,
      "reward_std": 0.03324119281023741,
      "reward_std_all": 0.03243981208652258,
      "rewards-max/forensics_sinkhorn_reward": 0.9762499034404755,
      "rewards-max/forensics_strict_edge_reward": 0.351301833987236,
      "rewards-max/forensics_tiou_reward": 0.966606467962265,
      "rewards/forensics_sinkhorn_reward": 0.7768586277961731,
      "rewards/forensics_strict_edge_reward": 0.04815663676708937,
      "rewards/forensics_tiou_reward": 0.5957340896129608,
      "step": 1020
    },
    {
      "completion_length": 26.421875,
      "epoch": 3.1321044546851,
      "grad_norm": 6.95152473449707,
      "kl": 0.6279296875,
      "learning_rate": 2.153846153846154e-07,
      "loss": 0.0919,
      "reward": 0.785031259059906,
      "reward_std": 0.030555421486496925,
      "reward_std_all": 0.02901457902044058,
      "rewards-max/forensics_sinkhorn_reward": 0.97600057721138,
      "rewards-max/forensics_strict_edge_reward": 0.4661562591791153,
      "rewards-max/forensics_tiou_reward": 0.9693961441516876,
      "rewards/forensics_sinkhorn_reward": 0.7814005017280579,
      "rewards/forensics_strict_edge_reward": 0.056371111422777176,
      "rewards/forensics_tiou_reward": 0.6100780963897705,
      "step": 1021
    },
    {
      "completion_length": 25.546875,
      "epoch": 3.1351766513056836,
      "grad_norm": 9.797558784484863,
      "kl": 0.3984375,
      "learning_rate": 2.146153846153846e-07,
      "loss": -0.0529,
      "reward": 0.8009800016880035,
      "reward_std": 0.006449623499065638,
      "reward_std_all": 0.008175985887646675,
      "rewards-max/forensics_sinkhorn_reward": 0.983436644077301,
      "rewards-max/forensics_strict_edge_reward": 0.5373988002538681,
      "rewards-max/forensics_tiou_reward": 0.9831276535987854,
      "rewards/forensics_sinkhorn_reward": 0.8228675425052643,
      "rewards/forensics_strict_edge_reward": 0.057185420766472816,
      "rewards/forensics_tiou_reward": 0.6884374618530273,
      "step": 1022
    },
    {
      "completion_length": 24.40625,
      "epoch": 3.1382488479262673,
      "grad_norm": 7.197600841522217,
      "kl": 0.4287109375,
      "learning_rate": 2.1384615384615385e-07,
      "loss": -0.0403,
      "reward": 0.7805612981319427,
      "reward_std": 0.02542073279619217,
      "reward_std_all": 0.02696822304278612,
      "rewards-max/forensics_sinkhorn_reward": 0.9869901537895203,
      "rewards-max/forensics_strict_edge_reward": 0.6232984811067581,
      "rewards-max/forensics_tiou_reward": 0.9934368431568146,
      "rewards/forensics_sinkhorn_reward": 0.7836914658546448,
      "rewards/forensics_strict_edge_reward": 0.06127459742128849,
      "rewards/forensics_tiou_reward": 0.6367023587226868,
      "step": 1023
    },
    {
      "completion_length": 23.6875,
      "epoch": 3.141321044546851,
      "grad_norm": 6.559778213500977,
      "kl": 0.544921875,
      "learning_rate": 2.1307692307692307e-07,
      "loss": -0.021,
      "reward": 0.7955471873283386,
      "reward_std": 0.01766378479078412,
      "reward_std_all": 0.017516242107376456,
      "rewards-max/forensics_sinkhorn_reward": 0.975926011800766,
      "rewards-max/forensics_strict_edge_reward": 0.435063898563385,
      "rewards-max/forensics_tiou_reward": 0.9668458998203278,
      "rewards/forensics_sinkhorn_reward": 0.8101006746292114,
      "rewards/forensics_strict_edge_reward": 0.035667150281369686,
      "rewards/forensics_tiou_reward": 0.6488651633262634,
      "step": 1024
    },
    {
      "completion_length": 23.5,
      "epoch": 3.144393241167435,
      "grad_norm": 10.054215431213379,
      "kl": 0.5068359375,
      "learning_rate": 2.123076923076923e-07,
      "loss": 0.0305,
      "reward": 0.7816548049449921,
      "reward_std": 0.04087766073644161,
      "reward_std_all": 0.037606608122587204,
      "rewards-max/forensics_sinkhorn_reward": 0.9825779497623444,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9825150966644287,
      "rewards/forensics_sinkhorn_reward": 0.746067225933075,
      "rewards/forensics_strict_edge_reward": 0.07471876218914986,
      "rewards/forensics_tiou_reward": 0.5337783545255661,
      "step": 1025
    },
    {
      "completion_length": 24.0625,
      "epoch": 3.1474654377880182,
      "grad_norm": 7.817506790161133,
      "kl": 0.4462890625,
      "learning_rate": 2.1153846153846152e-07,
      "loss": 0.0002,
      "reward": 0.7829635441303253,
      "reward_std": 0.03308263886719942,
      "reward_std_all": 0.03156796470284462,
      "rewards-max/forensics_sinkhorn_reward": 0.9782800078392029,
      "rewards-max/forensics_strict_edge_reward": 0.6153465881943703,
      "rewards-max/forensics_tiou_reward": 0.9684367775917053,
      "rewards/forensics_sinkhorn_reward": 0.780279129743576,
      "rewards/forensics_strict_edge_reward": 0.04872526042163372,
      "rewards/forensics_tiou_reward": 0.6162032783031464,
      "step": 1026
    },
    {
      "completion_length": 26.703125,
      "epoch": 3.150537634408602,
      "grad_norm": 8.109024047851562,
      "kl": 0.529296875,
      "learning_rate": 2.1076923076923076e-07,
      "loss": 0.1268,
      "reward": 0.7295166850090027,
      "reward_std": 0.05482583865523338,
      "reward_std_all": 0.06995837762951851,
      "rewards-max/forensics_sinkhorn_reward": 0.9778470396995544,
      "rewards-max/forensics_strict_edge_reward": 0.41550540924072266,
      "rewards-max/forensics_tiou_reward": 0.9725184738636017,
      "rewards/forensics_sinkhorn_reward": 0.7247463762760162,
      "rewards/forensics_strict_edge_reward": 0.05281299538910389,
      "rewards/forensics_tiou_reward": 0.5579224526882172,
      "step": 1027
    },
    {
      "completion_length": 22.265625,
      "epoch": 3.153609831029186,
      "grad_norm": 9.414910316467285,
      "kl": 0.3115234375,
      "learning_rate": 2.0999999999999997e-07,
      "loss": 0.0952,
      "reward": 0.7636343240737915,
      "reward_std": 0.04302801191806793,
      "reward_std_all": 0.04323320835828781,
      "rewards-max/forensics_sinkhorn_reward": 0.9822184145450592,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9832811653614044,
      "rewards/forensics_sinkhorn_reward": 0.7370770573616028,
      "rewards/forensics_strict_edge_reward": 0.0754009298980236,
      "rewards/forensics_tiou_reward": 0.6262103021144867,
      "step": 1028
    },
    {
      "completion_length": 27.3125,
      "epoch": 3.1566820276497696,
      "grad_norm": 8.301966667175293,
      "kl": 0.412109375,
      "learning_rate": 2.0923076923076921e-07,
      "loss": -0.0142,
      "reward": 0.8008908331394196,
      "reward_std": 0.0064716937486082315,
      "reward_std_all": 0.00776383513584733,
      "rewards-max/forensics_sinkhorn_reward": 0.9813936054706573,
      "rewards-max/forensics_strict_edge_reward": 0.6945812702178955,
      "rewards-max/forensics_tiou_reward": 0.9806720614433289,
      "rewards/forensics_sinkhorn_reward": 0.8045300841331482,
      "rewards/forensics_strict_edge_reward": 0.13237232714891434,
      "rewards/forensics_tiou_reward": 0.6750551164150238,
      "step": 1029
    },
    {
      "completion_length": 25.171875,
      "epoch": 3.1597542242703534,
      "grad_norm": 10.138498306274414,
      "kl": 0.4892578125,
      "learning_rate": 2.0846153846153843e-07,
      "loss": -0.0081,
      "reward": 0.751005232334137,
      "reward_std": 0.019630394876003265,
      "reward_std_all": 0.0375881427899003,
      "rewards-max/forensics_sinkhorn_reward": 0.9857592582702637,
      "rewards-max/forensics_strict_edge_reward": 0.5322106778621674,
      "rewards-max/forensics_tiou_reward": 0.9908415377140045,
      "rewards/forensics_sinkhorn_reward": 0.6952358782291412,
      "rewards/forensics_strict_edge_reward": 0.028010036796331406,
      "rewards/forensics_tiou_reward": 0.5010380148887634,
      "step": 1030
    },
    {
      "completion_length": 27.484375,
      "epoch": 3.162826420890937,
      "grad_norm": 11.897934913635254,
      "kl": 1.041015625,
      "learning_rate": 2.076923076923077e-07,
      "loss": 0.0367,
      "reward": 0.7696610689163208,
      "reward_std": 0.0375856701284647,
      "reward_std_all": 0.037837808951735497,
      "rewards-max/forensics_sinkhorn_reward": 0.9832844138145447,
      "rewards-max/forensics_strict_edge_reward": 0.7209237813949585,
      "rewards-max/forensics_tiou_reward": 0.9860803186893463,
      "rewards/forensics_sinkhorn_reward": 0.7662195861339569,
      "rewards/forensics_strict_edge_reward": 0.1072038784623146,
      "rewards/forensics_tiou_reward": 0.6327428221702576,
      "step": 1031
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.165898617511521,
      "grad_norm": 18.668542861938477,
      "kl": 0.9775390625,
      "learning_rate": 2.0692307692307694e-07,
      "loss": 0.0717,
      "reward": 0.7854989767074585,
      "reward_std": 0.03623528778553009,
      "reward_std_all": 0.03308426961302757,
      "rewards-max/forensics_sinkhorn_reward": 0.9791729152202606,
      "rewards-max/forensics_strict_edge_reward": 0.5840539336204529,
      "rewards-max/forensics_tiou_reward": 0.9737042784690857,
      "rewards/forensics_sinkhorn_reward": 0.8180105686187744,
      "rewards/forensics_strict_edge_reward": 0.09525402542203665,
      "rewards/forensics_tiou_reward": 0.6780916452407837,
      "step": 1032
    },
    {
      "completion_length": 25.359375,
      "epoch": 3.1689708141321047,
      "grad_norm": 9.217358589172363,
      "kl": 0.5048828125,
      "learning_rate": 2.0615384615384615e-07,
      "loss": 0.0217,
      "reward": 0.777643233537674,
      "reward_std": 0.027584376744925976,
      "reward_std_all": 0.029609959572553635,
      "rewards-max/forensics_sinkhorn_reward": 0.9656040966510773,
      "rewards-max/forensics_strict_edge_reward": 0.23852341622114182,
      "rewards-max/forensics_tiou_reward": 0.9386323392391205,
      "rewards/forensics_sinkhorn_reward": 0.7966166436672211,
      "rewards/forensics_strict_edge_reward": 0.034980098716914654,
      "rewards/forensics_tiou_reward": 0.6222644150257111,
      "step": 1033
    },
    {
      "completion_length": 28.5,
      "epoch": 3.172043010752688,
      "grad_norm": 7.898115634918213,
      "kl": 0.4296875,
      "learning_rate": 2.053846153846154e-07,
      "loss": 0.0605,
      "reward": 0.7667216658592224,
      "reward_std": 0.036932604387402534,
      "reward_std_all": 0.03790625371038914,
      "rewards-max/forensics_sinkhorn_reward": 0.9806700944900513,
      "rewards-max/forensics_strict_edge_reward": 0.6348889172077179,
      "rewards-max/forensics_tiou_reward": 0.9799172878265381,
      "rewards/forensics_sinkhorn_reward": 0.7528555393218994,
      "rewards/forensics_strict_edge_reward": 0.0969984419643879,
      "rewards/forensics_tiou_reward": 0.6269872784614563,
      "step": 1034
    },
    {
      "completion_length": 23.03125,
      "epoch": 3.175115207373272,
      "grad_norm": 7.883319854736328,
      "kl": 0.3857421875,
      "learning_rate": 2.046153846153846e-07,
      "loss": -0.0642,
      "reward": 0.7395506203174591,
      "reward_std": 0.04578028991818428,
      "reward_std_all": 0.058663574047386646,
      "rewards-max/forensics_sinkhorn_reward": 0.9781678020954132,
      "rewards-max/forensics_strict_edge_reward": 0.4581223875284195,
      "rewards-max/forensics_tiou_reward": 0.9693226516246796,
      "rewards/forensics_sinkhorn_reward": 0.7652004957199097,
      "rewards/forensics_strict_edge_reward": 0.07672977820038795,
      "rewards/forensics_tiou_reward": 0.6231759786605835,
      "step": 1035
    },
    {
      "completion_length": 24.328125,
      "epoch": 3.1781874039938556,
      "grad_norm": 8.692123413085938,
      "kl": 0.439453125,
      "learning_rate": 2.0384615384615384e-07,
      "loss": 0.0263,
      "reward": 0.8002460598945618,
      "reward_std": 0.006224357930477709,
      "reward_std_all": 0.00857937685213983,
      "rewards-max/forensics_sinkhorn_reward": 0.973048985004425,
      "rewards-max/forensics_strict_edge_reward": 0.6088263988494873,
      "rewards-max/forensics_tiou_reward": 0.959781140089035,
      "rewards/forensics_sinkhorn_reward": 0.8123137056827545,
      "rewards/forensics_strict_edge_reward": 0.06571842543780804,
      "rewards/forensics_tiou_reward": 0.6639096736907959,
      "step": 1036
    },
    {
      "completion_length": 28.109375,
      "epoch": 3.1812596006144394,
      "grad_norm": 7.234295845031738,
      "kl": 0.453125,
      "learning_rate": 2.0307692307692306e-07,
      "loss": -0.026,
      "reward": 0.7769410610198975,
      "reward_std": 0.04121206887066364,
      "reward_std_all": 0.03922376222908497,
      "rewards-max/forensics_sinkhorn_reward": 0.9816180765628815,
      "rewards-max/forensics_strict_edge_reward": 0.6005697548389435,
      "rewards-max/forensics_tiou_reward": 0.9796197712421417,
      "rewards/forensics_sinkhorn_reward": 0.778408020734787,
      "rewards/forensics_strict_edge_reward": 0.0937177762389183,
      "rewards/forensics_tiou_reward": 0.6235099136829376,
      "step": 1037
    },
    {
      "completion_length": 25.4375,
      "epoch": 3.184331797235023,
      "grad_norm": 8.901922225952148,
      "kl": 0.404296875,
      "learning_rate": 2.023076923076923e-07,
      "loss": -0.0239,
      "reward": 0.7618505656719208,
      "reward_std": 0.02488362859003246,
      "reward_std_all": 0.03632839396595955,
      "rewards-max/forensics_sinkhorn_reward": 0.9819457530975342,
      "rewards-max/forensics_strict_edge_reward": 0.6065306663513184,
      "rewards-max/forensics_tiou_reward": 0.9811548888683319,
      "rewards/forensics_sinkhorn_reward": 0.8091085851192474,
      "rewards/forensics_strict_edge_reward": 0.07288588769733906,
      "rewards/forensics_tiou_reward": 0.702595442533493,
      "step": 1038
    },
    {
      "completion_length": 26.375,
      "epoch": 3.187403993855607,
      "grad_norm": 6.41089391708374,
      "kl": 0.4345703125,
      "learning_rate": 2.015384615384615e-07,
      "loss": -0.0607,
      "reward": 0.8048273026943207,
      "reward_std": 0.003109655110165477,
      "reward_std_all": 0.0046872280072420835,
      "rewards-max/forensics_sinkhorn_reward": 0.9800582528114319,
      "rewards-max/forensics_strict_edge_reward": 0.4049251973628998,
      "rewards-max/forensics_tiou_reward": 0.9802796244621277,
      "rewards/forensics_sinkhorn_reward": 0.8518431186676025,
      "rewards/forensics_strict_edge_reward": 0.056891053915023804,
      "rewards/forensics_tiou_reward": 0.7338996529579163,
      "step": 1039
    },
    {
      "completion_length": 26.140625,
      "epoch": 3.1904761904761907,
      "grad_norm": 8.381470680236816,
      "kl": 0.34765625,
      "learning_rate": 2.0076923076923075e-07,
      "loss": -0.0497,
      "reward": 0.7908422946929932,
      "reward_std": 0.023634285666048527,
      "reward_std_all": 0.0227906396612525,
      "rewards-max/forensics_sinkhorn_reward": 0.9830695986747742,
      "rewards-max/forensics_strict_edge_reward": 0.7472657710313797,
      "rewards-max/forensics_tiou_reward": 0.9841189086437225,
      "rewards/forensics_sinkhorn_reward": 0.7904132604598999,
      "rewards/forensics_strict_edge_reward": 0.06884049437940121,
      "rewards/forensics_tiou_reward": 0.7026396989822388,
      "step": 1040
    },
    {
      "completion_length": 26.09375,
      "epoch": 3.193548387096774,
      "grad_norm": 8.910198211669922,
      "kl": 0.5966796875,
      "learning_rate": 2e-07,
      "loss": -0.0468,
      "reward": 0.7714217603206635,
      "reward_std": 0.05351885408163071,
      "reward_std_all": 0.049244124442338943,
      "rewards-max/forensics_sinkhorn_reward": 0.9781143367290497,
      "rewards-max/forensics_strict_edge_reward": 0.25777730345726013,
      "rewards-max/forensics_tiou_reward": 0.9635162353515625,
      "rewards/forensics_sinkhorn_reward": 0.7266733348369598,
      "rewards/forensics_strict_edge_reward": 0.028020435012876987,
      "rewards/forensics_tiou_reward": 0.5527710169553757,
      "step": 1041
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.196620583717358,
      "grad_norm": 7.554893493652344,
      "kl": 0.4970703125,
      "learning_rate": 1.9923076923076923e-07,
      "loss": 0.0574,
      "reward": 0.7803285121917725,
      "reward_std": 0.03226837981492281,
      "reward_std_all": 0.03154323436319828,
      "rewards-max/forensics_sinkhorn_reward": 0.9757994711399078,
      "rewards-max/forensics_strict_edge_reward": 0.3396413326263428,
      "rewards-max/forensics_tiou_reward": 0.9653138518333435,
      "rewards/forensics_sinkhorn_reward": 0.7910810112953186,
      "rewards/forensics_strict_edge_reward": 0.044378191232681274,
      "rewards/forensics_tiou_reward": 0.6278696060180664,
      "step": 1042
    },
    {
      "completion_length": 27.25,
      "epoch": 3.1996927803379416,
      "grad_norm": 8.368423461914062,
      "kl": 0.498046875,
      "learning_rate": 1.9846153846153844e-07,
      "loss": 0.1241,
      "reward": 0.7660074234008789,
      "reward_std": 0.046277355402708054,
      "reward_std_all": 0.04558354243636131,
      "rewards-max/forensics_sinkhorn_reward": 0.9633051455020905,
      "rewards-max/forensics_strict_edge_reward": 0.2042548879981041,
      "rewards-max/forensics_tiou_reward": 0.9416325688362122,
      "rewards/forensics_sinkhorn_reward": 0.6909301280975342,
      "rewards/forensics_strict_edge_reward": 0.027207113103941083,
      "rewards/forensics_tiou_reward": 0.5703553557395935,
      "step": 1043
    },
    {
      "completion_length": 25.671875,
      "epoch": 3.2027649769585254,
      "grad_norm": 7.9220170974731445,
      "kl": 0.455078125,
      "learning_rate": 1.9769230769230768e-07,
      "loss": 0.0256,
      "reward": 0.7903302013874054,
      "reward_std": 0.0177794280461967,
      "reward_std_all": 0.01860456052236259,
      "rewards-max/forensics_sinkhorn_reward": 0.9831064343452454,
      "rewards-max/forensics_strict_edge_reward": 0.500194251537323,
      "rewards-max/forensics_tiou_reward": 0.9850633144378662,
      "rewards/forensics_sinkhorn_reward": 0.7851018905639648,
      "rewards/forensics_strict_edge_reward": 0.0752674788236618,
      "rewards/forensics_tiou_reward": 0.6592302620410919,
      "step": 1044
    },
    {
      "completion_length": 24.203125,
      "epoch": 3.205837173579109,
      "grad_norm": 8.664068222045898,
      "kl": 0.5625,
      "learning_rate": 1.9692307692307693e-07,
      "loss": 0.0063,
      "reward": 0.8007498979568481,
      "reward_std": 0.010215724119916558,
      "reward_std_all": 0.010749673820100725,
      "rewards-max/forensics_sinkhorn_reward": 0.9816860854625702,
      "rewards-max/forensics_strict_edge_reward": 0.6271176636219025,
      "rewards-max/forensics_tiou_reward": 0.9806672632694244,
      "rewards/forensics_sinkhorn_reward": 0.8147832751274109,
      "rewards/forensics_strict_edge_reward": 0.06155494600534439,
      "rewards/forensics_tiou_reward": 0.7186314463615417,
      "step": 1045
    },
    {
      "completion_length": 26.25,
      "epoch": 3.208909370199693,
      "grad_norm": 8.001020431518555,
      "kl": 0.541015625,
      "learning_rate": 1.9615384615384614e-07,
      "loss": 0.0672,
      "reward": 0.7767317295074463,
      "reward_std": 0.04107978055253625,
      "reward_std_all": 0.03842673962935805,
      "rewards-max/forensics_sinkhorn_reward": 0.9842950403690338,
      "rewards-max/forensics_strict_edge_reward": 0.6085590124130249,
      "rewards-max/forensics_tiou_reward": 0.9866904020309448,
      "rewards/forensics_sinkhorn_reward": 0.7982807755470276,
      "rewards/forensics_strict_edge_reward": 0.09683714620769024,
      "rewards/forensics_tiou_reward": 0.6294112801551819,
      "step": 1046
    },
    {
      "completion_length": 27.6875,
      "epoch": 3.2119815668202767,
      "grad_norm": 7.539716720581055,
      "kl": 0.44140625,
      "learning_rate": 1.9538461538461538e-07,
      "loss": 0.007,
      "reward": 0.7891863882541656,
      "reward_std": 0.02654001209884882,
      "reward_std_all": 0.0252344673499465,
      "rewards-max/forensics_sinkhorn_reward": 0.9785955846309662,
      "rewards-max/forensics_strict_edge_reward": 0.6082877814769745,
      "rewards-max/forensics_tiou_reward": 0.9753022789955139,
      "rewards/forensics_sinkhorn_reward": 0.769304484128952,
      "rewards/forensics_strict_edge_reward": 0.04327262379229069,
      "rewards/forensics_tiou_reward": 0.6000668108463287,
      "step": 1047
    },
    {
      "completion_length": 26.1875,
      "epoch": 3.21505376344086,
      "grad_norm": 6.9496541023254395,
      "kl": 0.4013671875,
      "learning_rate": 1.946153846153846e-07,
      "loss": 0.1363,
      "reward": 0.7286942899227142,
      "reward_std": 0.05995565466582775,
      "reward_std_all": 0.07300426438450813,
      "rewards-max/forensics_sinkhorn_reward": 0.9688861072063446,
      "rewards-max/forensics_strict_edge_reward": 0.5987430810928345,
      "rewards-max/forensics_tiou_reward": 0.9524845480918884,
      "rewards/forensics_sinkhorn_reward": 0.6971008777618408,
      "rewards/forensics_strict_edge_reward": 0.060860276222229004,
      "rewards/forensics_tiou_reward": 0.6042238771915436,
      "step": 1048
    },
    {
      "completion_length": 25.359375,
      "epoch": 3.218125960061444,
      "grad_norm": 7.465199947357178,
      "kl": 0.4013671875,
      "learning_rate": 1.9384615384615383e-07,
      "loss": -0.151,
      "reward": 0.8046488165855408,
      "reward_std": 0.002203117939643562,
      "reward_std_all": 0.004360046936199069,
      "rewards-max/forensics_sinkhorn_reward": 0.9655947685241699,
      "rewards-max/forensics_strict_edge_reward": 0.6083009243011475,
      "rewards-max/forensics_tiou_reward": 0.9392853379249573,
      "rewards/forensics_sinkhorn_reward": 0.8306831121444702,
      "rewards/forensics_strict_edge_reward": 0.0653008408844471,
      "rewards/forensics_tiou_reward": 0.6428951919078827,
      "step": 1049
    },
    {
      "completion_length": 22.375,
      "epoch": 3.2211981566820276,
      "grad_norm": 12.72594928741455,
      "kl": 0.771484375,
      "learning_rate": 1.9307692307692307e-07,
      "loss": -0.0384,
      "reward": 0.7935182750225067,
      "reward_std": 0.0150132542476058,
      "reward_std_all": 0.01570306532084942,
      "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": 0.8168069124221802,
      "rewards/forensics_strict_edge_reward": 0.13976746052503586,
      "rewards/forensics_tiou_reward": 0.6994061470031738,
      "step": 1050
    },
    {
      "completion_length": 23.109375,
      "epoch": 3.2242703533026114,
      "grad_norm": 7.905198097229004,
      "kl": 0.5810546875,
      "learning_rate": 1.9230769230769231e-07,
      "loss": -0.0525,
      "reward": 0.8012998402118683,
      "reward_std": 0.00797388143837452,
      "reward_std_all": 0.008951770141720772,
      "rewards-max/forensics_sinkhorn_reward": 0.9888491928577423,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.9995000660419464,
      "rewards/forensics_sinkhorn_reward": 0.8290512263774872,
      "rewards/forensics_strict_edge_reward": 0.10657227411866188,
      "rewards/forensics_tiou_reward": 0.6887220442295074,
      "step": 1051
    },
    {
      "completion_length": 25.0625,
      "epoch": 3.227342549923195,
      "grad_norm": 7.656179428100586,
      "kl": 0.380859375,
      "learning_rate": 1.9153846153846153e-07,
      "loss": -0.004,
      "reward": 0.779772937297821,
      "reward_std": 0.036160387098789215,
      "reward_std_all": 0.034503910690546036,
      "rewards-max/forensics_sinkhorn_reward": 0.9856664836406708,
      "rewards-max/forensics_strict_edge_reward": 0.720185175538063,
      "rewards-max/forensics_tiou_reward": 0.9902088046073914,
      "rewards/forensics_sinkhorn_reward": 0.7999691367149353,
      "rewards/forensics_strict_edge_reward": 0.08288989588618279,
      "rewards/forensics_tiou_reward": 0.6283382773399353,
      "step": 1052
    },
    {
      "completion_length": 26.890625,
      "epoch": 3.230414746543779,
      "grad_norm": 7.333787441253662,
      "kl": 0.4306640625,
      "learning_rate": 1.9076923076923077e-07,
      "loss": -0.0361,
      "reward": 0.7998803555965424,
      "reward_std": 0.01174411189276725,
      "reward_std_all": 0.012069475371390581,
      "rewards-max/forensics_sinkhorn_reward": 0.9849825501441956,
      "rewards-max/forensics_strict_edge_reward": 0.44934946298599243,
      "rewards-max/forensics_tiou_reward": 0.9874527752399445,
      "rewards/forensics_sinkhorn_reward": 0.8301615417003632,
      "rewards/forensics_strict_edge_reward": 0.06840280257165432,
      "rewards/forensics_tiou_reward": 0.7122169733047485,
      "step": 1053
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.2334869431643627,
      "grad_norm": 6.4886932373046875,
      "kl": 0.5126953125,
      "learning_rate": 1.8999999999999998e-07,
      "loss": -0.0824,
      "reward": 0.800327330827713,
      "reward_std": 0.006618208950385451,
      "reward_std_all": 0.008163082879036665,
      "rewards-max/forensics_sinkhorn_reward": 0.9740684032440186,
      "rewards-max/forensics_strict_edge_reward": 0.41346001625061035,
      "rewards-max/forensics_tiou_reward": 0.9623647630214691,
      "rewards/forensics_sinkhorn_reward": 0.7996949553489685,
      "rewards/forensics_strict_edge_reward": 0.0676233060657978,
      "rewards/forensics_tiou_reward": 0.6370242238044739,
      "step": 1054
    },
    {
      "completion_length": 26.9375,
      "epoch": 3.236559139784946,
      "grad_norm": 7.368425369262695,
      "kl": 0.3779296875,
      "learning_rate": 1.8923076923076922e-07,
      "loss": -0.0946,
      "reward": 0.8045834302902222,
      "reward_std": 0.0019260564586147666,
      "reward_std_all": 0.004061357583850622,
      "rewards-max/forensics_sinkhorn_reward": 0.9773734211921692,
      "rewards-max/forensics_strict_edge_reward": 0.6378761231899261,
      "rewards-max/forensics_tiou_reward": 0.9745742678642273,
      "rewards/forensics_sinkhorn_reward": 0.8053101897239685,
      "rewards/forensics_strict_edge_reward": 0.06412661261856556,
      "rewards/forensics_tiou_reward": 0.6279443800449371,
      "step": 1055
    },
    {
      "completion_length": 24.9375,
      "epoch": 3.23963133640553,
      "grad_norm": 7.738780975341797,
      "kl": 0.5625,
      "learning_rate": 1.8846153846153846e-07,
      "loss": -0.1104,
      "reward": 0.8049427568912506,
      "reward_std": 0.0027586909127421677,
      "reward_std_all": 0.004331590607762337,
      "rewards-max/forensics_sinkhorn_reward": 0.979439377784729,
      "rewards-max/forensics_strict_edge_reward": 0.5135394930839539,
      "rewards-max/forensics_tiou_reward": 0.9772160053253174,
      "rewards/forensics_sinkhorn_reward": 0.8245347142219543,
      "rewards/forensics_strict_edge_reward": 0.08038248866796494,
      "rewards/forensics_tiou_reward": 0.6762522459030151,
      "step": 1056
    },
    {
      "completion_length": 23.375,
      "epoch": 3.2427035330261136,
      "grad_norm": 8.44286060333252,
      "kl": 0.4794921875,
      "learning_rate": 1.8769230769230767e-07,
      "loss": 0.0101,
      "reward": 0.7791110277175903,
      "reward_std": 0.038123817183077335,
      "reward_std_all": 0.036193653009831905,
      "rewards-max/forensics_sinkhorn_reward": 0.9843331873416901,
      "rewards-max/forensics_strict_edge_reward": 0.8142738044261932,
      "rewards-max/forensics_tiou_reward": 0.9888582825660706,
      "rewards/forensics_sinkhorn_reward": 0.8057995140552521,
      "rewards/forensics_strict_edge_reward": 0.0985436923801899,
      "rewards/forensics_tiou_reward": 0.6445631086826324,
      "step": 1057
    },
    {
      "completion_length": 24.421875,
      "epoch": 3.2457757296466974,
      "grad_norm": 8.274039268493652,
      "kl": 0.4755859375,
      "learning_rate": 1.8692307692307692e-07,
      "loss": -0.0478,
      "reward": 0.7616168856620789,
      "reward_std": 0.039913782849907875,
      "reward_std_all": 0.042671618051826954,
      "rewards-max/forensics_sinkhorn_reward": 0.9655316174030304,
      "rewards-max/forensics_strict_edge_reward": 0.6733579039573669,
      "rewards-max/forensics_tiou_reward": 0.9444484114646912,
      "rewards/forensics_sinkhorn_reward": 0.7413530349731445,
      "rewards/forensics_strict_edge_reward": 0.06535854935646057,
      "rewards/forensics_tiou_reward": 0.5507806837558746,
      "step": 1058
    },
    {
      "completion_length": 22.0,
      "epoch": 3.248847926267281,
      "grad_norm": 7.9638142585754395,
      "kl": 0.4765625,
      "learning_rate": 1.8615384615384613e-07,
      "loss": -0.0457,
      "reward": 0.8019372224807739,
      "reward_std": 0.00832244916819036,
      "reward_std_all": 0.00906249019317329,
      "rewards-max/forensics_sinkhorn_reward": 0.9871109426021576,
      "rewards-max/forensics_strict_edge_reward": 0.952418714761734,
      "rewards-max/forensics_tiou_reward": 0.9955528378486633,
      "rewards/forensics_sinkhorn_reward": 0.8153541386127472,
      "rewards/forensics_strict_edge_reward": 0.10939386487007141,
      "rewards/forensics_tiou_reward": 0.6885305047035217,
      "step": 1059
    },
    {
      "completion_length": 24.859375,
      "epoch": 3.251920122887865,
      "grad_norm": 6.823853492736816,
      "kl": 0.35546875,
      "learning_rate": 1.853846153846154e-07,
      "loss": -0.0201,
      "reward": 0.7953015565872192,
      "reward_std": 0.01822881307452917,
      "reward_std_all": 0.01776297716423869,
      "rewards-max/forensics_sinkhorn_reward": 0.9792607724666595,
      "rewards-max/forensics_strict_edge_reward": 0.6274375021457672,
      "rewards-max/forensics_tiou_reward": 0.9736345410346985,
      "rewards/forensics_sinkhorn_reward": 0.8227885067462921,
      "rewards/forensics_strict_edge_reward": 0.07500378601253033,
      "rewards/forensics_tiou_reward": 0.7139990925788879,
      "step": 1060
    },
    {
      "completion_length": 29.71875,
      "epoch": 3.2549923195084487,
      "grad_norm": 6.693057537078857,
      "kl": 0.4306640625,
      "learning_rate": 1.846153846153846e-07,
      "loss": -0.0132,
      "reward": 0.7746013700962067,
      "reward_std": 0.035650126170367,
      "reward_std_all": 0.03518363926559687,
      "rewards-max/forensics_sinkhorn_reward": 0.9840584695339203,
      "rewards-max/forensics_strict_edge_reward": 0.5308436453342438,
      "rewards-max/forensics_tiou_reward": 0.9863865673542023,
      "rewards/forensics_sinkhorn_reward": 0.7839483618736267,
      "rewards/forensics_strict_edge_reward": 0.09246553853154182,
      "rewards/forensics_tiou_reward": 0.6488244831562042,
      "step": 1061
    },
    {
      "completion_length": 23.359375,
      "epoch": 3.258064516129032,
      "grad_norm": 7.521206855773926,
      "kl": 0.4677734375,
      "learning_rate": 1.8384615384615385e-07,
      "loss": -0.0884,
      "reward": 0.8055688738822937,
      "reward_std": 0.0015975077403709292,
      "reward_std_all": 0.0034437752328813076,
      "rewards-max/forensics_sinkhorn_reward": 0.9870436191558838,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9947166740894318,
      "rewards/forensics_sinkhorn_reward": 0.8402898907661438,
      "rewards/forensics_strict_edge_reward": 0.08723952621221542,
      "rewards/forensics_tiou_reward": 0.6839397549629211,
      "step": 1062
    },
    {
      "completion_length": 24.3125,
      "epoch": 3.261136712749616,
      "grad_norm": 6.906801223754883,
      "kl": 0.421875,
      "learning_rate": 1.8307692307692306e-07,
      "loss": -0.0637,
      "reward": 0.8017525970935822,
      "reward_std": 0.009113782842177898,
      "reward_std_all": 0.009736804640851915,
      "rewards-max/forensics_sinkhorn_reward": 0.9873649477958679,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9953446388244629,
      "rewards/forensics_sinkhorn_reward": 0.8639800548553467,
      "rewards/forensics_strict_edge_reward": 0.10070755332708359,
      "rewards/forensics_tiou_reward": 0.7357819974422455,
      "step": 1063
    },
    {
      "completion_length": 25.078125,
      "epoch": 3.2642089093701996,
      "grad_norm": 5.943799018859863,
      "kl": 0.4541015625,
      "learning_rate": 1.823076923076923e-07,
      "loss": -0.0211,
      "reward": 0.7669785916805267,
      "reward_std": 0.02577682351693511,
      "reward_std_all": 0.036002242006361485,
      "rewards-max/forensics_sinkhorn_reward": 0.9692132771015167,
      "rewards-max/forensics_strict_edge_reward": 0.5724950432777405,
      "rewards-max/forensics_tiou_reward": 0.9531119167804718,
      "rewards/forensics_sinkhorn_reward": 0.7908128201961517,
      "rewards/forensics_strict_edge_reward": 0.09047254361212254,
      "rewards/forensics_tiou_reward": 0.6282959878444672,
      "step": 1064
    },
    {
      "completion_length": 26.484375,
      "epoch": 3.2672811059907834,
      "grad_norm": 7.302743434906006,
      "kl": 0.431640625,
      "learning_rate": 1.8153846153846152e-07,
      "loss": 0.0108,
      "reward": 0.7732787132263184,
      "reward_std": 0.032579658553004265,
      "reward_std_all": 0.03424456529319286,
      "rewards-max/forensics_sinkhorn_reward": 0.9802355468273163,
      "rewards-max/forensics_strict_edge_reward": 0.5743183493614197,
      "rewards-max/forensics_tiou_reward": 0.9784662127494812,
      "rewards/forensics_sinkhorn_reward": 0.7787342071533203,
      "rewards/forensics_strict_edge_reward": 0.07623298466205597,
      "rewards/forensics_tiou_reward": 0.6139993965625763,
      "step": 1065
    },
    {
      "completion_length": 22.828125,
      "epoch": 3.270353302611367,
      "grad_norm": 7.219225883483887,
      "kl": 0.455078125,
      "learning_rate": 1.8076923076923076e-07,
      "loss": -0.0204,
      "reward": 0.7867467701435089,
      "reward_std": 0.033127584494650364,
      "reward_std_all": 0.030701786279678345,
      "rewards-max/forensics_sinkhorn_reward": 0.9750640988349915,
      "rewards-max/forensics_strict_edge_reward": 0.36763034760951996,
      "rewards-max/forensics_tiou_reward": 0.9598184823989868,
      "rewards/forensics_sinkhorn_reward": 0.7780013978481293,
      "rewards/forensics_strict_edge_reward": 0.04174153506755829,
      "rewards/forensics_tiou_reward": 0.5962931215763092,
      "step": 1066
    },
    {
      "completion_length": 26.296875,
      "epoch": 3.273425499231951,
      "grad_norm": 8.50984001159668,
      "kl": 0.5107421875,
      "learning_rate": 1.8e-07,
      "loss": 0.0937,
      "reward": 0.7812144756317139,
      "reward_std": 0.03584587574005127,
      "reward_std_all": 0.03388496953994036,
      "rewards-max/forensics_sinkhorn_reward": 0.9806294441223145,
      "rewards-max/forensics_strict_edge_reward": 0.39087022840976715,
      "rewards-max/forensics_tiou_reward": 0.9780177474021912,
      "rewards/forensics_sinkhorn_reward": 0.7908696532249451,
      "rewards/forensics_strict_edge_reward": 0.051801095716655254,
      "rewards/forensics_tiou_reward": 0.6738743484020233,
      "step": 1067
    },
    {
      "completion_length": 25.421875,
      "epoch": 3.2764976958525347,
      "grad_norm": 7.714542865753174,
      "kl": 0.48828125,
      "learning_rate": 1.792307692307692e-07,
      "loss": -0.0317,
      "reward": 0.7759258449077606,
      "reward_std": 0.03235024120658636,
      "reward_std_all": 0.032687632367014885,
      "rewards-max/forensics_sinkhorn_reward": 0.9734580814838409,
      "rewards-max/forensics_strict_edge_reward": 0.48898807168006897,
      "rewards-max/forensics_tiou_reward": 0.9600272178649902,
      "rewards/forensics_sinkhorn_reward": 0.7319741249084473,
      "rewards/forensics_strict_edge_reward": 0.03949396591633558,
      "rewards/forensics_tiou_reward": 0.5425710082054138,
      "step": 1068
    },
    {
      "completion_length": 27.28125,
      "epoch": 3.279569892473118,
      "grad_norm": 10.694385528564453,
      "kl": 0.4072265625,
      "learning_rate": 1.7846153846153845e-07,
      "loss": -0.0529,
      "reward": 0.7997534871101379,
      "reward_std": 0.012488558189943433,
      "reward_std_all": 0.012483401224017143,
      "rewards-max/forensics_sinkhorn_reward": 0.984828382730484,
      "rewards-max/forensics_strict_edge_reward": 0.6792138814926147,
      "rewards-max/forensics_tiou_reward": 0.991394966840744,
      "rewards/forensics_sinkhorn_reward": 0.7881543636322021,
      "rewards/forensics_strict_edge_reward": 0.14426986873149872,
      "rewards/forensics_tiou_reward": 0.6887466609477997,
      "step": 1069
    },
    {
      "completion_length": 27.0,
      "epoch": 3.282642089093702,
      "grad_norm": 5.952958583831787,
      "kl": 0.615234375,
      "learning_rate": 1.776923076923077e-07,
      "loss": 0.0238,
      "reward": 0.7855882346630096,
      "reward_std": 0.03472785511985421,
      "reward_std_all": 0.03188085649162531,
      "rewards-max/forensics_sinkhorn_reward": 0.9652957022190094,
      "rewards-max/forensics_strict_edge_reward": 0.4274432845413685,
      "rewards-max/forensics_tiou_reward": 0.9450221359729767,
      "rewards/forensics_sinkhorn_reward": 0.7775404453277588,
      "rewards/forensics_strict_edge_reward": 0.040578297805041075,
      "rewards/forensics_tiou_reward": 0.6029751747846603,
      "step": 1070
    },
    {
      "completion_length": 25.078125,
      "epoch": 3.2857142857142856,
      "grad_norm": 7.680172920227051,
      "kl": 0.408203125,
      "learning_rate": 1.7692307692307693e-07,
      "loss": -0.0385,
      "reward": 0.7532984614372253,
      "reward_std": 0.029451316688209772,
      "reward_std_all": 0.0409182570874691,
      "rewards-max/forensics_sinkhorn_reward": 0.9744047522544861,
      "rewards-max/forensics_strict_edge_reward": 0.6408714950084686,
      "rewards-max/forensics_tiou_reward": 0.963689386844635,
      "rewards/forensics_sinkhorn_reward": 0.7298518717288971,
      "rewards/forensics_strict_edge_reward": 0.07011241745203733,
      "rewards/forensics_tiou_reward": 0.573671966791153,
      "step": 1071
    },
    {
      "completion_length": 24.65625,
      "epoch": 3.2887864823348694,
      "grad_norm": 8.641556739807129,
      "kl": 0.4794921875,
      "learning_rate": 1.7615384615384615e-07,
      "loss": 0.0438,
      "reward": 0.7719654440879822,
      "reward_std": 0.04529014974832535,
      "reward_std_all": 0.04275842569768429,
      "rewards-max/forensics_sinkhorn_reward": 0.9775421917438507,
      "rewards-max/forensics_strict_edge_reward": 0.5190997421741486,
      "rewards-max/forensics_tiou_reward": 0.9708452522754669,
      "rewards/forensics_sinkhorn_reward": 0.7407234013080597,
      "rewards/forensics_strict_edge_reward": 0.07719800621271133,
      "rewards/forensics_tiou_reward": 0.5933603048324585,
      "step": 1072
    },
    {
      "completion_length": 27.40625,
      "epoch": 3.291858678955453,
      "grad_norm": 8.496874809265137,
      "kl": 0.4541015625,
      "learning_rate": 1.7538461538461539e-07,
      "loss": -0.0242,
      "reward": 0.7899452149868011,
      "reward_std": 0.016867245081812143,
      "reward_std_all": 0.01826802222058177,
      "rewards-max/forensics_sinkhorn_reward": 0.9856932759284973,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.990477591753006,
      "rewards/forensics_sinkhorn_reward": 0.7451992928981781,
      "rewards/forensics_strict_edge_reward": 0.08319756388664246,
      "rewards/forensics_tiou_reward": 0.6152420043945312,
      "step": 1073
    },
    {
      "completion_length": 26.46875,
      "epoch": 3.294930875576037,
      "grad_norm": 7.914041996002197,
      "kl": 0.458984375,
      "learning_rate": 1.746153846153846e-07,
      "loss": 0.0737,
      "reward": 0.7830998599529266,
      "reward_std": 0.03133539669215679,
      "reward_std_all": 0.030098802410066128,
      "rewards-max/forensics_sinkhorn_reward": 0.9700550734996796,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.9536672234535217,
      "rewards/forensics_sinkhorn_reward": 0.7820132672786713,
      "rewards/forensics_strict_edge_reward": 0.09091359935700893,
      "rewards/forensics_tiou_reward": 0.6413022577762604,
      "step": 1074
    },
    {
      "completion_length": 24.9375,
      "epoch": 3.2980030721966207,
      "grad_norm": 7.868408203125,
      "kl": 0.4443359375,
      "learning_rate": 1.7384615384615384e-07,
      "loss": -0.0687,
      "reward": 0.7932799160480499,
      "reward_std": 0.02359762362902984,
      "reward_std_all": 0.022111095255240798,
      "rewards-max/forensics_sinkhorn_reward": 0.977594643831253,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9694579243659973,
      "rewards/forensics_sinkhorn_reward": 0.8494445383548737,
      "rewards/forensics_strict_edge_reward": 0.07686403393745422,
      "rewards/forensics_tiou_reward": 0.7191050350666046,
      "step": 1075
    },
    {
      "completion_length": 25.578125,
      "epoch": 3.3010752688172045,
      "grad_norm": 7.130821704864502,
      "kl": 0.4912109375,
      "learning_rate": 1.7307692307692305e-07,
      "loss": 0.014,
      "reward": 0.7728859782218933,
      "reward_std": 0.027458841912448406,
      "reward_std_all": 0.030255117919296026,
      "rewards-max/forensics_sinkhorn_reward": 0.9820515513420105,
      "rewards-max/forensics_strict_edge_reward": 0.5271099358797073,
      "rewards-max/forensics_tiou_reward": 0.9816550314426422,
      "rewards/forensics_sinkhorn_reward": 0.8177810311317444,
      "rewards/forensics_strict_edge_reward": 0.08746889233589172,
      "rewards/forensics_tiou_reward": 0.7016158699989319,
      "step": 1076
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.3041474654377883,
      "grad_norm": 7.578813076019287,
      "kl": 0.53125,
      "learning_rate": 1.723076923076923e-07,
      "loss": 0.0794,
      "reward": 0.7749649286270142,
      "reward_std": 0.04247916862368584,
      "reward_std_all": 0.040294949896633625,
      "rewards-max/forensics_sinkhorn_reward": 0.980889618396759,
      "rewards-max/forensics_strict_edge_reward": 0.6233813166618347,
      "rewards-max/forensics_tiou_reward": 0.9754282534122467,
      "rewards/forensics_sinkhorn_reward": 0.7065174877643585,
      "rewards/forensics_strict_edge_reward": 0.07716313377022743,
      "rewards/forensics_tiou_reward": 0.5417869091033936,
      "step": 1077
    },
    {
      "completion_length": 27.21875,
      "epoch": 3.3072196620583716,
      "grad_norm": 6.946925640106201,
      "kl": 0.388671875,
      "learning_rate": 1.7153846153846153e-07,
      "loss": -0.0106,
      "reward": 0.7426241636276245,
      "reward_std": 0.02424627263098955,
      "reward_std_all": 0.04590040259063244,
      "rewards-max/forensics_sinkhorn_reward": 0.9831541776657104,
      "rewards-max/forensics_strict_edge_reward": 0.6177990138530731,
      "rewards-max/forensics_tiou_reward": 0.9835511445999146,
      "rewards/forensics_sinkhorn_reward": 0.7622734010219574,
      "rewards/forensics_strict_edge_reward": 0.07983534783124924,
      "rewards/forensics_tiou_reward": 0.6387923359870911,
      "step": 1078
    },
    {
      "completion_length": 22.828125,
      "epoch": 3.3102918586789554,
      "grad_norm": 6.918896675109863,
      "kl": 0.32421875,
      "learning_rate": 1.7076923076923075e-07,
      "loss": -0.1187,
      "reward": 0.8013187348842621,
      "reward_std": 0.011305069201625884,
      "reward_std_all": 0.011084213852882385,
      "rewards-max/forensics_sinkhorn_reward": 0.9871570765972137,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9943637847900391,
      "rewards/forensics_sinkhorn_reward": 0.8597375452518463,
      "rewards/forensics_strict_edge_reward": 0.1747671216726303,
      "rewards/forensics_tiou_reward": 0.761967271566391,
      "step": 1079
    },
    {
      "completion_length": 29.921875,
      "epoch": 3.313364055299539,
      "grad_norm": 7.056856155395508,
      "kl": 0.431640625,
      "learning_rate": 1.7000000000000001e-07,
      "loss": 0.031,
      "reward": 0.7837491035461426,
      "reward_std": 0.03685235558077693,
      "reward_std_all": 0.03408739157021046,
      "rewards-max/forensics_sinkhorn_reward": 0.9812276363372803,
      "rewards-max/forensics_strict_edge_reward": 0.5598244965076447,
      "rewards-max/forensics_tiou_reward": 0.9776264727115631,
      "rewards/forensics_sinkhorn_reward": 0.7973012328147888,
      "rewards/forensics_strict_edge_reward": 0.08287299051880836,
      "rewards/forensics_tiou_reward": 0.6898819506168365,
      "step": 1080
    },
    {
      "completion_length": 24.0625,
      "epoch": 3.316436251920123,
      "grad_norm": 7.4096784591674805,
      "kl": 0.451171875,
      "learning_rate": 1.6923076923076923e-07,
      "loss": -0.0526,
      "reward": 0.8001392483711243,
      "reward_std": 0.010973868425935507,
      "reward_std_all": 0.011588169727474451,
      "rewards-max/forensics_sinkhorn_reward": 0.9877491593360901,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9967256188392639,
      "rewards/forensics_sinkhorn_reward": 0.8182654976844788,
      "rewards/forensics_strict_edge_reward": 0.11002916842699051,
      "rewards/forensics_tiou_reward": 0.7053637504577637,
      "step": 1081
    },
    {
      "completion_length": 25.375,
      "epoch": 3.3195084485407067,
      "grad_norm": 9.736969947814941,
      "kl": 0.47265625,
      "learning_rate": 1.6846153846153847e-07,
      "loss": 0.0314,
      "reward": 0.7967572808265686,
      "reward_std": 0.011747275246307254,
      "reward_std_all": 0.01283341646194458,
      "rewards-max/forensics_sinkhorn_reward": 0.9787600040435791,
      "rewards-max/forensics_strict_edge_reward": 0.457786500453949,
      "rewards-max/forensics_tiou_reward": 0.9749756157398224,
      "rewards/forensics_sinkhorn_reward": 0.7748649418354034,
      "rewards/forensics_strict_edge_reward": 0.06295046955347061,
      "rewards/forensics_tiou_reward": 0.6260544657707214,
      "step": 1082
    },
    {
      "completion_length": 22.015625,
      "epoch": 3.3225806451612905,
      "grad_norm": 8.044220924377441,
      "kl": 0.4052734375,
      "learning_rate": 1.6769230769230768e-07,
      "loss": -0.0905,
      "reward": 0.7842252254486084,
      "reward_std": 0.006766341160982847,
      "reward_std_all": 0.01688357349485159,
      "rewards-max/forensics_sinkhorn_reward": 0.9830938279628754,
      "rewards-max/forensics_strict_edge_reward": 0.4419635832309723,
      "rewards-max/forensics_tiou_reward": 0.9847917258739471,
      "rewards/forensics_sinkhorn_reward": 0.8149926066398621,
      "rewards/forensics_strict_edge_reward": 0.05830247513949871,
      "rewards/forensics_tiou_reward": 0.6907175779342651,
      "step": 1083
    },
    {
      "completion_length": 27.4375,
      "epoch": 3.3256528417818743,
      "grad_norm": 7.649552345275879,
      "kl": 0.353515625,
      "learning_rate": 1.6692307692307692e-07,
      "loss": 0.0826,
      "reward": 0.7684619426727295,
      "reward_std": 0.044228872284293175,
      "reward_std_all": 0.042772869579494,
      "rewards-max/forensics_sinkhorn_reward": 0.979235827922821,
      "rewards-max/forensics_strict_edge_reward": 0.6657527685165405,
      "rewards-max/forensics_tiou_reward": 0.9753478467464447,
      "rewards/forensics_sinkhorn_reward": 0.7858270108699799,
      "rewards/forensics_strict_edge_reward": 0.09099799767136574,
      "rewards/forensics_tiou_reward": 0.616504967212677,
      "step": 1084
    },
    {
      "completion_length": 22.546875,
      "epoch": 3.3287250384024576,
      "grad_norm": 7.3897175788879395,
      "kl": 0.3740234375,
      "learning_rate": 1.6615384615384613e-07,
      "loss": 0.0661,
      "reward": 0.7936018407344818,
      "reward_std": 0.01460346207022667,
      "reward_std_all": 0.015593270771205425,
      "rewards-max/forensics_sinkhorn_reward": 0.9806724190711975,
      "rewards-max/forensics_strict_edge_reward": 0.6228483319282532,
      "rewards-max/forensics_tiou_reward": 0.9818851053714752,
      "rewards/forensics_sinkhorn_reward": 0.8436513543128967,
      "rewards/forensics_strict_edge_reward": 0.08162796124815941,
      "rewards/forensics_tiou_reward": 0.7100463807582855,
      "step": 1085
    },
    {
      "completion_length": 23.921875,
      "epoch": 3.3317972350230414,
      "grad_norm": 7.883527755737305,
      "kl": 0.4873046875,
      "learning_rate": 1.6538461538461538e-07,
      "loss": -0.0208,
      "reward": 0.7854991853237152,
      "reward_std": 0.03196247527375817,
      "reward_std_all": 0.03027174749877304,
      "rewards-max/forensics_sinkhorn_reward": 0.9771120846271515,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9704310595989227,
      "rewards/forensics_sinkhorn_reward": 0.7828723192214966,
      "rewards/forensics_strict_edge_reward": 0.0696367397904396,
      "rewards/forensics_tiou_reward": 0.620156466960907,
      "step": 1086
    },
    {
      "completion_length": 24.734375,
      "epoch": 3.334869431643625,
      "grad_norm": 6.718643665313721,
      "kl": 0.63671875,
      "learning_rate": 1.646153846153846e-07,
      "loss": -0.0456,
      "reward": 0.7916137874126434,
      "reward_std": 0.02617912378627807,
      "reward_std_all": 0.024478535633534193,
      "rewards-max/forensics_sinkhorn_reward": 0.9666414558887482,
      "rewards-max/forensics_strict_edge_reward": 0.563819408416748,
      "rewards-max/forensics_tiou_reward": 0.9475301802158356,
      "rewards/forensics_sinkhorn_reward": 0.8318292200565338,
      "rewards/forensics_strict_edge_reward": 0.09222649596631527,
      "rewards/forensics_tiou_reward": 0.6745765507221222,
      "step": 1087
    },
    {
      "completion_length": 27.453125,
      "epoch": 3.337941628264209,
      "grad_norm": 7.304737091064453,
      "kl": 0.54296875,
      "learning_rate": 1.6384615384615383e-07,
      "loss": -0.0342,
      "reward": 0.8034422099590302,
      "reward_std": 0.006066214060410857,
      "reward_std_all": 0.006945237284526229,
      "rewards-max/forensics_sinkhorn_reward": 0.9831068813800812,
      "rewards-max/forensics_strict_edge_reward": 0.6360974311828613,
      "rewards-max/forensics_tiou_reward": 0.9865845143795013,
      "rewards/forensics_sinkhorn_reward": 0.8384964168071747,
      "rewards/forensics_strict_edge_reward": 0.09606154635548592,
      "rewards/forensics_tiou_reward": 0.7194192409515381,
      "step": 1088
    },
    {
      "completion_length": 26.71875,
      "epoch": 3.3410138248847927,
      "grad_norm": 7.260293483734131,
      "kl": 0.396484375,
      "learning_rate": 1.6307692307692307e-07,
      "loss": -0.0154,
      "reward": 0.7648736536502838,
      "reward_std": 0.024900217074900866,
      "reward_std_all": 0.03372016176581383,
      "rewards-max/forensics_sinkhorn_reward": 0.9824301600456238,
      "rewards-max/forensics_strict_edge_reward": 0.8053101599216461,
      "rewards-max/forensics_tiou_reward": 0.9838740229606628,
      "rewards/forensics_sinkhorn_reward": 0.7585499584674835,
      "rewards/forensics_strict_edge_reward": 0.08508823439478874,
      "rewards/forensics_tiou_reward": 0.5817480981349945,
      "step": 1089
    },
    {
      "completion_length": 22.859375,
      "epoch": 3.3440860215053765,
      "grad_norm": 7.047128677368164,
      "kl": 0.521484375,
      "learning_rate": 1.623076923076923e-07,
      "loss": 0.0161,
      "reward": 0.7963310778141022,
      "reward_std": 0.018487609457224607,
      "reward_std_all": 0.017650882713496685,
      "rewards-max/forensics_sinkhorn_reward": 0.9864358305931091,
      "rewards-max/forensics_strict_edge_reward": 0.7361255586147308,
      "rewards-max/forensics_tiou_reward": 0.9939048290252686,
      "rewards/forensics_sinkhorn_reward": 0.8320823907852173,
      "rewards/forensics_strict_edge_reward": 0.11322054266929626,
      "rewards/forensics_tiou_reward": 0.6976612210273743,
      "step": 1090
    },
    {
      "completion_length": 25.203125,
      "epoch": 3.3471582181259603,
      "grad_norm": 9.996383666992188,
      "kl": 0.427734375,
      "learning_rate": 1.6153846153846155e-07,
      "loss": -0.0659,
      "reward": 0.7999738454818726,
      "reward_std": 0.011596154887229204,
      "reward_std_all": 0.01208277978003025,
      "rewards-max/forensics_sinkhorn_reward": 0.973497748374939,
      "rewards-max/forensics_strict_edge_reward": 0.6620611250400543,
      "rewards-max/forensics_tiou_reward": 0.9632436037063599,
      "rewards/forensics_sinkhorn_reward": 0.8548546731472015,
      "rewards/forensics_strict_edge_reward": 0.09219199419021606,
      "rewards/forensics_tiou_reward": 0.7032820880413055,
      "step": 1091
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.3502304147465436,
      "grad_norm": 8.55565357208252,
      "kl": 0.408203125,
      "learning_rate": 1.6076923076923076e-07,
      "loss": -0.0379,
      "reward": 0.8026541173458099,
      "reward_std": 0.005011319415643811,
      "reward_std_all": 0.006430442328564823,
      "rewards-max/forensics_sinkhorn_reward": 0.9851414263248444,
      "rewards-max/forensics_strict_edge_reward": 0.7923295497894287,
      "rewards-max/forensics_tiou_reward": 0.9907248020172119,
      "rewards/forensics_sinkhorn_reward": 0.8229882717132568,
      "rewards/forensics_strict_edge_reward": 0.10524998232722282,
      "rewards/forensics_tiou_reward": 0.6517401337623596,
      "step": 1092
    },
    {
      "completion_length": 28.859375,
      "epoch": 3.3533026113671274,
      "grad_norm": 6.216616153717041,
      "kl": 0.42578125,
      "learning_rate": 1.6e-07,
      "loss": -0.0024,
      "reward": 0.782086193561554,
      "reward_std": 0.02769696363247931,
      "reward_std_all": 0.027771343011409044,
      "rewards-max/forensics_sinkhorn_reward": 0.9795427322387695,
      "rewards-max/forensics_strict_edge_reward": 0.37929777801036835,
      "rewards-max/forensics_tiou_reward": 0.9779480993747711,
      "rewards/forensics_sinkhorn_reward": 0.8079867362976074,
      "rewards/forensics_strict_edge_reward": 0.048877740278840065,
      "rewards/forensics_tiou_reward": 0.6966106295585632,
      "step": 1093
    },
    {
      "completion_length": 24.078125,
      "epoch": 3.356374807987711,
      "grad_norm": 6.300889492034912,
      "kl": 0.4736328125,
      "learning_rate": 1.5923076923076922e-07,
      "loss": -0.03,
      "reward": 0.799513041973114,
      "reward_std": 0.012985162436962128,
      "reward_std_all": 0.012902028625831008,
      "rewards-max/forensics_sinkhorn_reward": 0.9814829528331757,
      "rewards-max/forensics_strict_edge_reward": 0.6804680824279785,
      "rewards-max/forensics_tiou_reward": 0.9801510274410248,
      "rewards/forensics_sinkhorn_reward": 0.8309118449687958,
      "rewards/forensics_strict_edge_reward": 0.101871058344841,
      "rewards/forensics_tiou_reward": 0.6972506940364838,
      "step": 1094
    },
    {
      "completion_length": 25.90625,
      "epoch": 3.359447004608295,
      "grad_norm": 22.493837356567383,
      "kl": 0.9853515625,
      "learning_rate": 1.5846153846153846e-07,
      "loss": -0.0613,
      "reward": 0.8010403513908386,
      "reward_std": 0.010137398377992213,
      "reward_std_all": 0.010477811098098755,
      "rewards-max/forensics_sinkhorn_reward": 0.9824202656745911,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9827824831008911,
      "rewards/forensics_sinkhorn_reward": 0.8282378911972046,
      "rewards/forensics_strict_edge_reward": 0.07866938039660454,
      "rewards/forensics_tiou_reward": 0.7215800881385803,
      "step": 1095
    },
    {
      "completion_length": 22.09375,
      "epoch": 3.3625192012288787,
      "grad_norm": 10.147409439086914,
      "kl": 0.451171875,
      "learning_rate": 1.5769230769230767e-07,
      "loss": 0.0144,
      "reward": 0.7901986241340637,
      "reward_std": 0.026874998584389687,
      "reward_std_all": 0.02541065216064453,
      "rewards-max/forensics_sinkhorn_reward": 0.9800204038619995,
      "rewards-max/forensics_strict_edge_reward": 0.7372014820575714,
      "rewards-max/forensics_tiou_reward": 0.9771124124526978,
      "rewards/forensics_sinkhorn_reward": 0.7855680584907532,
      "rewards/forensics_strict_edge_reward": 0.065608911216259,
      "rewards/forensics_tiou_reward": 0.5638593137264252,
      "step": 1096
    },
    {
      "completion_length": 24.203125,
      "epoch": 3.3655913978494625,
      "grad_norm": 7.240817546844482,
      "kl": 0.529296875,
      "learning_rate": 1.569230769230769e-07,
      "loss": -0.0547,
      "reward": 0.7769312858581543,
      "reward_std": 0.00608657905831933,
      "reward_std_all": 0.02000182820484042,
      "rewards-max/forensics_sinkhorn_reward": 0.9741944074630737,
      "rewards-max/forensics_strict_edge_reward": 0.657658040523529,
      "rewards-max/forensics_tiou_reward": 0.9644163846969604,
      "rewards/forensics_sinkhorn_reward": 0.7633861303329468,
      "rewards/forensics_strict_edge_reward": 0.06903867796063423,
      "rewards/forensics_tiou_reward": 0.6051236689090729,
      "step": 1097
    },
    {
      "completion_length": 25.765625,
      "epoch": 3.3686635944700463,
      "grad_norm": 6.99114990234375,
      "kl": 0.396484375,
      "learning_rate": 1.5615384615384612e-07,
      "loss": -0.0361,
      "reward": 0.7376264035701752,
      "reward_std": 0.03380489442497492,
      "reward_std_all": 0.053639400750398636,
      "rewards-max/forensics_sinkhorn_reward": 0.9809552133083344,
      "rewards-max/forensics_strict_edge_reward": 0.4715774208307266,
      "rewards-max/forensics_tiou_reward": 0.9789665639400482,
      "rewards/forensics_sinkhorn_reward": 0.7502242922782898,
      "rewards/forensics_strict_edge_reward": 0.08013652637600899,
      "rewards/forensics_tiou_reward": 0.6467324197292328,
      "step": 1098
    },
    {
      "completion_length": 26.671875,
      "epoch": 3.3717357910906296,
      "grad_norm": 8.481124877929688,
      "kl": 0.53515625,
      "learning_rate": 1.5538461538461536e-07,
      "loss": 0.025,
      "reward": 0.7792261838912964,
      "reward_std": 0.03016385855153203,
      "reward_std_all": 0.030284568667411804,
      "rewards-max/forensics_sinkhorn_reward": 0.9712556600570679,
      "rewards-max/forensics_strict_edge_reward": 0.6633935272693634,
      "rewards-max/forensics_tiou_reward": 0.9597952663898468,
      "rewards/forensics_sinkhorn_reward": 0.7812570035457611,
      "rewards/forensics_strict_edge_reward": 0.05537596717476845,
      "rewards/forensics_tiou_reward": 0.6455189883708954,
      "step": 1099
    },
    {
      "completion_length": 26.3125,
      "epoch": 3.3748079877112134,
      "grad_norm": 7.07072639465332,
      "kl": 0.4970703125,
      "learning_rate": 1.5461538461538463e-07,
      "loss": 0.0002,
      "reward": 0.7737448811531067,
      "reward_std": 0.04435434751212597,
      "reward_std_all": 0.04187263618223369,
      "rewards-max/forensics_sinkhorn_reward": 0.9730709791183472,
      "rewards-max/forensics_strict_edge_reward": 0.5669038891792297,
      "rewards-max/forensics_tiou_reward": 0.9620139598846436,
      "rewards/forensics_sinkhorn_reward": 0.7946064472198486,
      "rewards/forensics_strict_edge_reward": 0.058986484073102474,
      "rewards/forensics_tiou_reward": 0.6470272243022919,
      "step": 1100
    },
    {
      "completion_length": 27.34375,
      "epoch": 3.377880184331797,
      "grad_norm": 8.029314041137695,
      "kl": 0.384765625,
      "learning_rate": 1.5384615384615385e-07,
      "loss": -0.0565,
      "reward": 0.792539656162262,
      "reward_std": 0.011438653571531177,
      "reward_std_all": 0.013803312554955482,
      "rewards-max/forensics_sinkhorn_reward": 0.9741097986698151,
      "rewards-max/forensics_strict_edge_reward": 0.45834553241729736,
      "rewards-max/forensics_tiou_reward": 0.9623377919197083,
      "rewards/forensics_sinkhorn_reward": 0.7931875288486481,
      "rewards/forensics_strict_edge_reward": 0.06045161373913288,
      "rewards/forensics_tiou_reward": 0.6223213076591492,
      "step": 1101
    },
    {
      "completion_length": 24.359375,
      "epoch": 3.380952380952381,
      "grad_norm": 10.69784927368164,
      "kl": 0.3994140625,
      "learning_rate": 1.5307692307692309e-07,
      "loss": -0.0318,
      "reward": 0.7679948210716248,
      "reward_std": 0.03755173087120056,
      "reward_std_all": 0.038433070294559,
      "rewards-max/forensics_sinkhorn_reward": 0.9860078394412994,
      "rewards-max/forensics_strict_edge_reward": 0.3763222470879555,
      "rewards-max/forensics_tiou_reward": 0.9926386475563049,
      "rewards/forensics_sinkhorn_reward": 0.7802963852882385,
      "rewards/forensics_strict_edge_reward": 0.049925814382731915,
      "rewards/forensics_tiou_reward": 0.5947907567024231,
      "step": 1102
    },
    {
      "completion_length": 26.453125,
      "epoch": 3.3840245775729647,
      "grad_norm": 8.047477722167969,
      "kl": 0.5244140625,
      "learning_rate": 1.523076923076923e-07,
      "loss": -0.0707,
      "reward": 0.7903751730918884,
      "reward_std": 0.026667926576919854,
      "reward_std_all": 0.02531839138828218,
      "rewards-max/forensics_sinkhorn_reward": 0.9588280618190765,
      "rewards-max/forensics_strict_edge_reward": 0.5183888971805573,
      "rewards-max/forensics_tiou_reward": 0.9263779819011688,
      "rewards/forensics_sinkhorn_reward": 0.8041556477546692,
      "rewards/forensics_strict_edge_reward": 0.05873877368867397,
      "rewards/forensics_tiou_reward": 0.6126944422721863,
      "step": 1103
    },
    {
      "completion_length": 26.34375,
      "epoch": 3.3870967741935485,
      "grad_norm": 8.251212120056152,
      "kl": 0.5205078125,
      "learning_rate": 1.5153846153846154e-07,
      "loss": -0.0523,
      "reward": 0.7900270223617554,
      "reward_std": 0.013274030294269323,
      "reward_std_all": 0.01653466559946537,
      "rewards-max/forensics_sinkhorn_reward": 0.9814043045043945,
      "rewards-max/forensics_strict_edge_reward": 0.7094893157482147,
      "rewards-max/forensics_tiou_reward": 0.9801588356494904,
      "rewards/forensics_sinkhorn_reward": 0.8175175189971924,
      "rewards/forensics_strict_edge_reward": 0.08446909114718437,
      "rewards/forensics_tiou_reward": 0.6782831251621246,
      "step": 1104
    },
    {
      "completion_length": 22.078125,
      "epoch": 3.3901689708141323,
      "grad_norm": 8.856691360473633,
      "kl": 0.486328125,
      "learning_rate": 1.5076923076923075e-07,
      "loss": -0.0566,
      "reward": 0.7888995110988617,
      "reward_std": 0.025462694466114044,
      "reward_std_all": 0.024589509703218937,
      "rewards-max/forensics_sinkhorn_reward": 0.9800242781639099,
      "rewards-max/forensics_strict_edge_reward": 0.7165639698505402,
      "rewards-max/forensics_tiou_reward": 0.9754171669483185,
      "rewards/forensics_sinkhorn_reward": 0.8180207312107086,
      "rewards/forensics_strict_edge_reward": 0.09122344478964806,
      "rewards/forensics_tiou_reward": 0.6714551746845245,
      "step": 1105
    },
    {
      "completion_length": 23.234375,
      "epoch": 3.3932411674347156,
      "grad_norm": 7.125377655029297,
      "kl": 0.4580078125,
      "learning_rate": 1.5e-07,
      "loss": 0.0139,
      "reward": 0.7897776365280151,
      "reward_std": 0.02202211134135723,
      "reward_std_all": 0.021548155695199966,
      "rewards-max/forensics_sinkhorn_reward": 0.9880739748477936,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9975005984306335,
      "rewards/forensics_sinkhorn_reward": 0.778386265039444,
      "rewards/forensics_strict_edge_reward": 0.10726242139935493,
      "rewards/forensics_tiou_reward": 0.6423884928226471,
      "step": 1106
    },
    {
      "completion_length": 23.359375,
      "epoch": 3.3963133640552994,
      "grad_norm": 8.323795318603516,
      "kl": 0.51171875,
      "learning_rate": 1.492307692307692e-07,
      "loss": 0.031,
      "reward": 0.7882315218448639,
      "reward_std": 0.022141582798212767,
      "reward_std_all": 0.022187280002981424,
      "rewards-max/forensics_sinkhorn_reward": 0.9826464653015137,
      "rewards-max/forensics_strict_edge_reward": 0.6499901115894318,
      "rewards-max/forensics_tiou_reward": 0.9818407893180847,
      "rewards/forensics_sinkhorn_reward": 0.8278547525405884,
      "rewards/forensics_strict_edge_reward": 0.09461449086666107,
      "rewards/forensics_tiou_reward": 0.688225120306015,
      "step": 1107
    },
    {
      "completion_length": 27.796875,
      "epoch": 3.399385560675883,
      "grad_norm": 12.012640953063965,
      "kl": 0.8525390625,
      "learning_rate": 1.4846153846153845e-07,
      "loss": 0.012,
      "reward": 0.7574091255664825,
      "reward_std": 0.04396177176386118,
      "reward_std_all": 0.04633750952780247,
      "rewards-max/forensics_sinkhorn_reward": 0.9845497310161591,
      "rewards-max/forensics_strict_edge_reward": 0.6284831166267395,
      "rewards-max/forensics_tiou_reward": 0.9855365753173828,
      "rewards/forensics_sinkhorn_reward": 0.7873897850513458,
      "rewards/forensics_strict_edge_reward": 0.10250771604478359,
      "rewards/forensics_tiou_reward": 0.6816222667694092,
      "step": 1108
    },
    {
      "completion_length": 28.1875,
      "epoch": 3.402457757296467,
      "grad_norm": 7.333463668823242,
      "kl": 0.447265625,
      "learning_rate": 1.476923076923077e-07,
      "loss": 0.0172,
      "reward": 0.7519248127937317,
      "reward_std": 0.030435738619416952,
      "reward_std_all": 0.044041181448847055,
      "rewards-max/forensics_sinkhorn_reward": 0.9796288907527924,
      "rewards-max/forensics_strict_edge_reward": 0.7126307189464569,
      "rewards-max/forensics_tiou_reward": 0.9776610136032104,
      "rewards/forensics_sinkhorn_reward": 0.7903591096401215,
      "rewards/forensics_strict_edge_reward": 0.07937153801321983,
      "rewards/forensics_tiou_reward": 0.6806334257125854,
      "step": 1109
    },
    {
      "completion_length": 26.859375,
      "epoch": 3.4055299539170507,
      "grad_norm": 7.1216511726379395,
      "kl": 0.4501953125,
      "learning_rate": 1.4692307692307693e-07,
      "loss": -0.1107,
      "reward": 0.8051947057247162,
      "reward_std": 0.002378587203565985,
      "reward_std_all": 0.004015674348920584,
      "rewards-max/forensics_sinkhorn_reward": 0.9827799797058105,
      "rewards-max/forensics_strict_edge_reward": 0.8106843829154968,
      "rewards-max/forensics_tiou_reward": 0.9837907254695892,
      "rewards/forensics_sinkhorn_reward": 0.8416378200054169,
      "rewards/forensics_strict_edge_reward": 0.08129589632153511,
      "rewards/forensics_tiou_reward": 0.7182222604751587,
      "step": 1110
    },
    {
      "completion_length": 29.578125,
      "epoch": 3.4086021505376345,
      "grad_norm": 8.441611289978027,
      "kl": 0.494140625,
      "learning_rate": 1.4615384615384617e-07,
      "loss": 0.0191,
      "reward": 0.7684431076049805,
      "reward_std": 0.011242377338930964,
      "reward_std_all": 0.027441917918622494,
      "rewards-max/forensics_sinkhorn_reward": 0.9765113592147827,
      "rewards-max/forensics_strict_edge_reward": 0.5559668838977814,
      "rewards-max/forensics_tiou_reward": 0.9647878110408783,
      "rewards/forensics_sinkhorn_reward": 0.7829716801643372,
      "rewards/forensics_strict_edge_reward": 0.07098167762160301,
      "rewards/forensics_tiou_reward": 0.6648009121417999,
      "step": 1111
    },
    {
      "completion_length": 24.5,
      "epoch": 3.4116743471582183,
      "grad_norm": 11.167174339294434,
      "kl": 0.423828125,
      "learning_rate": 1.4538461538461538e-07,
      "loss": -0.0308,
      "reward": 0.7881740927696228,
      "reward_std": 0.029425843036733568,
      "reward_std_all": 0.02764429314993322,
      "rewards-max/forensics_sinkhorn_reward": 0.9731641113758087,
      "rewards-max/forensics_strict_edge_reward": 0.5340558886528015,
      "rewards-max/forensics_tiou_reward": 0.9457988142967224,
      "rewards/forensics_sinkhorn_reward": 0.798535168170929,
      "rewards/forensics_strict_edge_reward": 0.05092519707977772,
      "rewards/forensics_tiou_reward": 0.6491785645484924,
      "step": 1112
    },
    {
      "completion_length": 26.78125,
      "epoch": 3.4147465437788016,
      "grad_norm": 8.017863273620605,
      "kl": 0.4296875,
      "learning_rate": 1.4461538461538462e-07,
      "loss": -0.014,
      "reward": 0.7962974905967712,
      "reward_std": 0.017582000233232975,
      "reward_std_all": 0.016968317795544863,
      "rewards-max/forensics_sinkhorn_reward": 0.9765187799930573,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9643073678016663,
      "rewards/forensics_sinkhorn_reward": 0.7944779694080353,
      "rewards/forensics_strict_edge_reward": 0.08458535000681877,
      "rewards/forensics_tiou_reward": 0.6553311347961426,
      "step": 1113
    },
    {
      "completion_length": 23.90625,
      "epoch": 3.4178187403993854,
      "grad_norm": 8.20090389251709,
      "kl": 0.4912109375,
      "learning_rate": 1.4384615384615384e-07,
      "loss": 0.0023,
      "reward": 0.8002930283546448,
      "reward_std": 0.01063226955011487,
      "reward_std_all": 0.011066826526075602,
      "rewards-max/forensics_sinkhorn_reward": 0.9718676507472992,
      "rewards-max/forensics_strict_edge_reward": 0.4697008728981018,
      "rewards-max/forensics_tiou_reward": 0.9584283530712128,
      "rewards/forensics_sinkhorn_reward": 0.8358198404312134,
      "rewards/forensics_strict_edge_reward": 0.06664030998945236,
      "rewards/forensics_tiou_reward": 0.675952136516571,
      "step": 1114
    },
    {
      "completion_length": 24.28125,
      "epoch": 3.420890937019969,
      "grad_norm": 8.406106948852539,
      "kl": 0.48828125,
      "learning_rate": 1.4307692307692308e-07,
      "loss": -0.0518,
      "reward": 0.7924798727035522,
      "reward_std": 0.008813170716166496,
      "reward_std_all": 0.01249281782656908,
      "rewards-max/forensics_sinkhorn_reward": 0.975578635931015,
      "rewards-max/forensics_strict_edge_reward": 0.4582431763410568,
      "rewards-max/forensics_tiou_reward": 0.9661845564842224,
      "rewards/forensics_sinkhorn_reward": 0.757091224193573,
      "rewards/forensics_strict_edge_reward": 0.04193546902388334,
      "rewards/forensics_tiou_reward": 0.6107158958911896,
      "step": 1115
    },
    {
      "completion_length": 24.3125,
      "epoch": 3.423963133640553,
      "grad_norm": 8.487955093383789,
      "kl": 0.48828125,
      "learning_rate": 1.423076923076923e-07,
      "loss": -0.0572,
      "reward": 0.7973885536193848,
      "reward_std": 0.012119106540922076,
      "reward_std_all": 0.0131682496285066,
      "rewards-max/forensics_sinkhorn_reward": 0.978970468044281,
      "rewards-max/forensics_strict_edge_reward": 0.5066563785076141,
      "rewards-max/forensics_tiou_reward": 0.9730105698108673,
      "rewards/forensics_sinkhorn_reward": 0.7690581977367401,
      "rewards/forensics_strict_edge_reward": 0.02880884986370802,
      "rewards/forensics_tiou_reward": 0.6109191030263901,
      "step": 1116
    },
    {
      "completion_length": 26.265625,
      "epoch": 3.4270353302611367,
      "grad_norm": 6.749483585357666,
      "kl": 0.4921875,
      "learning_rate": 1.4153846153846153e-07,
      "loss": 0.0216,
      "reward": 0.7263530790805817,
      "reward_std": 0.029617659281939268,
      "reward_std_all": 0.05858847498893738,
      "rewards-max/forensics_sinkhorn_reward": 0.9782349467277527,
      "rewards-max/forensics_strict_edge_reward": 0.4240923821926117,
      "rewards-max/forensics_tiou_reward": 0.9700339138507843,
      "rewards/forensics_sinkhorn_reward": 0.7370535135269165,
      "rewards/forensics_strict_edge_reward": 0.040829531848430634,
      "rewards/forensics_tiou_reward": 0.6193844974040985,
      "step": 1117
    },
    {
      "completion_length": 24.90625,
      "epoch": 3.4301075268817205,
      "grad_norm": 6.892155170440674,
      "kl": 0.55078125,
      "learning_rate": 1.4076923076923074e-07,
      "loss": 0.127,
      "reward": 0.772322416305542,
      "reward_std": 0.03182759368792176,
      "reward_std_all": 0.035662548150867224,
      "rewards-max/forensics_sinkhorn_reward": 0.9801456928253174,
      "rewards-max/forensics_strict_edge_reward": 0.6919131278991699,
      "rewards-max/forensics_tiou_reward": 0.9810494184494019,
      "rewards/forensics_sinkhorn_reward": 0.7697759866714478,
      "rewards/forensics_strict_edge_reward": 0.10306144878268242,
      "rewards/forensics_tiou_reward": 0.6442201733589172,
      "step": 1118
    },
    {
      "completion_length": 21.671875,
      "epoch": 3.4331797235023043,
      "grad_norm": 8.195488929748535,
      "kl": 0.3779296875,
      "learning_rate": 1.4e-07,
      "loss": -0.1083,
      "reward": 0.7749082744121552,
      "reward_std": 0.03774447925388813,
      "reward_std_all": 0.03875340148806572,
      "rewards-max/forensics_sinkhorn_reward": 0.9813561141490936,
      "rewards-max/forensics_strict_edge_reward": 0.5901983380317688,
      "rewards-max/forensics_tiou_reward": 0.9799736738204956,
      "rewards/forensics_sinkhorn_reward": 0.7914914190769196,
      "rewards/forensics_strict_edge_reward": 0.06692708190530539,
      "rewards/forensics_tiou_reward": 0.5933175384998322,
      "step": 1119
    },
    {
      "completion_length": 24.4375,
      "epoch": 3.436251920122888,
      "grad_norm": 7.340615272521973,
      "kl": 0.390625,
      "learning_rate": 1.3923076923076922e-07,
      "loss": -0.0777,
      "reward": 0.7864098250865936,
      "reward_std": 0.023149138316512108,
      "reward_std_all": 0.02373324683867395,
      "rewards-max/forensics_sinkhorn_reward": 0.9780797362327576,
      "rewards-max/forensics_strict_edge_reward": 0.28830963373184204,
      "rewards-max/forensics_tiou_reward": 0.9697015881538391,
      "rewards/forensics_sinkhorn_reward": 0.7613763213157654,
      "rewards/forensics_strict_edge_reward": 0.029938452877104282,
      "rewards/forensics_tiou_reward": 0.586475059390068,
      "step": 1120
    },
    {
      "completion_length": 22.859375,
      "epoch": 3.439324116743472,
      "grad_norm": 8.77281665802002,
      "kl": 0.513671875,
      "learning_rate": 1.3846153846153846e-07,
      "loss": 0.1006,
      "reward": 0.7842286229133606,
      "reward_std": 0.01690382556989789,
      "reward_std_all": 0.022125284653156996,
      "rewards-max/forensics_sinkhorn_reward": 0.9806243479251862,
      "rewards-max/forensics_strict_edge_reward": 0.7190383970737457,
      "rewards-max/forensics_tiou_reward": 0.9760976731777191,
      "rewards/forensics_sinkhorn_reward": 0.8198046386241913,
      "rewards/forensics_strict_edge_reward": 0.06577916257083416,
      "rewards/forensics_tiou_reward": 0.6820975840091705,
      "step": 1121
    },
    {
      "completion_length": 26.34375,
      "epoch": 3.442396313364055,
      "grad_norm": 8.699688911437988,
      "kl": 0.423828125,
      "learning_rate": 1.376923076923077e-07,
      "loss": 0.0354,
      "reward": 0.79567950963974,
      "reward_std": 0.017468495294451714,
      "reward_std_all": 0.017202572897076607,
      "rewards-max/forensics_sinkhorn_reward": 0.9834668636322021,
      "rewards-max/forensics_strict_edge_reward": 0.6802245080471039,
      "rewards-max/forensics_tiou_reward": 0.985212504863739,
      "rewards/forensics_sinkhorn_reward": 0.7832840085029602,
      "rewards/forensics_strict_edge_reward": 0.06682095304131508,
      "rewards/forensics_tiou_reward": 0.6186608076095581,
      "step": 1122
    },
    {
      "completion_length": 26.765625,
      "epoch": 3.445468509984639,
      "grad_norm": 7.542174816131592,
      "kl": 0.4453125,
      "learning_rate": 1.3692307692307692e-07,
      "loss": -0.0322,
      "reward": 0.787827730178833,
      "reward_std": 0.02063167782034725,
      "reward_std_all": 0.021251876605674624,
      "rewards-max/forensics_sinkhorn_reward": 0.973191887140274,
      "rewards-max/forensics_strict_edge_reward": 0.5989546254277229,
      "rewards-max/forensics_tiou_reward": 0.9636947810649872,
      "rewards/forensics_sinkhorn_reward": 0.8107272684574127,
      "rewards/forensics_strict_edge_reward": 0.08659618347883224,
      "rewards/forensics_tiou_reward": 0.7059511244297028,
      "step": 1123
    },
    {
      "completion_length": 30.171875,
      "epoch": 3.4485407066052227,
      "grad_norm": 7.463185787200928,
      "kl": 0.6484375,
      "learning_rate": 1.3615384615384616e-07,
      "loss": -0.0301,
      "reward": 0.8042540848255157,
      "reward_std": 0.0035511734313331544,
      "reward_std_all": 0.005242559942416847,
      "rewards-max/forensics_sinkhorn_reward": 0.9817118346691132,
      "rewards-max/forensics_strict_edge_reward": 0.45318591594696045,
      "rewards-max/forensics_tiou_reward": 0.9821093380451202,
      "rewards/forensics_sinkhorn_reward": 0.8433398008346558,
      "rewards/forensics_strict_edge_reward": 0.0397674348205328,
      "rewards/forensics_tiou_reward": 0.6944495141506195,
      "step": 1124
    },
    {
      "completion_length": 22.96875,
      "epoch": 3.4516129032258065,
      "grad_norm": 9.508462905883789,
      "kl": 0.4306640625,
      "learning_rate": 1.3538461538461537e-07,
      "loss": 0.0649,
      "reward": 0.7610759735107422,
      "reward_std": 0.021563903894275427,
      "reward_std_all": 0.036856313701719046,
      "rewards-max/forensics_sinkhorn_reward": 0.9789907932281494,
      "rewards-max/forensics_strict_edge_reward": 0.6265257000923157,
      "rewards-max/forensics_tiou_reward": 0.9699369370937347,
      "rewards/forensics_sinkhorn_reward": 0.7395515441894531,
      "rewards/forensics_strict_edge_reward": 0.07248508557677269,
      "rewards/forensics_tiou_reward": 0.5782111287117004,
      "step": 1125
    },
    {
      "completion_length": 24.25,
      "epoch": 3.4546850998463903,
      "grad_norm": 8.433640480041504,
      "kl": 0.4951171875,
      "learning_rate": 1.346153846153846e-07,
      "loss": -0.0533,
      "reward": 0.8007605671882629,
      "reward_std": 0.010115007404237986,
      "reward_std_all": 0.01078543858602643,
      "rewards-max/forensics_sinkhorn_reward": 0.9786654114723206,
      "rewards-max/forensics_strict_edge_reward": 0.4970944970846176,
      "rewards-max/forensics_tiou_reward": 0.9767196774482727,
      "rewards/forensics_sinkhorn_reward": 0.8497094213962555,
      "rewards/forensics_strict_edge_reward": 0.08891569077968597,
      "rewards/forensics_tiou_reward": 0.6857690215110779,
      "step": 1126
    },
    {
      "completion_length": 25.265625,
      "epoch": 3.457757296466974,
      "grad_norm": 8.691206932067871,
      "kl": 0.4912109375,
      "learning_rate": 1.3384615384615383e-07,
      "loss": -0.0379,
      "reward": 0.7965878844261169,
      "reward_std": 0.007221098057925701,
      "reward_std_all": 0.01049278024584055,
      "rewards-max/forensics_sinkhorn_reward": 0.9836570918560028,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9816009104251862,
      "rewards/forensics_sinkhorn_reward": 0.778884768486023,
      "rewards/forensics_strict_edge_reward": 0.10480378940701485,
      "rewards/forensics_tiou_reward": 0.6645862460136414,
      "step": 1127
    },
    {
      "completion_length": 28.578125,
      "epoch": 3.460829493087558,
      "grad_norm": 7.578115463256836,
      "kl": 0.3857421875,
      "learning_rate": 1.3307692307692307e-07,
      "loss": 0.0649,
      "reward": 0.7687710523605347,
      "reward_std": 0.03615673445165157,
      "reward_std_all": 0.03762619383633137,
      "rewards-max/forensics_sinkhorn_reward": 0.969568520784378,
      "rewards-max/forensics_strict_edge_reward": 0.24783098697662354,
      "rewards-max/forensics_tiou_reward": 0.9454940259456635,
      "rewards/forensics_sinkhorn_reward": 0.7345350682735443,
      "rewards/forensics_strict_edge_reward": 0.028686849400401115,
      "rewards/forensics_tiou_reward": 0.5844832062721252,
      "step": 1128
    },
    {
      "completion_length": 23.921875,
      "epoch": 3.463901689708141,
      "grad_norm": 8.927268028259277,
      "kl": 0.576171875,
      "learning_rate": 1.323076923076923e-07,
      "loss": 0.0111,
      "reward": 0.7501541376113892,
      "reward_std": 0.031992883421480656,
      "reward_std_all": 0.04800823889672756,
      "rewards-max/forensics_sinkhorn_reward": 0.978861391544342,
      "rewards-max/forensics_strict_edge_reward": 0.6779125332832336,
      "rewards-max/forensics_tiou_reward": 0.9760176241397858,
      "rewards/forensics_sinkhorn_reward": 0.7407063245773315,
      "rewards/forensics_strict_edge_reward": 0.06800803914666176,
      "rewards/forensics_tiou_reward": 0.5978453010320663,
      "step": 1129
    },
    {
      "completion_length": 24.71875,
      "epoch": 3.466973886328725,
      "grad_norm": 7.586572647094727,
      "kl": 0.3212890625,
      "learning_rate": 1.3153846153846155e-07,
      "loss": -0.0501,
      "reward": 0.7578601241111755,
      "reward_std": 0.03731144033372402,
      "reward_std_all": 0.04312529042363167,
      "rewards-max/forensics_sinkhorn_reward": 0.9803508222103119,
      "rewards-max/forensics_strict_edge_reward": 0.5933051109313965,
      "rewards-max/forensics_tiou_reward": 0.9768711924552917,
      "rewards/forensics_sinkhorn_reward": 0.7363434731960297,
      "rewards/forensics_strict_edge_reward": 0.09521175175905228,
      "rewards/forensics_tiou_reward": 0.61673504114151,
      "step": 1130
    },
    {
      "completion_length": 26.34375,
      "epoch": 3.4700460829493087,
      "grad_norm": 7.619264602661133,
      "kl": 0.443359375,
      "learning_rate": 1.3076923076923076e-07,
      "loss": -0.0583,
      "reward": 0.7907324135303497,
      "reward_std": 0.012294185347855091,
      "reward_std_all": 0.015079062897711992,
      "rewards-max/forensics_sinkhorn_reward": 0.9814891815185547,
      "rewards-max/forensics_strict_edge_reward": 0.7676310539245605,
      "rewards-max/forensics_tiou_reward": 0.9798512756824493,
      "rewards/forensics_sinkhorn_reward": 0.756822019815445,
      "rewards/forensics_strict_edge_reward": 0.11954877525568008,
      "rewards/forensics_tiou_reward": 0.6539525091648102,
      "step": 1131
    },
    {
      "completion_length": 24.640625,
      "epoch": 3.4731182795698925,
      "grad_norm": 6.741349697113037,
      "kl": 0.3701171875,
      "learning_rate": 1.3e-07,
      "loss": -0.0798,
      "reward": 0.7807644009590149,
      "reward_std": 0.026484920468647033,
      "reward_std_all": 0.027263460564427078,
      "rewards-max/forensics_sinkhorn_reward": 0.9777999818325043,
      "rewards-max/forensics_strict_edge_reward": 0.6495925933122635,
      "rewards-max/forensics_tiou_reward": 0.9713070690631866,
      "rewards/forensics_sinkhorn_reward": 0.8232356309890747,
      "rewards/forensics_strict_edge_reward": 0.08586856350302696,
      "rewards/forensics_tiou_reward": 0.6642768383026123,
      "step": 1132
    },
    {
      "completion_length": 27.453125,
      "epoch": 3.4761904761904763,
      "grad_norm": 11.557877540588379,
      "kl": 0.474609375,
      "learning_rate": 1.2923076923076924e-07,
      "loss": 0.0352,
      "reward": 0.7511657774448395,
      "reward_std": 0.031358528416603804,
      "reward_std_all": 0.040846534073352814,
      "rewards-max/forensics_sinkhorn_reward": 0.9769265353679657,
      "rewards-max/forensics_strict_edge_reward": 0.6363745033740997,
      "rewards-max/forensics_tiou_reward": 0.9704696536064148,
      "rewards/forensics_sinkhorn_reward": 0.670904666185379,
      "rewards/forensics_strict_edge_reward": 0.07310184091329575,
      "rewards/forensics_tiou_reward": 0.572350800037384,
      "step": 1133
    },
    {
      "completion_length": 32.078125,
      "epoch": 3.47926267281106,
      "grad_norm": 6.246967315673828,
      "kl": 0.4462890625,
      "learning_rate": 1.2846153846153845e-07,
      "loss": 0.0526,
      "reward": 0.7870170772075653,
      "reward_std": 0.029219494783319533,
      "reward_std_all": 0.02782326377928257,
      "rewards-max/forensics_sinkhorn_reward": 0.9866972863674164,
      "rewards-max/forensics_strict_edge_reward": 0.6271176636219025,
      "rewards-max/forensics_tiou_reward": 0.9931211471557617,
      "rewards/forensics_sinkhorn_reward": 0.8118334114551544,
      "rewards/forensics_strict_edge_reward": 0.07592049986124039,
      "rewards/forensics_tiou_reward": 0.7150661051273346,
      "step": 1134
    },
    {
      "completion_length": 25.421875,
      "epoch": 3.482334869431644,
      "grad_norm": 6.984889984130859,
      "kl": 0.384765625,
      "learning_rate": 1.276923076923077e-07,
      "loss": -0.0069,
      "reward": 0.8034331500530243,
      "reward_std": 0.005254682386294007,
      "reward_std_all": 0.006371104158461094,
      "rewards-max/forensics_sinkhorn_reward": 0.9844915270805359,
      "rewards-max/forensics_strict_edge_reward": 0.6964708566665649,
      "rewards-max/forensics_tiou_reward": 0.985584557056427,
      "rewards/forensics_sinkhorn_reward": 0.8446691632270813,
      "rewards/forensics_strict_edge_reward": 0.135654054582119,
      "rewards/forensics_tiou_reward": 0.7120084762573242,
      "step": 1135
    },
    {
      "completion_length": 22.984375,
      "epoch": 3.485407066052227,
      "grad_norm": 8.343689918518066,
      "kl": 0.59375,
      "learning_rate": 1.269230769230769e-07,
      "loss": 0.0194,
      "reward": 0.7743388116359711,
      "reward_std": 0.022531490307301283,
      "reward_std_all": 0.030572769232094288,
      "rewards-max/forensics_sinkhorn_reward": 0.984167069196701,
      "rewards-max/forensics_strict_edge_reward": 0.6184562146663666,
      "rewards-max/forensics_tiou_reward": 0.988193154335022,
      "rewards/forensics_sinkhorn_reward": 0.7714899778366089,
      "rewards/forensics_strict_edge_reward": 0.07008343189954758,
      "rewards/forensics_tiou_reward": 0.5837367475032806,
      "step": 1136
    },
    {
      "completion_length": 23.921875,
      "epoch": 3.488479262672811,
      "grad_norm": 10.125463485717773,
      "kl": 0.623046875,
      "learning_rate": 1.2615384615384615e-07,
      "loss": -0.0221,
      "reward": 0.787555456161499,
      "reward_std": 0.03000417910516262,
      "reward_std_all": 0.028216345235705376,
      "rewards-max/forensics_sinkhorn_reward": 0.9867368936538696,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9930828213691711,
      "rewards/forensics_sinkhorn_reward": 0.7636304497718811,
      "rewards/forensics_strict_edge_reward": 0.10020437091588974,
      "rewards/forensics_tiou_reward": 0.6101157665252686,
      "step": 1137
    },
    {
      "completion_length": 22.34375,
      "epoch": 3.4915514592933947,
      "grad_norm": 9.261688232421875,
      "kl": 0.3916015625,
      "learning_rate": 1.2538461538461536e-07,
      "loss": -0.0268,
      "reward": 0.7834219038486481,
      "reward_std": 0.04096589423716068,
      "reward_std_all": 0.03695589583367109,
      "rewards-max/forensics_sinkhorn_reward": 0.9798399806022644,
      "rewards-max/forensics_strict_edge_reward": 0.7056996524333954,
      "rewards-max/forensics_tiou_reward": 0.9732100665569305,
      "rewards/forensics_sinkhorn_reward": 0.8406306803226471,
      "rewards/forensics_strict_edge_reward": 0.09622934833168983,
      "rewards/forensics_tiou_reward": 0.7125698626041412,
      "step": 1138
    },
    {
      "completion_length": 24.109375,
      "epoch": 3.4946236559139785,
      "grad_norm": 13.678055763244629,
      "kl": 0.8173828125,
      "learning_rate": 1.246153846153846e-07,
      "loss": -0.0636,
      "reward": 0.7981860637664795,
      "reward_std": 0.014285784680396318,
      "reward_std_all": 0.014383328147232533,
      "rewards-max/forensics_sinkhorn_reward": 0.9823469519615173,
      "rewards-max/forensics_strict_edge_reward": 0.6513115167617798,
      "rewards-max/forensics_tiou_reward": 0.9831799864768982,
      "rewards/forensics_sinkhorn_reward": 0.7953223586082458,
      "rewards/forensics_strict_edge_reward": 0.09860492870211601,
      "rewards/forensics_tiou_reward": 0.6546567976474762,
      "step": 1139
    },
    {
      "completion_length": 23.578125,
      "epoch": 3.4976958525345623,
      "grad_norm": 7.187588214874268,
      "kl": 0.546875,
      "learning_rate": 1.2384615384615384e-07,
      "loss": -0.0434,
      "reward": 0.806024044752121,
      "reward_std": 0.0018094379338435829,
      "reward_std_all": 0.00336434505879879,
      "rewards-max/forensics_sinkhorn_reward": 0.9851460754871368,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9868040382862091,
      "rewards/forensics_sinkhorn_reward": 0.8583155870437622,
      "rewards/forensics_strict_edge_reward": 0.07248159311711788,
      "rewards/forensics_tiou_reward": 0.7542869746685028,
      "step": 1140
    },
    {
      "completion_length": 26.421875,
      "epoch": 3.500768049155146,
      "grad_norm": 6.0460710525512695,
      "kl": 0.44921875,
      "learning_rate": 1.2307692307692308e-07,
      "loss": -0.082,
      "reward": 0.8054247796535492,
      "reward_std": 0.001501347403973341,
      "reward_std_all": 0.0034492681734263897,
      "rewards-max/forensics_sinkhorn_reward": 0.9810336828231812,
      "rewards-max/forensics_strict_edge_reward": 0.4765545129776001,
      "rewards-max/forensics_tiou_reward": 0.9771198034286499,
      "rewards/forensics_sinkhorn_reward": 0.8281775116920471,
      "rewards/forensics_strict_edge_reward": 0.06243009492754936,
      "rewards/forensics_tiou_reward": 0.6897434592247009,
      "step": 1141
    },
    {
      "completion_length": 23.78125,
      "epoch": 3.50384024577573,
      "grad_norm": 8.422751426696777,
      "kl": 0.439453125,
      "learning_rate": 1.223076923076923e-07,
      "loss": 0.0217,
      "reward": 0.7962939143180847,
      "reward_std": 0.01661406923085451,
      "reward_std_all": 0.01657659187912941,
      "rewards-max/forensics_sinkhorn_reward": 0.980260044336319,
      "rewards-max/forensics_strict_edge_reward": 0.3881099596619606,
      "rewards-max/forensics_tiou_reward": 0.972063809633255,
      "rewards/forensics_sinkhorn_reward": 0.7812163531780243,
      "rewards/forensics_strict_edge_reward": 0.06020169332623482,
      "rewards/forensics_tiou_reward": 0.6444528102874756,
      "step": 1142
    },
    {
      "completion_length": 24.984375,
      "epoch": 3.506912442396313,
      "grad_norm": 8.180850982666016,
      "kl": 0.4873046875,
      "learning_rate": 1.2153846153846154e-07,
      "loss": 0.0259,
      "reward": 0.7386325597763062,
      "reward_std": 0.032763874158263206,
      "reward_std_all": 0.049656244926154613,
      "rewards-max/forensics_sinkhorn_reward": 0.9826170802116394,
      "rewards-max/forensics_strict_edge_reward": 0.5816013813018799,
      "rewards-max/forensics_tiou_reward": 0.9826786816120148,
      "rewards/forensics_sinkhorn_reward": 0.7626805901527405,
      "rewards/forensics_strict_edge_reward": 0.09379883110523224,
      "rewards/forensics_tiou_reward": 0.589766651391983,
      "step": 1143
    },
    {
      "completion_length": 24.53125,
      "epoch": 3.509984639016897,
      "grad_norm": 8.721921920776367,
      "kl": 0.4697265625,
      "learning_rate": 1.2076923076923075e-07,
      "loss": 0.0884,
      "reward": 0.7515222132205963,
      "reward_std": 0.030400443822145462,
      "reward_std_all": 0.04328414937481284,
      "rewards-max/forensics_sinkhorn_reward": 0.9831371903419495,
      "rewards-max/forensics_strict_edge_reward": 0.6914820820093155,
      "rewards-max/forensics_tiou_reward": 0.9780020713806152,
      "rewards/forensics_sinkhorn_reward": 0.711659848690033,
      "rewards/forensics_strict_edge_reward": 0.05003696121275425,
      "rewards/forensics_tiou_reward": 0.5279299914836884,
      "step": 1144
    },
    {
      "completion_length": 26.109375,
      "epoch": 3.5130568356374807,
      "grad_norm": 7.542220592498779,
      "kl": 0.599609375,
      "learning_rate": 1.2e-07,
      "loss": -0.0133,
      "reward": 0.804715633392334,
      "reward_std": 0.0016568060382269323,
      "reward_std_all": 0.0038657450350001454,
      "rewards-max/forensics_sinkhorn_reward": 0.9780277609825134,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9715897142887115,
      "rewards/forensics_sinkhorn_reward": 0.7907629907131195,
      "rewards/forensics_strict_edge_reward": 0.04926481284201145,
      "rewards/forensics_tiou_reward": 0.6102637052536011,
      "step": 1145
    },
    {
      "completion_length": 25.0625,
      "epoch": 3.5161290322580645,
      "grad_norm": 7.807558536529541,
      "kl": 0.5009765625,
      "learning_rate": 1.1923076923076923e-07,
      "loss": -0.0126,
      "reward": 0.7959108650684357,
      "reward_std": 0.01089141343254596,
      "reward_std_all": 0.012410697527229786,
      "rewards-max/forensics_sinkhorn_reward": 0.9872825443744659,
      "rewards-max/forensics_strict_edge_reward": 0.723440021276474,
      "rewards-max/forensics_tiou_reward": 0.9949682652950287,
      "rewards/forensics_sinkhorn_reward": 0.8052439987659454,
      "rewards/forensics_strict_edge_reward": 0.12247848510742188,
      "rewards/forensics_tiou_reward": 0.6725854575634003,
      "step": 1146
    },
    {
      "completion_length": 26.921875,
      "epoch": 3.5192012288786483,
      "grad_norm": 9.532981872558594,
      "kl": 0.484375,
      "learning_rate": 1.1846153846153846e-07,
      "loss": -0.0815,
      "reward": 0.7881172299385071,
      "reward_std": 0.020454718731343746,
      "reward_std_all": 0.02122487500309944,
      "rewards-max/forensics_sinkhorn_reward": 0.984594076871872,
      "rewards-max/forensics_strict_edge_reward": 0.6000702232122421,
      "rewards-max/forensics_tiou_reward": 0.9854572713375092,
      "rewards/forensics_sinkhorn_reward": 0.8138810396194458,
      "rewards/forensics_strict_edge_reward": 0.09767131134867668,
      "rewards/forensics_tiou_reward": 0.6784041523933411,
      "step": 1147
    },
    {
      "completion_length": 29.046875,
      "epoch": 3.522273425499232,
      "grad_norm": 9.23310375213623,
      "kl": 0.5048828125,
      "learning_rate": 1.1769230769230768e-07,
      "loss": -0.0415,
      "reward": 0.7716718316078186,
      "reward_std": 0.006258190725930035,
      "reward_std_all": 0.02224219497293234,
      "rewards-max/forensics_sinkhorn_reward": 0.9789469540119171,
      "rewards-max/forensics_strict_edge_reward": 0.5467688739299774,
      "rewards-max/forensics_tiou_reward": 0.9773077368736267,
      "rewards/forensics_sinkhorn_reward": 0.7530620098114014,
      "rewards/forensics_strict_edge_reward": 0.08725978434085846,
      "rewards/forensics_tiou_reward": 0.6342556178569794,
      "step": 1148
    },
    {
      "completion_length": 26.671875,
      "epoch": 3.525345622119816,
      "grad_norm": 7.8758769035339355,
      "kl": 0.369140625,
      "learning_rate": 1.1692307692307691e-07,
      "loss": -0.044,
      "reward": 0.7636886537075043,
      "reward_std": 0.04287336394190788,
      "reward_std_all": 0.04367738589644432,
      "rewards-max/forensics_sinkhorn_reward": 0.9752161502838135,
      "rewards-max/forensics_strict_edge_reward": 0.5427795201539993,
      "rewards-max/forensics_tiou_reward": 0.9662988185882568,
      "rewards/forensics_sinkhorn_reward": 0.7826468646526337,
      "rewards/forensics_strict_edge_reward": 0.061262357980012894,
      "rewards/forensics_tiou_reward": 0.6307640075683594,
      "step": 1149
    },
    {
      "completion_length": 24.4375,
      "epoch": 3.528417818740399,
      "grad_norm": 7.0416412353515625,
      "kl": 0.466796875,
      "learning_rate": 1.1615384615384615e-07,
      "loss": -0.0603,
      "reward": 0.7923890054225922,
      "reward_std": 0.024033826543018222,
      "reward_std_all": 0.022845877334475517,
      "rewards-max/forensics_sinkhorn_reward": 0.9845085442066193,
      "rewards-max/forensics_strict_edge_reward": 0.6616338789463043,
      "rewards-max/forensics_tiou_reward": 0.9893751442432404,
      "rewards/forensics_sinkhorn_reward": 0.8175631165504456,
      "rewards/forensics_strict_edge_reward": 0.08856735751032829,
      "rewards/forensics_tiou_reward": 0.6527620553970337,
      "step": 1150
    },
    {
      "completion_length": 30.234375,
      "epoch": 3.531490015360983,
      "grad_norm": 6.079166889190674,
      "kl": 0.5205078125,
      "learning_rate": 1.1538461538461539e-07,
      "loss": 0.0081,
      "reward": 0.7796356976032257,
      "reward_std": 0.027440744917839766,
      "reward_std_all": 0.02867064019665122,
      "rewards-max/forensics_sinkhorn_reward": 0.9859419763088226,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9910333752632141,
      "rewards/forensics_sinkhorn_reward": 0.8095962405204773,
      "rewards/forensics_strict_edge_reward": 0.09839346259832382,
      "rewards/forensics_tiou_reward": 0.6701188087463379,
      "step": 1151
    },
    {
      "completion_length": 23.9375,
      "epoch": 3.5345622119815667,
      "grad_norm": 7.974146842956543,
      "kl": 0.4326171875,
      "learning_rate": 1.1461538461538462e-07,
      "loss": -0.0074,
      "reward": 0.7948902249336243,
      "reward_std": 0.01992759993299842,
      "reward_std_all": 0.019005530513823032,
      "rewards-max/forensics_sinkhorn_reward": 0.9756087064743042,
      "rewards-max/forensics_strict_edge_reward": 0.6839397251605988,
      "rewards-max/forensics_tiou_reward": 0.962981104850769,
      "rewards/forensics_sinkhorn_reward": 0.8017567992210388,
      "rewards/forensics_strict_edge_reward": 0.06321447808295488,
      "rewards/forensics_tiou_reward": 0.6498333215713501,
      "step": 1152
    },
    {
      "completion_length": 24.40625,
      "epoch": 3.5376344086021505,
      "grad_norm": 10.315388679504395,
      "kl": 0.576171875,
      "learning_rate": 1.1384615384615384e-07,
      "loss": 0.1047,
      "reward": 0.7355612218379974,
      "reward_std": 0.04868042771704495,
      "reward_std_all": 0.057839374989271164,
      "rewards-max/forensics_sinkhorn_reward": 0.9690416157245636,
      "rewards-max/forensics_strict_edge_reward": 0.4433192014694214,
      "rewards-max/forensics_tiou_reward": 0.9538534581661224,
      "rewards/forensics_sinkhorn_reward": 0.670965701341629,
      "rewards/forensics_strict_edge_reward": 0.03302445728331804,
      "rewards/forensics_tiou_reward": 0.5193095356225967,
      "step": 1153
    },
    {
      "completion_length": 26.40625,
      "epoch": 3.5407066052227343,
      "grad_norm": 7.93467903137207,
      "kl": 0.466796875,
      "learning_rate": 1.1307692307692307e-07,
      "loss": 0.0128,
      "reward": 0.7857265770435333,
      "reward_std": 0.012880038935691118,
      "reward_std_all": 0.01864841766655445,
      "rewards-max/forensics_sinkhorn_reward": 0.9827115833759308,
      "rewards-max/forensics_strict_edge_reward": 0.6594628691673279,
      "rewards-max/forensics_tiou_reward": 0.9811005592346191,
      "rewards/forensics_sinkhorn_reward": 0.7954412400722504,
      "rewards/forensics_strict_edge_reward": 0.0715695284307003,
      "rewards/forensics_tiou_reward": 0.6330692768096924,
      "step": 1154
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.543778801843318,
      "grad_norm": 8.26378059387207,
      "kl": 0.4619140625,
      "learning_rate": 1.123076923076923e-07,
      "loss": 0.0631,
      "reward": 0.7629994750022888,
      "reward_std": 0.04558283369988203,
      "reward_std_all": 0.04839089885354042,
      "rewards-max/forensics_sinkhorn_reward": 0.9811879396438599,
      "rewards-max/forensics_strict_edge_reward": 0.7047859728336334,
      "rewards-max/forensics_tiou_reward": 0.9842066466808319,
      "rewards/forensics_sinkhorn_reward": 0.7463576793670654,
      "rewards/forensics_strict_edge_reward": 0.07253235206007957,
      "rewards/forensics_tiou_reward": 0.6191520392894745,
      "step": 1155
    },
    {
      "completion_length": 26.9375,
      "epoch": 3.546850998463902,
      "grad_norm": 7.377379417419434,
      "kl": 0.3681640625,
      "learning_rate": 1.1153846153846154e-07,
      "loss": -0.0695,
      "reward": 0.7923241555690765,
      "reward_std": 0.01283451507333666,
      "reward_std_all": 0.014688383555039763,
      "rewards-max/forensics_sinkhorn_reward": 0.9827646911144257,
      "rewards-max/forensics_strict_edge_reward": 0.5521318167448044,
      "rewards-max/forensics_tiou_reward": 0.9841064512729645,
      "rewards/forensics_sinkhorn_reward": 0.8029716312885284,
      "rewards/forensics_strict_edge_reward": 0.07487755827605724,
      "rewards/forensics_tiou_reward": 0.6405388414859772,
      "step": 1156
    },
    {
      "completion_length": 26.140625,
      "epoch": 3.549923195084485,
      "grad_norm": 7.875638008117676,
      "kl": 0.416015625,
      "learning_rate": 1.1076923076923077e-07,
      "loss": 0.033,
      "reward": 0.7735578417778015,
      "reward_std": 0.05449134297668934,
      "reward_std_all": 0.04929692484438419,
      "rewards-max/forensics_sinkhorn_reward": 0.9779609739780426,
      "rewards-max/forensics_strict_edge_reward": 0.541800245642662,
      "rewards-max/forensics_tiou_reward": 0.9743585288524628,
      "rewards/forensics_sinkhorn_reward": 0.7988287806510925,
      "rewards/forensics_strict_edge_reward": 0.057369984686374664,
      "rewards/forensics_tiou_reward": 0.6452093422412872,
      "step": 1157
    },
    {
      "completion_length": 23.890625,
      "epoch": 3.5529953917050694,
      "grad_norm": 8.986897468566895,
      "kl": 0.4794921875,
      "learning_rate": 1.0999999999999999e-07,
      "loss": -0.01,
      "reward": 0.8030929863452911,
      "reward_std": 0.0062729420606046915,
      "reward_std_all": 0.007184746558777988,
      "rewards-max/forensics_sinkhorn_reward": 0.9787042737007141,
      "rewards-max/forensics_strict_edge_reward": 0.49053023755550385,
      "rewards-max/forensics_tiou_reward": 0.9749219417572021,
      "rewards/forensics_sinkhorn_reward": 0.8269094526767731,
      "rewards/forensics_strict_edge_reward": 0.06500883214175701,
      "rewards/forensics_tiou_reward": 0.6845117807388306,
      "step": 1158
    },
    {
      "completion_length": 23.71875,
      "epoch": 3.5560675883256527,
      "grad_norm": 7.225614070892334,
      "kl": 0.4814453125,
      "learning_rate": 1.0923076923076922e-07,
      "loss": -0.006,
      "reward": 0.7907550930976868,
      "reward_std": 0.020314105786383152,
      "reward_std_all": 0.020295372931286693,
      "rewards-max/forensics_sinkhorn_reward": 0.9688283801078796,
      "rewards-max/forensics_strict_edge_reward": 0.45014238357543945,
      "rewards-max/forensics_tiou_reward": 0.951140969991684,
      "rewards/forensics_sinkhorn_reward": 0.802457332611084,
      "rewards/forensics_strict_edge_reward": 0.06932608783245087,
      "rewards/forensics_tiou_reward": 0.6488900780677795,
      "step": 1159
    },
    {
      "completion_length": 26.671875,
      "epoch": 3.5591397849462365,
      "grad_norm": 8.216452598571777,
      "kl": 0.4248046875,
      "learning_rate": 1.0846153846153845e-07,
      "loss": 0.1128,
      "reward": 0.7505985796451569,
      "reward_std": 0.06262477301061153,
      "reward_std_all": 0.06046636961400509,
      "rewards-max/forensics_sinkhorn_reward": 0.9668971002101898,
      "rewards-max/forensics_strict_edge_reward": 0.32717829942703247,
      "rewards-max/forensics_tiou_reward": 0.9476043283939362,
      "rewards/forensics_sinkhorn_reward": 0.7032444477081299,
      "rewards/forensics_strict_edge_reward": 0.043516576290130615,
      "rewards/forensics_tiou_reward": 0.5437141209840775,
      "step": 1160
    },
    {
      "completion_length": 23.3125,
      "epoch": 3.5622119815668203,
      "grad_norm": 9.923026084899902,
      "kl": 0.4755859375,
      "learning_rate": 1.076923076923077e-07,
      "loss": -0.0322,
      "reward": 0.7730721235275269,
      "reward_std": 0.02666368382051587,
      "reward_std_all": 0.03403069917112589,
      "rewards-max/forensics_sinkhorn_reward": 0.9699985682964325,
      "rewards-max/forensics_strict_edge_reward": 0.539385087788105,
      "rewards-max/forensics_tiou_reward": 0.9550627171993256,
      "rewards/forensics_sinkhorn_reward": 0.7656628489494324,
      "rewards/forensics_strict_edge_reward": 0.042562806978821754,
      "rewards/forensics_tiou_reward": 0.5728336274623871,
      "step": 1161
    },
    {
      "completion_length": 22.25,
      "epoch": 3.565284178187404,
      "grad_norm": 14.639093399047852,
      "kl": 0.966796875,
      "learning_rate": 1.0692307692307693e-07,
      "loss": 0.0958,
      "reward": 0.7750932276248932,
      "reward_std": 0.02048817928880453,
      "reward_std_all": 0.02876955270767212,
      "rewards-max/forensics_sinkhorn_reward": 0.9512249827384949,
      "rewards-max/forensics_strict_edge_reward": 0.2569960691034794,
      "rewards-max/forensics_tiou_reward": 0.8936023712158203,
      "rewards/forensics_sinkhorn_reward": 0.7568780779838562,
      "rewards/forensics_strict_edge_reward": 0.030278874561190605,
      "rewards/forensics_tiou_reward": 0.5708813667297363,
      "step": 1162
    },
    {
      "completion_length": 25.921875,
      "epoch": 3.568356374807988,
      "grad_norm": 7.255169868469238,
      "kl": 0.4873046875,
      "learning_rate": 1.0615384615384615e-07,
      "loss": -0.0467,
      "reward": 0.7937400937080383,
      "reward_std": 0.022452383302152157,
      "reward_std_all": 0.02096514869481325,
      "rewards-max/forensics_sinkhorn_reward": 0.9883373379707336,
      "rewards-max/forensics_strict_edge_reward": 0.7989390194416046,
      "rewards-max/forensics_tiou_reward": 0.9977815747261047,
      "rewards/forensics_sinkhorn_reward": 0.8060586154460907,
      "rewards/forensics_strict_edge_reward": 0.09444674849510193,
      "rewards/forensics_tiou_reward": 0.66676265001297,
      "step": 1163
    },
    {
      "completion_length": 26.515625,
      "epoch": 3.571428571428571,
      "grad_norm": 8.997652053833008,
      "kl": 0.52734375,
      "learning_rate": 1.0538461538461538e-07,
      "loss": -0.0464,
      "reward": 0.787063479423523,
      "reward_std": 0.008167112478986382,
      "reward_std_all": 0.014455446740612388,
      "rewards-max/forensics_sinkhorn_reward": 0.9663517773151398,
      "rewards-max/forensics_strict_edge_reward": 0.4834974706172943,
      "rewards-max/forensics_tiou_reward": 0.9369016289710999,
      "rewards/forensics_sinkhorn_reward": 0.8048863112926483,
      "rewards/forensics_strict_edge_reward": 0.07985768653452396,
      "rewards/forensics_tiou_reward": 0.6563601493835449,
      "step": 1164
    },
    {
      "completion_length": 25.109375,
      "epoch": 3.5745007680491554,
      "grad_norm": 10.28048324584961,
      "kl": 0.58984375,
      "learning_rate": 1.0461538461538461e-07,
      "loss": -0.1403,
      "reward": 0.7890298068523407,
      "reward_std": 0.01672494993545115,
      "reward_std_all": 0.018148750765249133,
      "rewards-max/forensics_sinkhorn_reward": 0.979302853345871,
      "rewards-max/forensics_strict_edge_reward": 0.5854259133338928,
      "rewards-max/forensics_tiou_reward": 0.9729959666728973,
      "rewards/forensics_sinkhorn_reward": 0.7824635207653046,
      "rewards/forensics_strict_edge_reward": 0.09415335580706596,
      "rewards/forensics_tiou_reward": 0.6688214838504791,
      "step": 1165
    },
    {
      "completion_length": 24.40625,
      "epoch": 3.5775729646697387,
      "grad_norm": 9.722684860229492,
      "kl": 0.3828125,
      "learning_rate": 1.0384615384615385e-07,
      "loss": -0.0992,
      "reward": 0.7658049464225769,
      "reward_std": 0.003352384897880256,
      "reward_std_all": 0.02374671248253435,
      "rewards-max/forensics_sinkhorn_reward": 0.9837652146816254,
      "rewards-max/forensics_strict_edge_reward": 0.7676310539245605,
      "rewards-max/forensics_tiou_reward": 0.9867816269397736,
      "rewards/forensics_sinkhorn_reward": 0.7752965688705444,
      "rewards/forensics_strict_edge_reward": 0.11677954718470573,
      "rewards/forensics_tiou_reward": 0.6152225136756897,
      "step": 1166
    },
    {
      "completion_length": 29.359375,
      "epoch": 3.5806451612903225,
      "grad_norm": 9.293853759765625,
      "kl": 0.35546875,
      "learning_rate": 1.0307692307692307e-07,
      "loss": -0.0972,
      "reward": 0.7796667516231537,
      "reward_std": 0.015743681229650974,
      "reward_std_all": 0.021551883779466152,
      "rewards-max/forensics_sinkhorn_reward": 0.9873684942722321,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9954085052013397,
      "rewards/forensics_sinkhorn_reward": 0.7786728739738464,
      "rewards/forensics_strict_edge_reward": 0.10831725224852562,
      "rewards/forensics_tiou_reward": 0.7023284137248993,
      "step": 1167
    },
    {
      "completion_length": 28.3125,
      "epoch": 3.5837173579109063,
      "grad_norm": 6.868742942810059,
      "kl": 0.56640625,
      "learning_rate": 1.023076923076923e-07,
      "loss": 0.0246,
      "reward": 0.7874801158905029,
      "reward_std": 0.025694130919873714,
      "reward_std_all": 0.02495502680540085,
      "rewards-max/forensics_sinkhorn_reward": 0.981837123632431,
      "rewards-max/forensics_strict_edge_reward": 0.7043021023273468,
      "rewards-max/forensics_tiou_reward": 0.981065422296524,
      "rewards/forensics_sinkhorn_reward": 0.7588922679424286,
      "rewards/forensics_strict_edge_reward": 0.08393122255802155,
      "rewards/forensics_tiou_reward": 0.6333701908588409,
      "step": 1168
    },
    {
      "completion_length": 26.703125,
      "epoch": 3.58678955453149,
      "grad_norm": 8.48513412475586,
      "kl": 0.4345703125,
      "learning_rate": 1.0153846153846153e-07,
      "loss": -0.0499,
      "reward": 0.7860111892223358,
      "reward_std": 0.018105865397956222,
      "reward_std_all": 0.020165566122159362,
      "rewards-max/forensics_sinkhorn_reward": 0.9841855466365814,
      "rewards-max/forensics_strict_edge_reward": 0.7490177750587463,
      "rewards-max/forensics_tiou_reward": 0.988264799118042,
      "rewards/forensics_sinkhorn_reward": 0.8420317769050598,
      "rewards/forensics_strict_edge_reward": 0.06132511980831623,
      "rewards/forensics_tiou_reward": 0.6968088746070862,
      "step": 1169
    },
    {
      "completion_length": 27.3125,
      "epoch": 3.589861751152074,
      "grad_norm": 7.195263862609863,
      "kl": 0.513671875,
      "learning_rate": 1.0076923076923076e-07,
      "loss": 0.0407,
      "reward": 0.7568749189376831,
      "reward_std": 0.037961967289447784,
      "reward_std_all": 0.04268278554081917,
      "rewards-max/forensics_sinkhorn_reward": 0.9813650548458099,
      "rewards-max/forensics_strict_edge_reward": 0.5275306105613708,
      "rewards-max/forensics_tiou_reward": 0.9833036661148071,
      "rewards/forensics_sinkhorn_reward": 0.7333929240703583,
      "rewards/forensics_strict_edge_reward": 0.06567912735044956,
      "rewards/forensics_tiou_reward": 0.5793061405420303,
      "step": 1170
    },
    {
      "completion_length": 25.640625,
      "epoch": 3.592933947772657,
      "grad_norm": 8.416614532470703,
      "kl": 0.41015625,
      "learning_rate": 1e-07,
      "loss": -0.0626,
      "reward": 0.8054389655590057,
      "reward_std": 0.0024253505980595946,
      "reward_std_all": 0.003975272411480546,
      "rewards-max/forensics_sinkhorn_reward": 0.9821559488773346,
      "rewards-max/forensics_strict_edge_reward": 0.6621623635292053,
      "rewards-max/forensics_tiou_reward": 0.9826016426086426,
      "rewards/forensics_sinkhorn_reward": 0.845622181892395,
      "rewards/forensics_strict_edge_reward": 0.08414825797080994,
      "rewards/forensics_tiou_reward": 0.7179352045059204,
      "step": 1171
    },
    {
      "completion_length": 28.4375,
      "epoch": 3.5960061443932414,
      "grad_norm": 6.293784141540527,
      "kl": 0.517578125,
      "learning_rate": 9.923076923076922e-08,
      "loss": -0.0163,
      "reward": 0.7898548543453217,
      "reward_std": 0.028494223952293396,
      "reward_std_all": 0.026522335596382618,
      "rewards-max/forensics_sinkhorn_reward": 0.9838781952857971,
      "rewards-max/forensics_strict_edge_reward": 0.4472678154706955,
      "rewards-max/forensics_tiou_reward": 0.9879595041275024,
      "rewards/forensics_sinkhorn_reward": 0.8086727559566498,
      "rewards/forensics_strict_edge_reward": 0.06607596203684807,
      "rewards/forensics_tiou_reward": 0.6566649973392487,
      "step": 1172
    },
    {
      "completion_length": 26.40625,
      "epoch": 3.5990783410138247,
      "grad_norm": 7.407046794891357,
      "kl": 0.5166015625,
      "learning_rate": 9.846153846153846e-08,
      "loss": -0.0161,
      "reward": 0.7746462523937225,
      "reward_std": 0.020683358423411846,
      "reward_std_all": 0.029210141859948635,
      "rewards-max/forensics_sinkhorn_reward": 0.966616302728653,
      "rewards-max/forensics_strict_edge_reward": 0.7000000029802322,
      "rewards-max/forensics_tiou_reward": 0.9454410970211029,
      "rewards/forensics_sinkhorn_reward": 0.7671089768409729,
      "rewards/forensics_strict_edge_reward": 0.0775494184345007,
      "rewards/forensics_tiou_reward": 0.564082533121109,
      "step": 1173
    },
    {
      "completion_length": 26.625,
      "epoch": 3.6021505376344085,
      "grad_norm": 9.055042266845703,
      "kl": 0.412109375,
      "learning_rate": 9.769230769230769e-08,
      "loss": -0.022,
      "reward": 0.7827673554420471,
      "reward_std": 0.012045623734593391,
      "reward_std_all": 0.01872314792126417,
      "rewards-max/forensics_sinkhorn_reward": 0.9822142720222473,
      "rewards-max/forensics_strict_edge_reward": 0.46017518639564514,
      "rewards-max/forensics_tiou_reward": 0.9794923961162567,
      "rewards/forensics_sinkhorn_reward": 0.7377933263778687,
      "rewards/forensics_strict_edge_reward": 0.06894098967313766,
      "rewards/forensics_tiou_reward": 0.6286898553371429,
      "step": 1174
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.6052227342549923,
      "grad_norm": 7.865375995635986,
      "kl": 0.3662109375,
      "learning_rate": 9.692307692307692e-08,
      "loss": -0.0527,
      "reward": 0.8050509095191956,
      "reward_std": 0.0026617232942953706,
      "reward_std_all": 0.004380014608614147,
      "rewards-max/forensics_sinkhorn_reward": 0.9857923686504364,
      "rewards-max/forensics_strict_edge_reward": 0.7012533247470856,
      "rewards-max/forensics_tiou_reward": 0.9927740097045898,
      "rewards/forensics_sinkhorn_reward": 0.8467175960540771,
      "rewards/forensics_strict_edge_reward": 0.07379639707505703,
      "rewards/forensics_tiou_reward": 0.7261864840984344,
      "step": 1175
    },
    {
      "completion_length": 23.78125,
      "epoch": 3.608294930875576,
      "grad_norm": 9.465533256530762,
      "kl": 0.37890625,
      "learning_rate": 9.615384615384616e-08,
      "loss": -0.0773,
      "reward": 0.8049156963825226,
      "reward_std": 0.002969856490381062,
      "reward_std_all": 0.004556668805889785,
      "rewards-max/forensics_sinkhorn_reward": 0.9817058145999908,
      "rewards-max/forensics_strict_edge_reward": 0.5969027578830719,
      "rewards-max/forensics_tiou_reward": 0.9828505516052246,
      "rewards/forensics_sinkhorn_reward": 0.8351246416568756,
      "rewards/forensics_strict_edge_reward": 0.08496807515621185,
      "rewards/forensics_tiou_reward": 0.7101875841617584,
      "step": 1176
    },
    {
      "completion_length": 21.40625,
      "epoch": 3.61136712749616,
      "grad_norm": 7.538238048553467,
      "kl": 0.4599609375,
      "learning_rate": 9.538461538461538e-08,
      "loss": -0.0416,
      "reward": 0.7930909395217896,
      "reward_std": 0.019299494568258524,
      "reward_std_all": 0.0190709694288671,
      "rewards-max/forensics_sinkhorn_reward": 0.9868243336677551,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9939998686313629,
      "rewards/forensics_sinkhorn_reward": 0.8122131526470184,
      "rewards/forensics_strict_edge_reward": 0.10305413976311684,
      "rewards/forensics_tiou_reward": 0.6866848766803741,
      "step": 1177
    },
    {
      "completion_length": 30.046875,
      "epoch": 3.6144393241167436,
      "grad_norm": 6.761477947235107,
      "kl": 0.37109375,
      "learning_rate": 9.461538461538461e-08,
      "loss": -0.1149,
      "reward": 0.8057733774185181,
      "reward_std": 0.002553127007558942,
      "reward_std_all": 0.003744568210095167,
      "rewards-max/forensics_sinkhorn_reward": 0.9833424389362335,
      "rewards-max/forensics_strict_edge_reward": 0.5736550390720367,
      "rewards-max/forensics_tiou_reward": 0.9848341643810272,
      "rewards/forensics_sinkhorn_reward": 0.8374998271465302,
      "rewards/forensics_strict_edge_reward": 0.1158176027238369,
      "rewards/forensics_tiou_reward": 0.7380496859550476,
      "step": 1178
    },
    {
      "completion_length": 25.09375,
      "epoch": 3.6175115207373274,
      "grad_norm": 9.731247901916504,
      "kl": 0.521484375,
      "learning_rate": 9.384615384615384e-08,
      "loss": 0.0509,
      "reward": 0.791938990354538,
      "reward_std": 0.02240242762491107,
      "reward_std_all": 0.021435112226754427,
      "rewards-max/forensics_sinkhorn_reward": 0.9817413091659546,
      "rewards-max/forensics_strict_edge_reward": 0.7268245220184326,
      "rewards-max/forensics_tiou_reward": 0.9805825352668762,
      "rewards/forensics_sinkhorn_reward": 0.8027375042438507,
      "rewards/forensics_strict_edge_reward": 0.0700303241610527,
      "rewards/forensics_tiou_reward": 0.6243840754032135,
      "step": 1179
    },
    {
      "completion_length": 23.203125,
      "epoch": 3.6205837173579107,
      "grad_norm": 8.3106107711792,
      "kl": 0.396484375,
      "learning_rate": 9.307692307692306e-08,
      "loss": 0.0261,
      "reward": 0.7738874554634094,
      "reward_std": 0.02139190584421158,
      "reward_std_all": 0.027624546317383647,
      "rewards-max/forensics_sinkhorn_reward": 0.9826933145523071,
      "rewards-max/forensics_strict_edge_reward": 0.7533732652664185,
      "rewards-max/forensics_tiou_reward": 0.9822736382484436,
      "rewards/forensics_sinkhorn_reward": 0.7986709475517273,
      "rewards/forensics_strict_edge_reward": 0.09137685224413872,
      "rewards/forensics_tiou_reward": 0.6449896991252899,
      "step": 1180
    },
    {
      "completion_length": 27.21875,
      "epoch": 3.6236559139784945,
      "grad_norm": 6.005672454833984,
      "kl": 0.580078125,
      "learning_rate": 9.23076923076923e-08,
      "loss": 0.0857,
      "reward": 0.7580105066299438,
      "reward_std": 0.03051166981458664,
      "reward_std_all": 0.04037475027143955,
      "rewards-max/forensics_sinkhorn_reward": 0.9745435118675232,
      "rewards-max/forensics_strict_edge_reward": 0.39278973639011383,
      "rewards-max/forensics_tiou_reward": 0.960253119468689,
      "rewards/forensics_sinkhorn_reward": 0.7489725351333618,
      "rewards/forensics_strict_edge_reward": 0.05136045999825001,
      "rewards/forensics_tiou_reward": 0.6123754382133484,
      "step": 1181
    },
    {
      "completion_length": 24.8125,
      "epoch": 3.6267281105990783,
      "grad_norm": 7.795929431915283,
      "kl": 0.4384765625,
      "learning_rate": 9.153846153846153e-08,
      "loss": -0.0426,
      "reward": 0.7863581478595734,
      "reward_std": 0.018571078777313232,
      "reward_std_all": 0.0203481437638402,
      "rewards-max/forensics_sinkhorn_reward": 0.9755503833293915,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9692121148109436,
      "rewards/forensics_sinkhorn_reward": 0.8181019425392151,
      "rewards/forensics_strict_edge_reward": 0.07471423409879208,
      "rewards/forensics_tiou_reward": 0.6598846614360809,
      "step": 1182
    },
    {
      "completion_length": 23.484375,
      "epoch": 3.629800307219662,
      "grad_norm": 8.195841789245605,
      "kl": 0.470703125,
      "learning_rate": 9.076923076923076e-08,
      "loss": -0.013,
      "reward": 0.7858696579933167,
      "reward_std": 0.0227413522079587,
      "reward_std_all": 0.023276702500879765,
      "rewards-max/forensics_sinkhorn_reward": 0.9842991232872009,
      "rewards-max/forensics_strict_edge_reward": 0.7181241810321808,
      "rewards-max/forensics_tiou_reward": 0.9875116944313049,
      "rewards/forensics_sinkhorn_reward": 0.804748147726059,
      "rewards/forensics_strict_edge_reward": 0.08577391877770424,
      "rewards/forensics_tiou_reward": 0.6448196768760681,
      "step": 1183
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.632872503840246,
      "grad_norm": 9.661170959472656,
      "kl": 0.4697265625,
      "learning_rate": 9e-08,
      "loss": 0.0492,
      "reward": 0.7529606521129608,
      "reward_std": 0.07100199535489082,
      "reward_std_all": 0.0660274289548397,
      "rewards-max/forensics_sinkhorn_reward": 0.9830861389636993,
      "rewards-max/forensics_strict_edge_reward": 0.6231098771095276,
      "rewards-max/forensics_tiou_reward": 0.9833919107913971,
      "rewards/forensics_sinkhorn_reward": 0.7600789070129395,
      "rewards/forensics_strict_edge_reward": 0.061044398695230484,
      "rewards/forensics_tiou_reward": 0.6391517221927643,
      "step": 1184
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.6359447004608296,
      "grad_norm": 9.216626167297363,
      "kl": 0.65625,
      "learning_rate": 8.923076923076923e-08,
      "loss": 0.0439,
      "reward": 0.7814225554466248,
      "reward_std": 0.03171005006879568,
      "reward_std_all": 0.030662880279123783,
      "rewards-max/forensics_sinkhorn_reward": 0.9795728027820587,
      "rewards-max/forensics_strict_edge_reward": 0.48032720386981964,
      "rewards-max/forensics_tiou_reward": 0.9741788506507874,
      "rewards/forensics_sinkhorn_reward": 0.7314567863941193,
      "rewards/forensics_strict_edge_reward": 0.04134150501340628,
      "rewards/forensics_tiou_reward": 0.626977413892746,
      "step": 1185
    },
    {
      "completion_length": 27.03125,
      "epoch": 3.6390168970814134,
      "grad_norm": 7.834023475646973,
      "kl": 0.44140625,
      "learning_rate": 8.846153846153847e-08,
      "loss": -0.0177,
      "reward": 0.8041785061359406,
      "reward_std": 0.004231927392538637,
      "reward_std_all": 0.00558156450279057,
      "rewards-max/forensics_sinkhorn_reward": 0.9822400510311127,
      "rewards-max/forensics_strict_edge_reward": 0.6551513969898224,
      "rewards-max/forensics_tiou_reward": 0.9800547659397125,
      "rewards/forensics_sinkhorn_reward": 0.8370473980903625,
      "rewards/forensics_strict_edge_reward": 0.08480677753686905,
      "rewards/forensics_tiou_reward": 0.7026365697383881,
      "step": 1186
    },
    {
      "completion_length": 24.328125,
      "epoch": 3.6420890937019967,
      "grad_norm": 8.421483993530273,
      "kl": 0.41796875,
      "learning_rate": 8.769230769230769e-08,
      "loss": -0.0361,
      "reward": 0.7942984104156494,
      "reward_std": 0.021304602501913905,
      "reward_std_all": 0.02036071359179914,
      "rewards-max/forensics_sinkhorn_reward": 0.9687622785568237,
      "rewards-max/forensics_strict_edge_reward": 0.4206818640232086,
      "rewards-max/forensics_tiou_reward": 0.9459215998649597,
      "rewards/forensics_sinkhorn_reward": 0.8328590393066406,
      "rewards/forensics_strict_edge_reward": 0.08121813088655472,
      "rewards/forensics_tiou_reward": 0.6657547056674957,
      "step": 1187
    },
    {
      "completion_length": 22.03125,
      "epoch": 3.6451612903225805,
      "grad_norm": 8.435989379882812,
      "kl": 0.4140625,
      "learning_rate": 8.692307692307692e-08,
      "loss": -0.0455,
      "reward": 0.7934973537921906,
      "reward_std": 0.009303807455580682,
      "reward_std_all": 0.012213897658511996,
      "rewards-max/forensics_sinkhorn_reward": 0.9809754192829132,
      "rewards-max/forensics_strict_edge_reward": 0.8120969831943512,
      "rewards-max/forensics_tiou_reward": 0.9823476672172546,
      "rewards/forensics_sinkhorn_reward": 0.8445175886154175,
      "rewards/forensics_strict_edge_reward": 0.10729024559259415,
      "rewards/forensics_tiou_reward": 0.6812952160835266,
      "step": 1188
    },
    {
      "completion_length": 22.734375,
      "epoch": 3.6482334869431643,
      "grad_norm": 8.182562828063965,
      "kl": 0.625,
      "learning_rate": 8.615384615384615e-08,
      "loss": 0.0448,
      "reward": 0.7789419889450073,
      "reward_std": 0.04090243112295866,
      "reward_std_all": 0.038113128393888474,
      "rewards-max/forensics_sinkhorn_reward": 0.9861496686935425,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9929197728633881,
      "rewards/forensics_sinkhorn_reward": 0.7493255138397217,
      "rewards/forensics_strict_edge_reward": 0.10508308559656143,
      "rewards/forensics_tiou_reward": 0.5996353626251221,
      "step": 1189
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.651305683563748,
      "grad_norm": 8.420660972595215,
      "kl": 0.5439453125,
      "learning_rate": 8.538461538461537e-08,
      "loss": 0.0158,
      "reward": 0.8014068603515625,
      "reward_std": 0.009220471139997244,
      "reward_std_all": 0.009534788317978382,
      "rewards-max/forensics_sinkhorn_reward": 0.9770821928977966,
      "rewards-max/forensics_strict_edge_reward": 0.6531337201595306,
      "rewards-max/forensics_tiou_reward": 0.9724378287792206,
      "rewards/forensics_sinkhorn_reward": 0.8172462582588196,
      "rewards/forensics_strict_edge_reward": 0.09483916312456131,
      "rewards/forensics_tiou_reward": 0.703386127948761,
      "step": 1190
    },
    {
      "completion_length": 26.453125,
      "epoch": 3.654377880184332,
      "grad_norm": 8.356512069702148,
      "kl": 0.3994140625,
      "learning_rate": 8.461538461538461e-08,
      "loss": -0.0224,
      "reward": 0.7751553356647491,
      "reward_std": 0.054260218515992165,
      "reward_std_all": 0.04860064387321472,
      "rewards-max/forensics_sinkhorn_reward": 0.9823625385761261,
      "rewards-max/forensics_strict_edge_reward": 0.5002432614564896,
      "rewards-max/forensics_tiou_reward": 0.980565220117569,
      "rewards/forensics_sinkhorn_reward": 0.7914498448371887,
      "rewards/forensics_strict_edge_reward": 0.062465205788612366,
      "rewards/forensics_tiou_reward": 0.660107433795929,
      "step": 1191
    },
    {
      "completion_length": 26.703125,
      "epoch": 3.6574500768049156,
      "grad_norm": 21.14797592163086,
      "kl": 0.9140625,
      "learning_rate": 8.384615384615384e-08,
      "loss": -0.0489,
      "reward": 0.7974308431148529,
      "reward_std": 0.015576406149193645,
      "reward_std_all": 0.015320304781198502,
      "rewards-max/forensics_sinkhorn_reward": 0.9774138927459717,
      "rewards-max/forensics_strict_edge_reward": 0.5932122468948364,
      "rewards-max/forensics_tiou_reward": 0.9704385697841644,
      "rewards/forensics_sinkhorn_reward": 0.8160403370857239,
      "rewards/forensics_strict_edge_reward": 0.09007482975721359,
      "rewards/forensics_tiou_reward": 0.6908567547798157,
      "step": 1192
    },
    {
      "completion_length": 25.171875,
      "epoch": 3.6605222734254994,
      "grad_norm": 8.104427337646484,
      "kl": 0.4482421875,
      "learning_rate": 8.307692307692307e-08,
      "loss": 0.0424,
      "reward": 0.7758110165596008,
      "reward_std": 0.04726473428308964,
      "reward_std_all": 0.04339562635868788,
      "rewards-max/forensics_sinkhorn_reward": 0.9820889532566071,
      "rewards-max/forensics_strict_edge_reward": 0.5135407596826553,
      "rewards-max/forensics_tiou_reward": 0.9818005561828613,
      "rewards/forensics_sinkhorn_reward": 0.7862381339073181,
      "rewards/forensics_strict_edge_reward": 0.06023945473134518,
      "rewards/forensics_tiou_reward": 0.6482980847358704,
      "step": 1193
    },
    {
      "completion_length": 30.03125,
      "epoch": 3.6635944700460827,
      "grad_norm": 7.2338151931762695,
      "kl": 0.4365234375,
      "learning_rate": 8.23076923076923e-08,
      "loss": -0.0177,
      "reward": 0.802205502986908,
      "reward_std": 0.0063731917180120945,
      "reward_std_all": 0.007738159503787756,
      "rewards-max/forensics_sinkhorn_reward": 0.9874676167964935,
      "rewards-max/forensics_strict_edge_reward": 0.471409872174263,
      "rewards-max/forensics_tiou_reward": 0.995041012763977,
      "rewards/forensics_sinkhorn_reward": 0.8410476744174957,
      "rewards/forensics_strict_edge_reward": 0.04032129142433405,
      "rewards/forensics_tiou_reward": 0.7262412309646606,
      "step": 1194
    },
    {
      "completion_length": 25.828125,
      "epoch": 3.6666666666666665,
      "grad_norm": 7.6433844566345215,
      "kl": 0.4130859375,
      "learning_rate": 8.153846153846153e-08,
      "loss": -0.0549,
      "reward": 0.8011377453804016,
      "reward_std": 0.010117119643837214,
      "reward_std_all": 0.010406631045043468,
      "rewards-max/forensics_sinkhorn_reward": 0.9684984683990479,
      "rewards-max/forensics_strict_edge_reward": 0.6394134163856506,
      "rewards-max/forensics_tiou_reward": 0.9551628828048706,
      "rewards/forensics_sinkhorn_reward": 0.8352031409740448,
      "rewards/forensics_strict_edge_reward": 0.11960256658494473,
      "rewards/forensics_tiou_reward": 0.7109109163284302,
      "step": 1195
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.6697388632872503,
      "grad_norm": 21.188461303710938,
      "kl": 0.3818359375,
      "learning_rate": 8.076923076923077e-08,
      "loss": 0.0851,
      "reward": 0.7693406939506531,
      "reward_std": 0.03197546675801277,
      "reward_std_all": 0.036588579416275024,
      "rewards-max/forensics_sinkhorn_reward": 0.9329032897949219,
      "rewards-max/forensics_strict_edge_reward": 0.75,
      "rewards-max/forensics_tiou_reward": 0.8746653199195862,
      "rewards/forensics_sinkhorn_reward": 0.715070366859436,
      "rewards/forensics_strict_edge_reward": 0.06625980138778687,
      "rewards/forensics_tiou_reward": 0.5283058285713196,
      "step": 1196
    },
    {
      "completion_length": 24.625,
      "epoch": 3.672811059907834,
      "grad_norm": 9.06967830657959,
      "kl": 0.353515625,
      "learning_rate": 8e-08,
      "loss": -0.0409,
      "reward": 0.7803235650062561,
      "reward_std": 0.03771994076669216,
      "reward_std_all": 0.03564323019236326,
      "rewards-max/forensics_sinkhorn_reward": 0.977954238653183,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9746575355529785,
      "rewards/forensics_sinkhorn_reward": 0.7873087525367737,
      "rewards/forensics_strict_edge_reward": 0.0687226839363575,
      "rewards/forensics_tiou_reward": 0.6216574609279633,
      "step": 1197
    },
    {
      "completion_length": 24.359375,
      "epoch": 3.675883256528418,
      "grad_norm": 7.636488914489746,
      "kl": 0.423828125,
      "learning_rate": 7.923076923076923e-08,
      "loss": 0.0408,
      "reward": 0.7907366752624512,
      "reward_std": 0.024408234748989344,
      "reward_std_all": 0.0239215730689466,
      "rewards-max/forensics_sinkhorn_reward": 0.9816509485244751,
      "rewards-max/forensics_strict_edge_reward": 0.6157284677028656,
      "rewards-max/forensics_tiou_reward": 0.9797018766403198,
      "rewards/forensics_sinkhorn_reward": 0.8095057904720306,
      "rewards/forensics_strict_edge_reward": 0.08696220070123672,
      "rewards/forensics_tiou_reward": 0.648937463760376,
      "step": 1198
    },
    {
      "completion_length": 23.390625,
      "epoch": 3.6789554531490016,
      "grad_norm": 9.81421184539795,
      "kl": 0.4931640625,
      "learning_rate": 7.846153846153846e-08,
      "loss": -0.0639,
      "reward": 0.7909279763698578,
      "reward_std": 0.016593517269939184,
      "reward_std_all": 0.017358053009957075,
      "rewards-max/forensics_sinkhorn_reward": 0.9855585098266602,
      "rewards-max/forensics_strict_edge_reward": 0.8079957664012909,
      "rewards-max/forensics_tiou_reward": 0.9917324781417847,
      "rewards/forensics_sinkhorn_reward": 0.822636604309082,
      "rewards/forensics_strict_edge_reward": 0.1332228183746338,
      "rewards/forensics_tiou_reward": 0.7152347266674042,
      "step": 1199
    },
    {
      "completion_length": 23.015625,
      "epoch": 3.6820276497695854,
      "grad_norm": 11.633851051330566,
      "kl": 0.9140625,
      "learning_rate": 7.769230769230768e-08,
      "loss": -0.0063,
      "reward": 0.7360740602016449,
      "reward_std": 0.0327329789288342,
      "reward_std_all": 0.051319003105163574,
      "rewards-max/forensics_sinkhorn_reward": 0.9785649478435516,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9749135076999664,
      "rewards/forensics_sinkhorn_reward": 0.743215024471283,
      "rewards/forensics_strict_edge_reward": 0.08429044112563133,
      "rewards/forensics_tiou_reward": 0.5941123068332672,
      "step": 1200
    },
    {
      "completion_length": 20.921875,
      "epoch": 3.6850998463901687,
      "grad_norm": 9.606436729431152,
      "kl": 0.4599609375,
      "learning_rate": 7.692307692307692e-08,
      "loss": 0.0053,
      "reward": 0.7850929498672485,
      "reward_std": 0.011814715806394815,
      "reward_std_all": 0.018339091911911964,
      "rewards-max/forensics_sinkhorn_reward": 0.9855708479881287,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.9934206008911133,
      "rewards/forensics_sinkhorn_reward": 0.8205370604991913,
      "rewards/forensics_strict_edge_reward": 0.09616516903042793,
      "rewards/forensics_tiou_reward": 0.6619389057159424,
      "step": 1201
    },
    {
      "completion_length": 25.796875,
      "epoch": 3.688172043010753,
      "grad_norm": 9.24333667755127,
      "kl": 0.3818359375,
      "learning_rate": 7.615384615384615e-08,
      "loss": 0.0726,
      "reward": 0.7544882893562317,
      "reward_std": 0.05663428269326687,
      "reward_std_all": 0.055620115250349045,
      "rewards-max/forensics_sinkhorn_reward": 0.9779886901378632,
      "rewards-max/forensics_strict_edge_reward": 0.5303425639867783,
      "rewards-max/forensics_tiou_reward": 0.9635987281799316,
      "rewards/forensics_sinkhorn_reward": 0.7085601091384888,
      "rewards/forensics_strict_edge_reward": 0.06627602688968182,
      "rewards/forensics_tiou_reward": 0.5055705308914185,
      "step": 1202
    },
    {
      "completion_length": 25.359375,
      "epoch": 3.6912442396313363,
      "grad_norm": 8.896133422851562,
      "kl": 0.3466796875,
      "learning_rate": 7.538461538461538e-08,
      "loss": -0.0554,
      "reward": 0.7756425142288208,
      "reward_std": 0.006393568590283394,
      "reward_std_all": 0.020495933014899492,
      "rewards-max/forensics_sinkhorn_reward": 0.9789203405380249,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9730998277664185,
      "rewards/forensics_sinkhorn_reward": 0.7540467381477356,
      "rewards/forensics_strict_edge_reward": 0.08632930740714073,
      "rewards/forensics_tiou_reward": 0.5812602043151855,
      "step": 1203
    },
    {
      "completion_length": 23.078125,
      "epoch": 3.69431643625192,
      "grad_norm": 8.453564643859863,
      "kl": 0.34375,
      "learning_rate": 7.46153846153846e-08,
      "loss": -0.0658,
      "reward": 0.792881041765213,
      "reward_std": 0.023179658805020154,
      "reward_std_all": 0.02213911316357553,
      "rewards-max/forensics_sinkhorn_reward": 0.9814424216747284,
      "rewards-max/forensics_strict_edge_reward": 0.6697181463241577,
      "rewards-max/forensics_tiou_reward": 0.9800978302955627,
      "rewards/forensics_sinkhorn_reward": 0.8084973692893982,
      "rewards/forensics_strict_edge_reward": 0.0813923180103302,
      "rewards/forensics_tiou_reward": 0.6555664241313934,
      "step": 1204
    },
    {
      "completion_length": 25.34375,
      "epoch": 3.697388632872504,
      "grad_norm": 8.982531547546387,
      "kl": 0.41015625,
      "learning_rate": 7.384615384615384e-08,
      "loss": -0.028,
      "reward": 0.7536602914333344,
      "reward_std": 0.023333157412707806,
      "reward_std_all": 0.03695312514901161,
      "rewards-max/forensics_sinkhorn_reward": 0.987866222858429,
      "rewards-max/forensics_strict_edge_reward": 0.6620611250400543,
      "rewards-max/forensics_tiou_reward": 0.9970012307167053,
      "rewards/forensics_sinkhorn_reward": 0.7963211238384247,
      "rewards/forensics_strict_edge_reward": 0.13537534326314926,
      "rewards/forensics_tiou_reward": 0.6728427410125732,
      "step": 1205
    },
    {
      "completion_length": 26.421875,
      "epoch": 3.7004608294930876,
      "grad_norm": 8.188956260681152,
      "kl": 0.3798828125,
      "learning_rate": 7.307692307692308e-08,
      "loss": 0.0664,
      "reward": 0.7717198729515076,
      "reward_std": 0.04058617167174816,
      "reward_std_all": 0.03968401253223419,
      "rewards-max/forensics_sinkhorn_reward": 0.9847475290298462,
      "rewards-max/forensics_strict_edge_reward": 0.7501528263092041,
      "rewards-max/forensics_tiou_reward": 0.985460638999939,
      "rewards/forensics_sinkhorn_reward": 0.7791836559772491,
      "rewards/forensics_strict_edge_reward": 0.08089381083846092,
      "rewards/forensics_tiou_reward": 0.6469810009002686,
      "step": 1206
    },
    {
      "completion_length": 21.375,
      "epoch": 3.7035330261136714,
      "grad_norm": 10.264449119567871,
      "kl": 0.439453125,
      "learning_rate": 7.230769230769231e-08,
      "loss": -0.054,
      "reward": 0.7919376492500305,
      "reward_std": 0.02614351361989975,
      "reward_std_all": 0.024233716540038586,
      "rewards-max/forensics_sinkhorn_reward": 0.9801077246665955,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9777000248432159,
      "rewards/forensics_sinkhorn_reward": 0.8435719907283783,
      "rewards/forensics_strict_edge_reward": 0.10997799038887024,
      "rewards/forensics_tiou_reward": 0.699436366558075,
      "step": 1207
    },
    {
      "completion_length": 25.390625,
      "epoch": 3.7066052227342547,
      "grad_norm": 7.077463150024414,
      "kl": 0.544921875,
      "learning_rate": 7.153846153846154e-08,
      "loss": 0.0191,
      "reward": 0.7790374159812927,
      "reward_std": 0.02413320471532643,
      "reward_std_all": 0.02799722063355148,
      "rewards-max/forensics_sinkhorn_reward": 0.9861210882663727,
      "rewards-max/forensics_strict_edge_reward": 0.5840405821800232,
      "rewards-max/forensics_tiou_reward": 0.9907706081867218,
      "rewards/forensics_sinkhorn_reward": 0.7841395735740662,
      "rewards/forensics_strict_edge_reward": 0.11199627444148064,
      "rewards/forensics_tiou_reward": 0.6506784856319427,
      "step": 1208
    },
    {
      "completion_length": 25.84375,
      "epoch": 3.709677419354839,
      "grad_norm": 7.597142219543457,
      "kl": 0.5537109375,
      "learning_rate": 7.076923076923076e-08,
      "loss": 0.1255,
      "reward": 0.7543314099311829,
      "reward_std": 0.07900156825780869,
      "reward_std_all": 0.07156091369688511,
      "rewards-max/forensics_sinkhorn_reward": 0.9714392125606537,
      "rewards-max/forensics_strict_edge_reward": 0.38935355097055435,
      "rewards-max/forensics_tiou_reward": 0.9547728002071381,
      "rewards/forensics_sinkhorn_reward": 0.6739112138748169,
      "rewards/forensics_strict_edge_reward": 0.04020506516098976,
      "rewards/forensics_tiou_reward": 0.4987190216779709,
      "step": 1209
    },
    {
      "completion_length": 27.484375,
      "epoch": 3.7127496159754223,
      "grad_norm": 9.865917205810547,
      "kl": 0.4931640625,
      "learning_rate": 7e-08,
      "loss": -0.0663,
      "reward": 0.7937293946743011,
      "reward_std": 0.015797855332493782,
      "reward_std_all": 0.016337382607162,
      "rewards-max/forensics_sinkhorn_reward": 0.9876543581485748,
      "rewards-max/forensics_strict_edge_reward": 0.8335314393043518,
      "rewards-max/forensics_tiou_reward": 0.9960836470127106,
      "rewards/forensics_sinkhorn_reward": 0.8325551152229309,
      "rewards/forensics_strict_edge_reward": 0.126944150775671,
      "rewards/forensics_tiou_reward": 0.7422401607036591,
      "step": 1210
    },
    {
      "completion_length": 22.90625,
      "epoch": 3.715821812596006,
      "grad_norm": 7.906296730041504,
      "kl": 0.4326171875,
      "learning_rate": 6.923076923076923e-08,
      "loss": 0.0101,
      "reward": 0.7887361347675323,
      "reward_std": 0.02850592415779829,
      "reward_std_all": 0.02654343470931053,
      "rewards-max/forensics_sinkhorn_reward": 0.9829533398151398,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.9876768290996552,
      "rewards/forensics_sinkhorn_reward": 0.7862852811813354,
      "rewards/forensics_strict_edge_reward": 0.08420152589678764,
      "rewards/forensics_tiou_reward": 0.6058691740036011,
      "step": 1211
    },
    {
      "completion_length": 23.359375,
      "epoch": 3.71889400921659,
      "grad_norm": 7.737532615661621,
      "kl": 0.40625,
      "learning_rate": 6.846153846153846e-08,
      "loss": -0.0947,
      "reward": 0.7982287704944611,
      "reward_std": 0.013858611695468426,
      "reward_std_all": 0.013723718002438545,
      "rewards-max/forensics_sinkhorn_reward": 0.9752973914146423,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9657649993896484,
      "rewards/forensics_sinkhorn_reward": 0.8100211024284363,
      "rewards/forensics_strict_edge_reward": 0.08790020644664764,
      "rewards/forensics_tiou_reward": 0.6367762684822083,
      "step": 1212
    },
    {
      "completion_length": 23.875,
      "epoch": 3.7219662058371736,
      "grad_norm": 8.73564624786377,
      "kl": 0.4130859375,
      "learning_rate": 6.769230769230769e-08,
      "loss": -0.0411,
      "reward": 0.759592592716217,
      "reward_std": 0.006700359284877777,
      "reward_std_all": 0.027436834760010242,
      "rewards-max/forensics_sinkhorn_reward": 0.9877872467041016,
      "rewards-max/forensics_strict_edge_reward": 0.8465203046798706,
      "rewards-max/forensics_tiou_reward": 0.9983125329017639,
      "rewards/forensics_sinkhorn_reward": 0.7597673535346985,
      "rewards/forensics_strict_edge_reward": 0.12268870696425438,
      "rewards/forensics_tiou_reward": 0.5814027786254883,
      "step": 1213
    },
    {
      "completion_length": 28.265625,
      "epoch": 3.7250384024577574,
      "grad_norm": 7.310576438903809,
      "kl": 0.423828125,
      "learning_rate": 6.692307692307691e-08,
      "loss": -0.0426,
      "reward": 0.7570168375968933,
      "reward_std": 0.023156628478318453,
      "reward_std_all": 0.039026989601552486,
      "rewards-max/forensics_sinkhorn_reward": 0.9783646166324615,
      "rewards-max/forensics_strict_edge_reward": 0.4745236784219742,
      "rewards-max/forensics_tiou_reward": 0.9715593755245209,
      "rewards/forensics_sinkhorn_reward": 0.77240389585495,
      "rewards/forensics_strict_edge_reward": 0.08342895656824112,
      "rewards/forensics_tiou_reward": 0.6582018435001373,
      "step": 1214
    },
    {
      "completion_length": 20.390625,
      "epoch": 3.7281105990783407,
      "grad_norm": 9.202960968017578,
      "kl": 0.4580078125,
      "learning_rate": 6.615384615384615e-08,
      "loss": 0.0412,
      "reward": 0.7956935465335846,
      "reward_std": 0.01236990699544549,
      "reward_std_all": 0.013444395153783262,
      "rewards-max/forensics_sinkhorn_reward": 0.9796096980571747,
      "rewards-max/forensics_strict_edge_reward": 0.4490247070789337,
      "rewards-max/forensics_tiou_reward": 0.9663205444812775,
      "rewards/forensics_sinkhorn_reward": 0.812707394361496,
      "rewards/forensics_strict_edge_reward": 0.07339128479361534,
      "rewards/forensics_tiou_reward": 0.6422945857048035,
      "step": 1215
    },
    {
      "completion_length": 29.921875,
      "epoch": 3.731182795698925,
      "grad_norm": 7.459553241729736,
      "kl": 0.390625,
      "learning_rate": 6.538461538461538e-08,
      "loss": 0.0047,
      "reward": 0.7909011840820312,
      "reward_std": 0.022811679169535637,
      "reward_std_all": 0.022041870281100273,
      "rewards-max/forensics_sinkhorn_reward": 0.9863681197166443,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9917176365852356,
      "rewards/forensics_sinkhorn_reward": 0.8021213412284851,
      "rewards/forensics_strict_edge_reward": 0.1291266717016697,
      "rewards/forensics_tiou_reward": 0.7313831746578217,
      "step": 1216
    },
    {
      "completion_length": 27.875,
      "epoch": 3.7342549923195083,
      "grad_norm": 7.0145955085754395,
      "kl": 0.4091796875,
      "learning_rate": 6.461538461538462e-08,
      "loss": -0.1493,
      "reward": 0.8053338527679443,
      "reward_std": 0.003508379391860217,
      "reward_std_all": 0.004770425031892955,
      "rewards-max/forensics_sinkhorn_reward": 0.9821831285953522,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9831063747406006,
      "rewards/forensics_sinkhorn_reward": 0.8814628422260284,
      "rewards/forensics_strict_edge_reward": 0.10594190284609795,
      "rewards/forensics_tiou_reward": 0.7667186260223389,
      "step": 1217
    },
    {
      "completion_length": 21.40625,
      "epoch": 3.737327188940092,
      "grad_norm": 8.324495315551758,
      "kl": 0.38671875,
      "learning_rate": 6.384615384615385e-08,
      "loss": 0.0074,
      "reward": 0.7913691997528076,
      "reward_std": 0.01673134695738554,
      "reward_std_all": 0.017627394758164883,
      "rewards-max/forensics_sinkhorn_reward": 0.9875001311302185,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9956616461277008,
      "rewards/forensics_sinkhorn_reward": 0.7843213081359863,
      "rewards/forensics_strict_edge_reward": 0.10843363404273987,
      "rewards/forensics_tiou_reward": 0.6174173355102539,
      "step": 1218
    },
    {
      "completion_length": 24.8125,
      "epoch": 3.740399385560676,
      "grad_norm": 7.225686073303223,
      "kl": 0.4775390625,
      "learning_rate": 6.307692307692307e-08,
      "loss": -0.0506,
      "reward": 0.7532716691493988,
      "reward_std": 0.05466373357921839,
      "reward_std_all": 0.054689071141183376,
      "rewards-max/forensics_sinkhorn_reward": 0.9766739010810852,
      "rewards-max/forensics_strict_edge_reward": 0.5587328374385834,
      "rewards-max/forensics_tiou_reward": 0.9672259986400604,
      "rewards/forensics_sinkhorn_reward": 0.7569147050380707,
      "rewards/forensics_strict_edge_reward": 0.08050550892949104,
      "rewards/forensics_tiou_reward": 0.6094276010990143,
      "step": 1219
    },
    {
      "completion_length": 24.6875,
      "epoch": 3.7434715821812596,
      "grad_norm": 7.882329940795898,
      "kl": 0.5224609375,
      "learning_rate": 6.23076923076923e-08,
      "loss": 0.0028,
      "reward": 0.8004399240016937,
      "reward_std": 0.00967912282794714,
      "reward_std_all": 0.010605049668811262,
      "rewards-max/forensics_sinkhorn_reward": 0.9777988791465759,
      "rewards-max/forensics_strict_edge_reward": 0.4329373836517334,
      "rewards-max/forensics_tiou_reward": 0.971395343542099,
      "rewards/forensics_sinkhorn_reward": 0.8016681969165802,
      "rewards/forensics_strict_edge_reward": 0.04710105434060097,
      "rewards/forensics_tiou_reward": 0.6722172498703003,
      "step": 1220
    },
    {
      "completion_length": 23.65625,
      "epoch": 3.7465437788018434,
      "grad_norm": 31.98575782775879,
      "kl": 1.81640625,
      "learning_rate": 6.153846153846154e-08,
      "loss": 0.0306,
      "reward": 0.7806635499000549,
      "reward_std": 0.04595372639596462,
      "reward_std_all": 0.04122501611709595,
      "rewards-max/forensics_sinkhorn_reward": 0.9772142767906189,
      "rewards-max/forensics_strict_edge_reward": 0.7274474501609802,
      "rewards-max/forensics_tiou_reward": 0.9683722257614136,
      "rewards/forensics_sinkhorn_reward": 0.8342936336994171,
      "rewards/forensics_strict_edge_reward": 0.1000509262084961,
      "rewards/forensics_tiou_reward": 0.7016338109970093,
      "step": 1221
    },
    {
      "completion_length": 22.765625,
      "epoch": 3.749615975422427,
      "grad_norm": 8.659804344177246,
      "kl": 0.3955078125,
      "learning_rate": 6.076923076923077e-08,
      "loss": 0.0262,
      "reward": 0.7748423516750336,
      "reward_std": 0.034184225369244814,
      "reward_std_all": 0.03414753591641784,
      "rewards-max/forensics_sinkhorn_reward": 0.9825289249420166,
      "rewards-max/forensics_strict_edge_reward": 0.6271176636219025,
      "rewards-max/forensics_tiou_reward": 0.9812925457954407,
      "rewards/forensics_sinkhorn_reward": 0.7922729849815369,
      "rewards/forensics_strict_edge_reward": 0.08507628738880157,
      "rewards/forensics_tiou_reward": 0.6810714602470398,
      "step": 1222
    },
    {
      "completion_length": 25.265625,
      "epoch": 3.752688172043011,
      "grad_norm": 8.250619888305664,
      "kl": 0.5380859375,
      "learning_rate": 6e-08,
      "loss": -0.0778,
      "reward": 0.7821480333805084,
      "reward_std": 0.03274893620982766,
      "reward_std_all": 0.031094071455299854,
      "rewards-max/forensics_sinkhorn_reward": 0.9858733117580414,
      "rewards-max/forensics_strict_edge_reward": 0.5798844993114471,
      "rewards-max/forensics_tiou_reward": 0.9914337694644928,
      "rewards/forensics_sinkhorn_reward": 0.791786253452301,
      "rewards/forensics_strict_edge_reward": 0.073517806828022,
      "rewards/forensics_tiou_reward": 0.6743563413619995,
      "step": 1223
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.7557603686635943,
      "grad_norm": 8.973526954650879,
      "kl": 0.4609375,
      "learning_rate": 5.923076923076923e-08,
      "loss": 0.0619,
      "reward": 0.7749577164649963,
      "reward_std": 0.031492212787270546,
      "reward_std_all": 0.033040585927665234,
      "rewards-max/forensics_sinkhorn_reward": 0.9819567501544952,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9822387397289276,
      "rewards/forensics_sinkhorn_reward": 0.8024111688137054,
      "rewards/forensics_strict_edge_reward": 0.09491075202822685,
      "rewards/forensics_tiou_reward": 0.6721704006195068,
      "step": 1224
    },
    {
      "completion_length": 30.640625,
      "epoch": 3.758832565284178,
      "grad_norm": 6.498519420623779,
      "kl": 0.4248046875,
      "learning_rate": 5.8461538461538455e-08,
      "loss": -0.0323,
      "reward": 0.7593197524547577,
      "reward_std": 0.030176709406077862,
      "reward_std_all": 0.039569443091750145,
      "rewards-max/forensics_sinkhorn_reward": 0.9815641343593597,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9805340766906738,
      "rewards/forensics_sinkhorn_reward": 0.7585338652133942,
      "rewards/forensics_strict_edge_reward": 0.07242441922426224,
      "rewards/forensics_tiou_reward": 0.646978884935379,
      "step": 1225
    },
    {
      "completion_length": 24.15625,
      "epoch": 3.761904761904762,
      "grad_norm": 8.932144165039062,
      "kl": 0.37109375,
      "learning_rate": 5.7692307692307695e-08,
      "loss": 0.0723,
      "reward": 0.7738448977470398,
      "reward_std": 0.051376255229115486,
      "reward_std_all": 0.04672708921134472,
      "rewards-max/forensics_sinkhorn_reward": 0.9789519011974335,
      "rewards-max/forensics_strict_edge_reward": 0.765928328037262,
      "rewards-max/forensics_tiou_reward": 0.9744454324245453,
      "rewards/forensics_sinkhorn_reward": 0.7813826203346252,
      "rewards/forensics_strict_edge_reward": 0.09423854947090149,
      "rewards/forensics_tiou_reward": 0.6034775078296661,
      "step": 1226
    },
    {
      "completion_length": 23.5,
      "epoch": 3.7649769585253456,
      "grad_norm": 8.081343650817871,
      "kl": 0.4775390625,
      "learning_rate": 5.692307692307692e-08,
      "loss": -0.1603,
      "reward": 0.7825081944465637,
      "reward_std": 0.0060163275338709354,
      "reward_std_all": 0.01650245673954487,
      "rewards-max/forensics_sinkhorn_reward": 0.9887205064296722,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9989681541919708,
      "rewards/forensics_sinkhorn_reward": 0.7960713803768158,
      "rewards/forensics_strict_edge_reward": 0.1326860934495926,
      "rewards/forensics_tiou_reward": 0.63278928399086,
      "step": 1227
    },
    {
      "completion_length": 24.21875,
      "epoch": 3.7680491551459294,
      "grad_norm": 8.83740520477295,
      "kl": 0.41796875,
      "learning_rate": 5.615384615384615e-08,
      "loss": 0.0656,
      "reward": 0.7468980252742767,
      "reward_std": 0.06297405809164047,
      "reward_std_all": 0.06393903121352196,
      "rewards-max/forensics_sinkhorn_reward": 0.955463707447052,
      "rewards-max/forensics_strict_edge_reward": 0.3065662086009979,
      "rewards-max/forensics_tiou_reward": 0.9275356829166412,
      "rewards/forensics_sinkhorn_reward": 0.6987912058830261,
      "rewards/forensics_strict_edge_reward": 0.031892098020762205,
      "rewards/forensics_tiou_reward": 0.5483762472867966,
      "step": 1228
    },
    {
      "completion_length": 25.1875,
      "epoch": 3.771121351766513,
      "grad_norm": 6.545607566833496,
      "kl": 0.3955078125,
      "learning_rate": 5.538461538461538e-08,
      "loss": -0.0525,
      "reward": 0.8042561411857605,
      "reward_std": 0.0024564527557231486,
      "reward_std_all": 0.004502363270148635,
      "rewards-max/forensics_sinkhorn_reward": 0.9731268882751465,
      "rewards-max/forensics_strict_edge_reward": 0.4715074747800827,
      "rewards-max/forensics_tiou_reward": 0.9506192803382874,
      "rewards/forensics_sinkhorn_reward": 0.8148638904094696,
      "rewards/forensics_strict_edge_reward": 0.05542798526585102,
      "rewards/forensics_tiou_reward": 0.6319078803062439,
      "step": 1229
    },
    {
      "completion_length": 24.046875,
      "epoch": 3.774193548387097,
      "grad_norm": 8.220480918884277,
      "kl": 0.5068359375,
      "learning_rate": 5.461538461538461e-08,
      "loss": 0.031,
      "reward": 0.7936643958091736,
      "reward_std": 0.013951039873063564,
      "reward_std_all": 0.015136091969907284,
      "rewards-max/forensics_sinkhorn_reward": 0.9784715473651886,
      "rewards-max/forensics_strict_edge_reward": 0.3586900904774666,
      "rewards-max/forensics_tiou_reward": 0.9620238542556763,
      "rewards/forensics_sinkhorn_reward": 0.8566047847270966,
      "rewards/forensics_strict_edge_reward": 0.04589667869731784,
      "rewards/forensics_tiou_reward": 0.7382062375545502,
      "step": 1230
    },
    {
      "completion_length": 24.109375,
      "epoch": 3.7772657450076803,
      "grad_norm": 8.901339530944824,
      "kl": 0.5390625,
      "learning_rate": 5.384615384615385e-08,
      "loss": -0.0105,
      "reward": 0.778063952922821,
      "reward_std": 0.032149984617717564,
      "reward_std_all": 0.0319794382667169,
      "rewards-max/forensics_sinkhorn_reward": 0.9807229340076447,
      "rewards-max/forensics_strict_edge_reward": 0.5818096846342087,
      "rewards-max/forensics_tiou_reward": 0.9759288132190704,
      "rewards/forensics_sinkhorn_reward": 0.7928625047206879,
      "rewards/forensics_strict_edge_reward": 0.06829981319606304,
      "rewards/forensics_tiou_reward": 0.626762717962265,
      "step": 1231
    },
    {
      "completion_length": 23.265625,
      "epoch": 3.780337941628264,
      "grad_norm": 12.621438026428223,
      "kl": 0.7099609375,
      "learning_rate": 5.307692307692308e-08,
      "loss": -0.0315,
      "reward": 0.802385002374649,
      "reward_std": 0.007872681191656739,
      "reward_std_all": 0.00858548772521317,
      "rewards-max/forensics_sinkhorn_reward": 0.9836721122264862,
      "rewards-max/forensics_strict_edge_reward": 0.6363584399223328,
      "rewards-max/forensics_tiou_reward": 0.9850880205631256,
      "rewards/forensics_sinkhorn_reward": 0.8269326984882355,
      "rewards/forensics_strict_edge_reward": 0.08096745982766151,
      "rewards/forensics_tiou_reward": 0.7347300052642822,
      "step": 1232
    },
    {
      "completion_length": 26.515625,
      "epoch": 3.783410138248848,
      "grad_norm": 7.4478840827941895,
      "kl": 0.4580078125,
      "learning_rate": 5.2307692307692304e-08,
      "loss": 0.0028,
      "reward": 0.7827872335910797,
      "reward_std": 0.023009218275547028,
      "reward_std_all": 0.02430067304521799,
      "rewards-max/forensics_sinkhorn_reward": 0.9785285890102386,
      "rewards-max/forensics_strict_edge_reward": 0.558412492275238,
      "rewards-max/forensics_tiou_reward": 0.9764644801616669,
      "rewards/forensics_sinkhorn_reward": 0.7920398414134979,
      "rewards/forensics_strict_edge_reward": 0.0871184840798378,
      "rewards/forensics_tiou_reward": 0.6391527950763702,
      "step": 1233
    },
    {
      "completion_length": 25.578125,
      "epoch": 3.7864823348694316,
      "grad_norm": 11.044312477111816,
      "kl": 0.7236328125,
      "learning_rate": 5.153846153846154e-08,
      "loss": 0.0597,
      "reward": 0.7776849567890167,
      "reward_std": 0.049355385825037956,
      "reward_std_all": 0.04468671977519989,
      "rewards-max/forensics_sinkhorn_reward": 0.9777804911136627,
      "rewards-max/forensics_strict_edge_reward": 0.7209268808364868,
      "rewards-max/forensics_tiou_reward": 0.9726678729057312,
      "rewards/forensics_sinkhorn_reward": 0.7364372313022614,
      "rewards/forensics_strict_edge_reward": 0.10426340810954571,
      "rewards/forensics_tiou_reward": 0.6250760555267334,
      "step": 1234
    },
    {
      "completion_length": 27.859375,
      "epoch": 3.7895545314900154,
      "grad_norm": 7.39655876159668,
      "kl": 0.3955078125,
      "learning_rate": 5.0769230769230764e-08,
      "loss": 0.049,
      "reward": 0.7845830619335175,
      "reward_std": 0.02249402366578579,
      "reward_std_all": 0.023319802014157176,
      "rewards-max/forensics_sinkhorn_reward": 0.9777861535549164,
      "rewards-max/forensics_strict_edge_reward": 0.2993010804057121,
      "rewards-max/forensics_tiou_reward": 0.9713631570339203,
      "rewards/forensics_sinkhorn_reward": 0.780714750289917,
      "rewards/forensics_strict_edge_reward": 0.05145206954330206,
      "rewards/forensics_tiou_reward": 0.6469780206680298,
      "step": 1235
    },
    {
      "completion_length": 26.578125,
      "epoch": 3.792626728110599,
      "grad_norm": 7.735136032104492,
      "kl": 0.478515625,
      "learning_rate": 5e-08,
      "loss": -0.0417,
      "reward": 0.7937632203102112,
      "reward_std": 0.02228350774385035,
      "reward_std_all": 0.021077584009617567,
      "rewards-max/forensics_sinkhorn_reward": 0.9807664752006531,
      "rewards-max/forensics_strict_edge_reward": 0.5605834722518921,
      "rewards-max/forensics_tiou_reward": 0.9789583086967468,
      "rewards/forensics_sinkhorn_reward": 0.856695145368576,
      "rewards/forensics_strict_edge_reward": 0.09120291471481323,
      "rewards/forensics_tiou_reward": 0.7284344136714935,
      "step": 1236
    },
    {
      "completion_length": 27.015625,
      "epoch": 3.795698924731183,
      "grad_norm": 7.930419921875,
      "kl": 0.388671875,
      "learning_rate": 4.923076923076923e-08,
      "loss": -0.0197,
      "reward": 0.7541925311088562,
      "reward_std": 0.011053910944610834,
      "reward_std_all": 0.03182855620980263,
      "rewards-max/forensics_sinkhorn_reward": 0.9856267273426056,
      "rewards-max/forensics_strict_edge_reward": 0.6896875947713852,
      "rewards-max/forensics_tiou_reward": 0.991121917963028,
      "rewards/forensics_sinkhorn_reward": 0.7765031456947327,
      "rewards/forensics_strict_edge_reward": 0.09205682948231697,
      "rewards/forensics_tiou_reward": 0.6466626524925232,
      "step": 1237
    },
    {
      "completion_length": 25.734375,
      "epoch": 3.7987711213517663,
      "grad_norm": 7.161715507507324,
      "kl": 0.552734375,
      "learning_rate": 4.846153846153846e-08,
      "loss": 0.015,
      "reward": 0.7887736856937408,
      "reward_std": 0.031199063640087843,
      "reward_std_all": 0.028533363714814186,
      "rewards-max/forensics_sinkhorn_reward": 0.9855785369873047,
      "rewards-max/forensics_strict_edge_reward": 0.799918919801712,
      "rewards-max/forensics_tiou_reward": 0.9899392127990723,
      "rewards/forensics_sinkhorn_reward": 0.8049427568912506,
      "rewards/forensics_strict_edge_reward": 0.09965800866484642,
      "rewards/forensics_tiou_reward": 0.7053819596767426,
      "step": 1238
    },
    {
      "completion_length": 24.59375,
      "epoch": 3.80184331797235,
      "grad_norm": 7.121955394744873,
      "kl": 0.474609375,
      "learning_rate": 4.769230769230769e-08,
      "loss": -0.0206,
      "reward": 0.799063116312027,
      "reward_std": 0.01238703727722168,
      "reward_std_all": 0.01286309864372015,
      "rewards-max/forensics_sinkhorn_reward": 0.9829594492912292,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9826907217502594,
      "rewards/forensics_sinkhorn_reward": 0.8070686757564545,
      "rewards/forensics_strict_edge_reward": 0.05345092713832855,
      "rewards/forensics_tiou_reward": 0.6751653850078583,
      "step": 1239
    },
    {
      "completion_length": 25.4375,
      "epoch": 3.804915514592934,
      "grad_norm": 7.254278182983398,
      "kl": 0.517578125,
      "learning_rate": 4.692307692307692e-08,
      "loss": 0.0177,
      "reward": 0.7969174385070801,
      "reward_std": 0.018486877903342247,
      "reward_std_all": 0.017396976705640554,
      "rewards-max/forensics_sinkhorn_reward": 0.9839060604572296,
      "rewards-max/forensics_strict_edge_reward": 0.6483739018440247,
      "rewards-max/forensics_tiou_reward": 0.9848065972328186,
      "rewards/forensics_sinkhorn_reward": 0.8658084273338318,
      "rewards/forensics_strict_edge_reward": 0.1261814571917057,
      "rewards/forensics_tiou_reward": 0.7482547163963318,
      "step": 1240
    },
    {
      "completion_length": 23.203125,
      "epoch": 3.8079877112135176,
      "grad_norm": 7.475066661834717,
      "kl": 0.466796875,
      "learning_rate": 4.615384615384615e-08,
      "loss": 0.0031,
      "reward": 0.7924013733863831,
      "reward_std": 0.023554551415145397,
      "reward_std_all": 0.022366515127941966,
      "rewards-max/forensics_sinkhorn_reward": 0.9701966047286987,
      "rewards-max/forensics_strict_edge_reward": 0.548613041639328,
      "rewards-max/forensics_tiou_reward": 0.9508950412273407,
      "rewards/forensics_sinkhorn_reward": 0.7807972729206085,
      "rewards/forensics_strict_edge_reward": 0.05992231331765652,
      "rewards/forensics_tiou_reward": 0.6148500144481659,
      "step": 1241
    },
    {
      "completion_length": 24.90625,
      "epoch": 3.8110599078341014,
      "grad_norm": 7.102672576904297,
      "kl": 0.439453125,
      "learning_rate": 4.538461538461538e-08,
      "loss": 0.0144,
      "reward": 0.8007246255874634,
      "reward_std": 0.010035219776909798,
      "reward_std_all": 0.010535428300499916,
      "rewards-max/forensics_sinkhorn_reward": 0.9840664565563202,
      "rewards-max/forensics_strict_edge_reward": 0.5646951347589493,
      "rewards-max/forensics_tiou_reward": 0.9869841039180756,
      "rewards/forensics_sinkhorn_reward": 0.8103031516075134,
      "rewards/forensics_strict_edge_reward": 0.08410526439547539,
      "rewards/forensics_tiou_reward": 0.7134162485599518,
      "step": 1242
    },
    {
      "completion_length": 23.0625,
      "epoch": 3.814132104454685,
      "grad_norm": 7.171328544616699,
      "kl": 0.4453125,
      "learning_rate": 4.461538461538461e-08,
      "loss": 0.0255,
      "reward": 0.7730170786380768,
      "reward_std": 0.03746923431754112,
      "reward_std_all": 0.03713439777493477,
      "rewards-max/forensics_sinkhorn_reward": 0.9780508875846863,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9746838808059692,
      "rewards/forensics_sinkhorn_reward": 0.784283310174942,
      "rewards/forensics_strict_edge_reward": 0.07003672420978546,
      "rewards/forensics_tiou_reward": 0.6207693219184875,
      "step": 1243
    },
    {
      "completion_length": 24.71875,
      "epoch": 3.817204301075269,
      "grad_norm": 7.465111255645752,
      "kl": 0.4580078125,
      "learning_rate": 4.3846153846153846e-08,
      "loss": -0.0397,
      "reward": 0.8005511462688446,
      "reward_std": 0.011668641935102642,
      "reward_std_all": 0.011677019065245986,
      "rewards-max/forensics_sinkhorn_reward": 0.9831916987895966,
      "rewards-max/forensics_strict_edge_reward": 0.6666710078716278,
      "rewards-max/forensics_tiou_reward": 0.9862793385982513,
      "rewards/forensics_sinkhorn_reward": 0.8412482440471649,
      "rewards/forensics_strict_edge_reward": 0.09964410960674286,
      "rewards/forensics_tiou_reward": 0.7712308764457703,
      "step": 1244
    },
    {
      "completion_length": 22.4375,
      "epoch": 3.8202764976958523,
      "grad_norm": 9.146687507629395,
      "kl": 0.486328125,
      "learning_rate": 4.307692307692307e-08,
      "loss": -0.0924,
      "reward": 0.8016782701015472,
      "reward_std": 0.008508300175890326,
      "reward_std_all": 0.009225510060787201,
      "rewards-max/forensics_sinkhorn_reward": 0.9833073914051056,
      "rewards-max/forensics_strict_edge_reward": 0.5598244965076447,
      "rewards-max/forensics_tiou_reward": 0.9835446178913116,
      "rewards/forensics_sinkhorn_reward": 0.8402076959609985,
      "rewards/forensics_strict_edge_reward": 0.06662335619330406,
      "rewards/forensics_tiou_reward": 0.7023111879825592,
      "step": 1245
    },
    {
      "completion_length": 25.015625,
      "epoch": 3.823348694316436,
      "grad_norm": 6.712521076202393,
      "kl": 0.57421875,
      "learning_rate": 4.230769230769231e-08,
      "loss": -0.0006,
      "reward": 0.8030715882778168,
      "reward_std": 0.004625785397365689,
      "reward_std_all": 0.006128163076937199,
      "rewards-max/forensics_sinkhorn_reward": 0.9864895343780518,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9917022287845612,
      "rewards/forensics_sinkhorn_reward": 0.8421220779418945,
      "rewards/forensics_strict_edge_reward": 0.13334056362509727,
      "rewards/forensics_tiou_reward": 0.6790767908096313,
      "step": 1246
    },
    {
      "completion_length": 21.421875,
      "epoch": 3.82642089093702,
      "grad_norm": 7.894344329833984,
      "kl": 0.494140625,
      "learning_rate": 4.1538461538461534e-08,
      "loss": 0.0454,
      "reward": 0.7088963985443115,
      "reward_std": 0.03746416326612234,
      "reward_std_all": 0.07030638307332993,
      "rewards-max/forensics_sinkhorn_reward": 0.9828448593616486,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9827525317668915,
      "rewards/forensics_sinkhorn_reward": 0.7325071394443512,
      "rewards/forensics_strict_edge_reward": 0.08638608455657959,
      "rewards/forensics_tiou_reward": 0.6061455607414246,
      "step": 1247
    },
    {
      "completion_length": 29.84375,
      "epoch": 3.8294930875576036,
      "grad_norm": 10.351165771484375,
      "kl": 0.47265625,
      "learning_rate": 4.076923076923077e-08,
      "loss": -0.0373,
      "reward": 0.7955181300640106,
      "reward_std": 0.017534923157654703,
      "reward_std_all": 0.017529650358483195,
      "rewards-max/forensics_sinkhorn_reward": 0.9706844687461853,
      "rewards-max/forensics_strict_edge_reward": 0.4030446261167526,
      "rewards-max/forensics_tiou_reward": 0.9487945437431335,
      "rewards/forensics_sinkhorn_reward": 0.7556638717651367,
      "rewards/forensics_strict_edge_reward": 0.04219529405236244,
      "rewards/forensics_tiou_reward": 0.6412143111228943,
      "step": 1248
    },
    {
      "completion_length": 26.609375,
      "epoch": 3.8325652841781874,
      "grad_norm": 6.635225296020508,
      "kl": 0.3916015625,
      "learning_rate": 4e-08,
      "loss": 0.0101,
      "reward": 0.740779846906662,
      "reward_std": 0.04390971455723047,
      "reward_std_all": 0.055044230073690414,
      "rewards-max/forensics_sinkhorn_reward": 0.9677266776561737,
      "rewards-max/forensics_strict_edge_reward": 0.3254912942647934,
      "rewards-max/forensics_tiou_reward": 0.9458546936511993,
      "rewards/forensics_sinkhorn_reward": 0.7341951727867126,
      "rewards/forensics_strict_edge_reward": 0.04852787684649229,
      "rewards/forensics_tiou_reward": 0.5945957899093628,
      "step": 1249
    },
    {
      "completion_length": 26.640625,
      "epoch": 3.835637480798771,
      "grad_norm": 8.555872917175293,
      "kl": 0.625,
      "learning_rate": 3.923076923076923e-08,
      "loss": 0.0621,
      "reward": 0.7697598934173584,
      "reward_std": 0.03859625477343798,
      "reward_std_all": 0.03898357227444649,
      "rewards-max/forensics_sinkhorn_reward": 0.9840797781944275,
      "rewards-max/forensics_strict_edge_reward": 0.7442346066236496,
      "rewards-max/forensics_tiou_reward": 0.9899778962135315,
      "rewards/forensics_sinkhorn_reward": 0.7020111978054047,
      "rewards/forensics_strict_edge_reward": 0.11226006969809532,
      "rewards/forensics_tiou_reward": 0.5843055844306946,
      "step": 1250
    },
    {
      "completion_length": 26.421875,
      "epoch": 3.838709677419355,
      "grad_norm": 6.483493328094482,
      "kl": 0.4443359375,
      "learning_rate": 3.846153846153846e-08,
      "loss": 0.0108,
      "reward": 0.804044246673584,
      "reward_std": 0.003863850375637412,
      "reward_std_all": 0.0052627939730882645,
      "rewards-max/forensics_sinkhorn_reward": 0.9699650704860687,
      "rewards-max/forensics_strict_edge_reward": 0.44073912501335144,
      "rewards-max/forensics_tiou_reward": 0.9526437520980835,
      "rewards/forensics_sinkhorn_reward": 0.7999041676521301,
      "rewards/forensics_strict_edge_reward": 0.06925516575574875,
      "rewards/forensics_tiou_reward": 0.6824100613594055,
      "step": 1251
    },
    {
      "completion_length": 25.46875,
      "epoch": 3.8417818740399383,
      "grad_norm": 8.087485313415527,
      "kl": 0.40625,
      "learning_rate": 3.769230769230769e-08,
      "loss": -0.0516,
      "reward": 0.7549595236778259,
      "reward_std": 0.023032850585877895,
      "reward_std_all": 0.0361322145909071,
      "rewards-max/forensics_sinkhorn_reward": 0.9852519929409027,
      "rewards-max/forensics_strict_edge_reward": 0.5623422563076019,
      "rewards-max/forensics_tiou_reward": 0.9900299906730652,
      "rewards/forensics_sinkhorn_reward": 0.7474082410335541,
      "rewards/forensics_strict_edge_reward": 0.04959848336875439,
      "rewards/forensics_tiou_reward": 0.5784483700990677,
      "step": 1252
    },
    {
      "completion_length": 24.640625,
      "epoch": 3.8448540706605225,
      "grad_norm": 9.114455223083496,
      "kl": 0.517578125,
      "learning_rate": 3.692307692307692e-08,
      "loss": -0.0193,
      "reward": 0.7994534373283386,
      "reward_std": 0.013429728802293539,
      "reward_std_all": 0.013057676143944263,
      "rewards-max/forensics_sinkhorn_reward": 0.9862919449806213,
      "rewards-max/forensics_strict_edge_reward": 0.6736744344234467,
      "rewards-max/forensics_tiou_reward": 0.9933468699455261,
      "rewards/forensics_sinkhorn_reward": 0.8409172892570496,
      "rewards/forensics_strict_edge_reward": 0.09904950484633446,
      "rewards/forensics_tiou_reward": 0.7331065535545349,
      "step": 1253
    },
    {
      "completion_length": 24.34375,
      "epoch": 3.847926267281106,
      "grad_norm": 8.61190128326416,
      "kl": 0.4169921875,
      "learning_rate": 3.6153846153846155e-08,
      "loss": 0.0626,
      "reward": 0.760404646396637,
      "reward_std": 0.0413419040851295,
      "reward_std_all": 0.043296465650200844,
      "rewards-max/forensics_sinkhorn_reward": 0.9777190089225769,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9730135798454285,
      "rewards/forensics_sinkhorn_reward": 0.7533313632011414,
      "rewards/forensics_strict_edge_reward": 0.0916699692606926,
      "rewards/forensics_tiou_reward": 0.5812940001487732,
      "step": 1254
    },
    {
      "completion_length": 23.390625,
      "epoch": 3.8509984639016897,
      "grad_norm": 7.960991859436035,
      "kl": 0.4794921875,
      "learning_rate": 3.538461538461538e-08,
      "loss": -0.0344,
      "reward": 0.7374943196773529,
      "reward_std": 0.06474540941417217,
      "reward_std_all": 0.06602023169398308,
      "rewards-max/forensics_sinkhorn_reward": 0.978290319442749,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9741885364055634,
      "rewards/forensics_sinkhorn_reward": 0.7504125833511353,
      "rewards/forensics_strict_edge_reward": 0.09174048155546188,
      "rewards/forensics_tiou_reward": 0.5983934998512268,
      "step": 1255
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.8540706605222734,
      "grad_norm": 9.178834915161133,
      "kl": 0.4013671875,
      "learning_rate": 3.4615384615384616e-08,
      "loss": -0.0396,
      "reward": 0.796399861574173,
      "reward_std": 0.008829079568386078,
      "reward_std_all": 0.01078396337106824,
      "rewards-max/forensics_sinkhorn_reward": 0.9827641248703003,
      "rewards-max/forensics_strict_edge_reward": 0.6777393221855164,
      "rewards-max/forensics_tiou_reward": 0.9828922748565674,
      "rewards/forensics_sinkhorn_reward": 0.7226147949695587,
      "rewards/forensics_strict_edge_reward": 0.08835194259881973,
      "rewards/forensics_tiou_reward": 0.6650774478912354,
      "step": 1256
    },
    {
      "completion_length": 22.765625,
      "epoch": 3.857142857142857,
      "grad_norm": 7.747682571411133,
      "kl": 0.4775390625,
      "learning_rate": 3.384615384615384e-08,
      "loss": -0.0096,
      "reward": 0.7990767359733582,
      "reward_std": 0.01170419529080391,
      "reward_std_all": 0.01232875045388937,
      "rewards-max/forensics_sinkhorn_reward": 0.9735161662101746,
      "rewards-max/forensics_strict_edge_reward": 0.5426253378391266,
      "rewards-max/forensics_tiou_reward": 0.9586793780326843,
      "rewards/forensics_sinkhorn_reward": 0.8148502707481384,
      "rewards/forensics_strict_edge_reward": 0.05051586404442787,
      "rewards/forensics_tiou_reward": 0.6418268084526062,
      "step": 1257
    },
    {
      "completion_length": 22.96875,
      "epoch": 3.860215053763441,
      "grad_norm": 10.554880142211914,
      "kl": 0.3974609375,
      "learning_rate": 3.3076923076923076e-08,
      "loss": -0.0503,
      "reward": 0.8044154644012451,
      "reward_std": 0.0045118804555386305,
      "reward_std_all": 0.005517232697457075,
      "rewards-max/forensics_sinkhorn_reward": 0.9847997725009918,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9900259077548981,
      "rewards/forensics_sinkhorn_reward": 0.8343338966369629,
      "rewards/forensics_strict_edge_reward": 0.12428371608257294,
      "rewards/forensics_tiou_reward": 0.6892260909080505,
      "step": 1258
    },
    {
      "completion_length": 22.21875,
      "epoch": 3.8632872503840243,
      "grad_norm": 8.258501052856445,
      "kl": 0.75,
      "learning_rate": 3.230769230769231e-08,
      "loss": 0.0251,
      "reward": 0.7837359309196472,
      "reward_std": 0.0403152946382761,
      "reward_std_all": 0.03644641488790512,
      "rewards-max/forensics_sinkhorn_reward": 0.9817340970039368,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9829685986042023,
      "rewards/forensics_sinkhorn_reward": 0.8226963877677917,
      "rewards/forensics_strict_edge_reward": 0.11112193390727043,
      "rewards/forensics_tiou_reward": 0.7003602385520935,
      "step": 1259
    },
    {
      "completion_length": 27.5625,
      "epoch": 3.8663594470046085,
      "grad_norm": 7.224905490875244,
      "kl": 0.4296875,
      "learning_rate": 3.153846153846154e-08,
      "loss": -0.0549,
      "reward": 0.7893167734146118,
      "reward_std": 0.03058625222183764,
      "reward_std_all": 0.028071572072803974,
      "rewards-max/forensics_sinkhorn_reward": 0.9841863512992859,
      "rewards-max/forensics_strict_edge_reward": 0.5487441271543503,
      "rewards-max/forensics_tiou_reward": 0.9890499413013458,
      "rewards/forensics_sinkhorn_reward": 0.8290771543979645,
      "rewards/forensics_strict_edge_reward": 0.10753587260842323,
      "rewards/forensics_tiou_reward": 0.7188666462898254,
      "step": 1260
    },
    {
      "completion_length": 26.65625,
      "epoch": 3.869431643625192,
      "grad_norm": 7.555004119873047,
      "kl": 0.4677734375,
      "learning_rate": 3.076923076923077e-08,
      "loss": 0.0444,
      "reward": 0.7699134051799774,
      "reward_std": 0.04840238578617573,
      "reward_std_all": 0.04570429585874081,
      "rewards-max/forensics_sinkhorn_reward": 0.9794159233570099,
      "rewards-max/forensics_strict_edge_reward": 0.5832466036081314,
      "rewards-max/forensics_tiou_reward": 0.9737230539321899,
      "rewards/forensics_sinkhorn_reward": 0.7572659552097321,
      "rewards/forensics_strict_edge_reward": 0.08329692110419273,
      "rewards/forensics_tiou_reward": 0.5769550800323486,
      "step": 1261
    },
    {
      "completion_length": 27.46875,
      "epoch": 3.8725038402457757,
      "grad_norm": 7.467609882354736,
      "kl": 0.470703125,
      "learning_rate": 3e-08,
      "loss": -0.0291,
      "reward": 0.7958205342292786,
      "reward_std": 0.01112619275227189,
      "reward_std_all": 0.012367004062980413,
      "rewards-max/forensics_sinkhorn_reward": 0.9849605560302734,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9894698858261108,
      "rewards/forensics_sinkhorn_reward": 0.802171915769577,
      "rewards/forensics_strict_edge_reward": 0.11116793751716614,
      "rewards/forensics_tiou_reward": 0.685651957988739,
      "step": 1262
    },
    {
      "completion_length": 27.3125,
      "epoch": 3.8755760368663594,
      "grad_norm": 7.723165988922119,
      "kl": 0.435546875,
      "learning_rate": 2.9230769230769228e-08,
      "loss": 0.0418,
      "reward": 0.7985414266586304,
      "reward_std": 0.013289196183905005,
      "reward_std_all": 0.013466862961649895,
      "rewards-max/forensics_sinkhorn_reward": 0.9811231791973114,
      "rewards-max/forensics_strict_edge_reward": 0.7482926547527313,
      "rewards-max/forensics_tiou_reward": 0.9794377684593201,
      "rewards/forensics_sinkhorn_reward": 0.7810222506523132,
      "rewards/forensics_strict_edge_reward": 0.10873143374919891,
      "rewards/forensics_tiou_reward": 0.6611970663070679,
      "step": 1263
    },
    {
      "completion_length": 24.328125,
      "epoch": 3.878648233486943,
      "grad_norm": 9.580599784851074,
      "kl": 0.408203125,
      "learning_rate": 2.846153846153846e-08,
      "loss": -0.0557,
      "reward": 0.8035602867603302,
      "reward_std": 0.003223102306947112,
      "reward_std_all": 0.005072775646112859,
      "rewards-max/forensics_sinkhorn_reward": 0.9759395122528076,
      "rewards-max/forensics_strict_edge_reward": 0.6126891374588013,
      "rewards-max/forensics_tiou_reward": 0.9652274549007416,
      "rewards/forensics_sinkhorn_reward": 0.8296006917953491,
      "rewards/forensics_strict_edge_reward": 0.05728661268949509,
      "rewards/forensics_tiou_reward": 0.7003399431705475,
      "step": 1264
    },
    {
      "completion_length": 27.5,
      "epoch": 3.881720430107527,
      "grad_norm": 9.561162948608398,
      "kl": 0.677734375,
      "learning_rate": 2.769230769230769e-08,
      "loss": 0.0162,
      "reward": 0.7935406863689423,
      "reward_std": 0.01577805168926716,
      "reward_std_all": 0.016562893986701965,
      "rewards-max/forensics_sinkhorn_reward": 0.9835447669029236,
      "rewards-max/forensics_strict_edge_reward": 0.4967339038848877,
      "rewards-max/forensics_tiou_reward": 0.984542578458786,
      "rewards/forensics_sinkhorn_reward": 0.7600825428962708,
      "rewards/forensics_strict_edge_reward": 0.07140149921178818,
      "rewards/forensics_tiou_reward": 0.6253823935985565,
      "step": 1265
    },
    {
      "completion_length": 25.796875,
      "epoch": 3.8847926267281108,
      "grad_norm": 7.073470115661621,
      "kl": 0.4267578125,
      "learning_rate": 2.6923076923076925e-08,
      "loss": -0.1002,
      "reward": 0.7434865236282349,
      "reward_std": 0.025562027003616095,
      "reward_std_all": 0.0443398542702198,
      "rewards-max/forensics_sinkhorn_reward": 0.9775323271751404,
      "rewards-max/forensics_strict_edge_reward": 0.6119300425052643,
      "rewards-max/forensics_tiou_reward": 0.9723609387874603,
      "rewards/forensics_sinkhorn_reward": 0.7573242783546448,
      "rewards/forensics_strict_edge_reward": 0.07034815102815628,
      "rewards/forensics_tiou_reward": 0.6409263610839844,
      "step": 1266
    },
    {
      "completion_length": 24.296875,
      "epoch": 3.8878648233486945,
      "grad_norm": 9.531238555908203,
      "kl": 0.744140625,
      "learning_rate": 2.6153846153846152e-08,
      "loss": 0.0407,
      "reward": 0.7787692844867706,
      "reward_std": 0.03184039145708084,
      "reward_std_all": 0.03166462015360594,
      "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": 0.796749085187912,
      "rewards/forensics_strict_edge_reward": 0.10065729916095734,
      "rewards/forensics_tiou_reward": 0.6478928029537201,
      "step": 1267
    },
    {
      "completion_length": 26.359375,
      "epoch": 3.890937019969278,
      "grad_norm": 8.89073371887207,
      "kl": 0.3349609375,
      "learning_rate": 2.5384615384615382e-08,
      "loss": -0.0243,
      "reward": 0.8006956875324249,
      "reward_std": 0.009635969181545079,
      "reward_std_all": 0.010331329889595509,
      "rewards-max/forensics_sinkhorn_reward": 0.9752892553806305,
      "rewards-max/forensics_strict_edge_reward": 0.7268245220184326,
      "rewards-max/forensics_tiou_reward": 0.9601251184940338,
      "rewards/forensics_sinkhorn_reward": 0.7840767502784729,
      "rewards/forensics_strict_edge_reward": 0.09123993292450905,
      "rewards/forensics_tiou_reward": 0.6309652924537659,
      "step": 1268
    },
    {
      "completion_length": 26.46875,
      "epoch": 3.8940092165898617,
      "grad_norm": 9.36065673828125,
      "kl": 0.41796875,
      "learning_rate": 2.4615384615384616e-08,
      "loss": 0.0368,
      "reward": 0.764168918132782,
      "reward_std": 0.03589736111462116,
      "reward_std_all": 0.040344465523958206,
      "rewards-max/forensics_sinkhorn_reward": 0.9816239774227142,
      "rewards-max/forensics_strict_edge_reward": 0.6303281784057617,
      "rewards-max/forensics_tiou_reward": 0.9728663265705109,
      "rewards/forensics_sinkhorn_reward": 0.7781603932380676,
      "rewards/forensics_strict_edge_reward": 0.07121346704661846,
      "rewards/forensics_tiou_reward": 0.6211086511611938,
      "step": 1269
    },
    {
      "completion_length": 26.234375,
      "epoch": 3.8970814132104454,
      "grad_norm": 14.167485237121582,
      "kl": 0.9169921875,
      "learning_rate": 2.3846153846153846e-08,
      "loss": 0.1935,
      "reward": 0.773595929145813,
      "reward_std": 0.049556247889995575,
      "reward_std_all": 0.04521438665688038,
      "rewards-max/forensics_sinkhorn_reward": 0.983514130115509,
      "rewards-max/forensics_strict_edge_reward": 0.6806710362434387,
      "rewards-max/forensics_tiou_reward": 0.9869424104690552,
      "rewards/forensics_sinkhorn_reward": 0.7614015936851501,
      "rewards/forensics_strict_edge_reward": 0.18004147708415985,
      "rewards/forensics_tiou_reward": 0.6571693420410156,
      "step": 1270
    },
    {
      "completion_length": 24.015625,
      "epoch": 3.900153609831029,
      "grad_norm": 7.01627254486084,
      "kl": 0.640625,
      "learning_rate": 2.3076923076923076e-08,
      "loss": -0.0232,
      "reward": 0.7932348847389221,
      "reward_std": 0.024254062213003635,
      "reward_std_all": 0.022524168249219656,
      "rewards-max/forensics_sinkhorn_reward": 0.984443724155426,
      "rewards-max/forensics_strict_edge_reward": 0.5007785111665726,
      "rewards-max/forensics_tiou_reward": 0.9867502450942993,
      "rewards/forensics_sinkhorn_reward": 0.8484255373477936,
      "rewards/forensics_strict_edge_reward": 0.10795671865344048,
      "rewards/forensics_tiou_reward": 0.723926842212677,
      "step": 1271
    },
    {
      "completion_length": 23.171875,
      "epoch": 3.903225806451613,
      "grad_norm": 7.7939019203186035,
      "kl": 0.384765625,
      "learning_rate": 2.2307692307692306e-08,
      "loss": 0.0896,
      "reward": 0.7572092413902283,
      "reward_std": 0.04287527874112129,
      "reward_std_all": 0.04819735698401928,
      "rewards-max/forensics_sinkhorn_reward": 0.9658398330211639,
      "rewards-max/forensics_strict_edge_reward": 0.40730567276477814,
      "rewards-max/forensics_tiou_reward": 0.9431703388690948,
      "rewards/forensics_sinkhorn_reward": 0.7427719533443451,
      "rewards/forensics_strict_edge_reward": 0.04158003441989422,
      "rewards/forensics_tiou_reward": 0.5546555817127228,
      "step": 1272
    },
    {
      "completion_length": 25.78125,
      "epoch": 3.9062980030721968,
      "grad_norm": 10.269336700439453,
      "kl": 0.482421875,
      "learning_rate": 2.1538461538461537e-08,
      "loss": 0.0285,
      "reward": 0.7636694312095642,
      "reward_std": 0.04124036291614175,
      "reward_std_all": 0.0419964543543756,
      "rewards-max/forensics_sinkhorn_reward": 0.9827752411365509,
      "rewards-max/forensics_strict_edge_reward": 0.4094945788383484,
      "rewards-max/forensics_tiou_reward": 0.984455406665802,
      "rewards/forensics_sinkhorn_reward": 0.720381885766983,
      "rewards/forensics_strict_edge_reward": 0.06076483242213726,
      "rewards/forensics_tiou_reward": 0.5739938914775848,
      "step": 1273
    },
    {
      "completion_length": 23.609375,
      "epoch": 3.9093701996927805,
      "grad_norm": 8.626166343688965,
      "kl": 0.4072265625,
      "learning_rate": 2.0769230769230767e-08,
      "loss": -0.0228,
      "reward": 0.7855215072631836,
      "reward_std": 0.02797328680753708,
      "reward_std_all": 0.026999671012163162,
      "rewards-max/forensics_sinkhorn_reward": 0.9831899702548981,
      "rewards-max/forensics_strict_edge_reward": 0.6006690710783005,
      "rewards-max/forensics_tiou_reward": 0.9825886785984039,
      "rewards/forensics_sinkhorn_reward": 0.8178232610225677,
      "rewards/forensics_strict_edge_reward": 0.06664603762328625,
      "rewards/forensics_tiou_reward": 0.6583267152309418,
      "step": 1274
    },
    {
      "completion_length": 29.890625,
      "epoch": 3.912442396313364,
      "grad_norm": 8.343360900878906,
      "kl": 0.5888671875,
      "learning_rate": 2e-08,
      "loss": -0.0687,
      "reward": 0.776080995798111,
      "reward_std": 0.012411969248205423,
      "reward_std_all": 0.020976636558771133,
      "rewards-max/forensics_sinkhorn_reward": 0.9818906486034393,
      "rewards-max/forensics_strict_edge_reward": 0.7403165996074677,
      "rewards-max/forensics_tiou_reward": 0.9833154082298279,
      "rewards/forensics_sinkhorn_reward": 0.7538551688194275,
      "rewards/forensics_strict_edge_reward": 0.08022320829331875,
      "rewards/forensics_tiou_reward": 0.6688726246356964,
      "step": 1275
    },
    {
      "completion_length": 25.640625,
      "epoch": 3.9155145929339477,
      "grad_norm": 8.85676383972168,
      "kl": 0.494140625,
      "learning_rate": 1.923076923076923e-08,
      "loss": 0.0072,
      "reward": 0.7948687970638275,
      "reward_std": 0.019298496190458536,
      "reward_std_all": 0.018568684114143252,
      "rewards-max/forensics_sinkhorn_reward": 0.9710297584533691,
      "rewards-max/forensics_strict_edge_reward": 0.5227026343345642,
      "rewards-max/forensics_tiou_reward": 0.9521034359931946,
      "rewards/forensics_sinkhorn_reward": 0.7866873741149902,
      "rewards/forensics_strict_edge_reward": 0.07035551220178604,
      "rewards/forensics_tiou_reward": 0.6290317177772522,
      "step": 1276
    },
    {
      "completion_length": 24.203125,
      "epoch": 3.9185867895545314,
      "grad_norm": 9.97804069519043,
      "kl": 0.693359375,
      "learning_rate": 1.846153846153846e-08,
      "loss": 0.0279,
      "reward": 0.7682115435600281,
      "reward_std": 0.05777024291455746,
      "reward_std_all": 0.05309536587446928,
      "rewards-max/forensics_sinkhorn_reward": 0.9788991212844849,
      "rewards-max/forensics_strict_edge_reward": 0.38722455501556396,
      "rewards-max/forensics_tiou_reward": 0.9660095572471619,
      "rewards/forensics_sinkhorn_reward": 0.786234587430954,
      "rewards/forensics_strict_edge_reward": 0.0370671022683382,
      "rewards/forensics_tiou_reward": 0.6172438263893127,
      "step": 1277
    },
    {
      "completion_length": 29.578125,
      "epoch": 3.921658986175115,
      "grad_norm": 12.238926887512207,
      "kl": 0.4228515625,
      "learning_rate": 1.769230769230769e-08,
      "loss": -0.0771,
      "reward": 0.7856571674346924,
      "reward_std": 0.027001410780940205,
      "reward_std_all": 0.02632769919000566,
      "rewards-max/forensics_sinkhorn_reward": 0.9818413853645325,
      "rewards-max/forensics_strict_edge_reward": 0.5360626578330994,
      "rewards-max/forensics_tiou_reward": 0.9814470112323761,
      "rewards/forensics_sinkhorn_reward": 0.8003786504268646,
      "rewards/forensics_strict_edge_reward": 0.07633552327752113,
      "rewards/forensics_tiou_reward": 0.6257896423339844,
      "step": 1278
    },
    {
      "completion_length": 22.40625,
      "epoch": 3.924731182795699,
      "grad_norm": 7.954892635345459,
      "kl": 0.458984375,
      "learning_rate": 1.692307692307692e-08,
      "loss": 0.0548,
      "reward": 0.7932868003845215,
      "reward_std": 0.017675881274044514,
      "reward_std_all": 0.017723570112138987,
      "rewards-max/forensics_sinkhorn_reward": 0.9816353023052216,
      "rewards-max/forensics_strict_edge_reward": 0.5905451774597168,
      "rewards-max/forensics_tiou_reward": 0.9813247621059418,
      "rewards/forensics_sinkhorn_reward": 0.7789453864097595,
      "rewards/forensics_strict_edge_reward": 0.040101575665175915,
      "rewards/forensics_tiou_reward": 0.591927707195282,
      "step": 1279
    },
    {
      "completion_length": 28.25,
      "epoch": 3.9278033794162828,
      "grad_norm": 7.156748294830322,
      "kl": 0.48828125,
      "learning_rate": 1.6153846153846155e-08,
      "loss": 0.0715,
      "reward": 0.7661137878894806,
      "reward_std": 0.054163048043847084,
      "reward_std_all": 0.05061734467744827,
      "rewards-max/forensics_sinkhorn_reward": 0.9796037375926971,
      "rewards-max/forensics_strict_edge_reward": 0.4208007901906967,
      "rewards-max/forensics_tiou_reward": 0.9736413359642029,
      "rewards/forensics_sinkhorn_reward": 0.7536011636257172,
      "rewards/forensics_strict_edge_reward": 0.046798624098300934,
      "rewards/forensics_tiou_reward": 0.5792476534843445,
      "step": 1280
    },
    {
      "completion_length": 25.140625,
      "epoch": 3.9308755760368665,
      "grad_norm": 8.229083061218262,
      "kl": 0.658203125,
      "learning_rate": 1.5384615384615385e-08,
      "loss": 0.0249,
      "reward": 0.7977930903434753,
      "reward_std": 0.008826898410916328,
      "reward_std_all": 0.010556994006037712,
      "rewards-max/forensics_sinkhorn_reward": 0.981309711933136,
      "rewards-max/forensics_strict_edge_reward": 0.5798844993114471,
      "rewards-max/forensics_tiou_reward": 0.9826705753803253,
      "rewards/forensics_sinkhorn_reward": 0.8253332674503326,
      "rewards/forensics_strict_edge_reward": 0.06432950124144554,
      "rewards/forensics_tiou_reward": 0.6471114754676819,
      "step": 1281
    },
    {
      "completion_length": 23.484375,
      "epoch": 3.93394777265745,
      "grad_norm": 6.862236499786377,
      "kl": 0.37890625,
      "learning_rate": 1.4615384615384614e-08,
      "loss": -0.0128,
      "reward": 0.7873952388763428,
      "reward_std": 0.009688617545180023,
      "reward_std_all": 0.016478104516863823,
      "rewards-max/forensics_sinkhorn_reward": 0.9773257374763489,
      "rewards-max/forensics_strict_edge_reward": 0.3786467704921961,
      "rewards-max/forensics_tiou_reward": 0.965317964553833,
      "rewards/forensics_sinkhorn_reward": 0.8011711537837982,
      "rewards/forensics_strict_edge_reward": 0.0687672896310687,
      "rewards/forensics_tiou_reward": 0.6201498061418533,
      "step": 1282
    },
    {
      "completion_length": 22.578125,
      "epoch": 3.9370199692780337,
      "grad_norm": 8.00267505645752,
      "kl": 0.509765625,
      "learning_rate": 1.3846153846153846e-08,
      "loss": -0.0785,
      "reward": 0.7625242471694946,
      "reward_std": 0.03068565670400858,
      "reward_std_all": 0.0374557189643383,
      "rewards-max/forensics_sinkhorn_reward": 0.9877511560916901,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9965972900390625,
      "rewards/forensics_sinkhorn_reward": 0.7730274796485901,
      "rewards/forensics_strict_edge_reward": 0.1098649762570858,
      "rewards/forensics_tiou_reward": 0.6254750490188599,
      "step": 1283
    },
    {
      "completion_length": 26.796875,
      "epoch": 3.9400921658986174,
      "grad_norm": 7.001213073730469,
      "kl": 0.484375,
      "learning_rate": 1.3076923076923076e-08,
      "loss": 0.0321,
      "reward": 0.7841679751873016,
      "reward_std": 0.03214059863239527,
      "reward_std_all": 0.030296042561531067,
      "rewards-max/forensics_sinkhorn_reward": 0.9774461388587952,
      "rewards-max/forensics_strict_edge_reward": 0.49332477152347565,
      "rewards-max/forensics_tiou_reward": 0.9699226915836334,
      "rewards/forensics_sinkhorn_reward": 0.8067913949489594,
      "rewards/forensics_strict_edge_reward": 0.058548253029584885,
      "rewards/forensics_tiou_reward": 0.6599321663379669,
      "step": 1284
    },
    {
      "completion_length": 27.40625,
      "epoch": 3.943164362519201,
      "grad_norm": 7.681600093841553,
      "kl": 0.423828125,
      "learning_rate": 1.2307692307692308e-08,
      "loss": -0.0734,
      "reward": 0.7976633310317993,
      "reward_std": 0.01121147081721574,
      "reward_std_all": 0.012366014765575528,
      "rewards-max/forensics_sinkhorn_reward": 0.9713177978992462,
      "rewards-max/forensics_strict_edge_reward": 0.5526288747787476,
      "rewards-max/forensics_tiou_reward": 0.954491525888443,
      "rewards/forensics_sinkhorn_reward": 0.7868110537528992,
      "rewards/forensics_strict_edge_reward": 0.04493139125406742,
      "rewards/forensics_tiou_reward": 0.606326699256897,
      "step": 1285
    },
    {
      "completion_length": 24.359375,
      "epoch": 3.946236559139785,
      "grad_norm": 11.84724235534668,
      "kl": 0.638671875,
      "learning_rate": 1.1538461538461538e-08,
      "loss": 0.0435,
      "reward": 0.7882505059242249,
      "reward_std": 0.03156610857695341,
      "reward_std_all": 0.029152432456612587,
      "rewards-max/forensics_sinkhorn_reward": 0.9689492583274841,
      "rewards-max/forensics_strict_edge_reward": 0.4043564051389694,
      "rewards-max/forensics_tiou_reward": 0.953286349773407,
      "rewards/forensics_sinkhorn_reward": 0.8116621375083923,
      "rewards/forensics_strict_edge_reward": 0.0583789087831974,
      "rewards/forensics_tiou_reward": 0.6729951202869415,
      "step": 1286
    },
    {
      "completion_length": 27.03125,
      "epoch": 3.9493087557603688,
      "grad_norm": 8.30078125,
      "kl": 0.501953125,
      "learning_rate": 1.0769230769230768e-08,
      "loss": 0.166,
      "reward": 0.7428398728370667,
      "reward_std": 0.055452555418014526,
      "reward_std_all": 0.061040109023451805,
      "rewards-max/forensics_sinkhorn_reward": 0.9715654850006104,
      "rewards-max/forensics_strict_edge_reward": 0.5529413521289825,
      "rewards-max/forensics_tiou_reward": 0.9573107957839966,
      "rewards/forensics_sinkhorn_reward": 0.7179244160652161,
      "rewards/forensics_strict_edge_reward": 0.06383643858134747,
      "rewards/forensics_tiou_reward": 0.5262847542762756,
      "step": 1287
    },
    {
      "completion_length": 27.328125,
      "epoch": 3.9523809523809526,
      "grad_norm": 8.744545936584473,
      "kl": 0.486328125,
      "learning_rate": 1e-08,
      "loss": -0.0049,
      "reward": 0.7808689475059509,
      "reward_std": 0.041449847631156445,
      "reward_std_all": 0.03790230676531792,
      "rewards-max/forensics_sinkhorn_reward": 0.974670946598053,
      "rewards-max/forensics_strict_edge_reward": 0.3180827274918556,
      "rewards-max/forensics_tiou_reward": 0.9563027918338776,
      "rewards/forensics_sinkhorn_reward": 0.7833943068981171,
      "rewards/forensics_strict_edge_reward": 0.0234648953191936,
      "rewards/forensics_tiou_reward": 0.6191657781600952,
      "step": 1288
    },
    {
      "completion_length": 23.953125,
      "epoch": 3.955453149001536,
      "grad_norm": 11.697248458862305,
      "kl": 0.6240234375,
      "learning_rate": 9.23076923076923e-09,
      "loss": 0.0014,
      "reward": 0.7754030227661133,
      "reward_std": 0.0346922492608428,
      "reward_std_all": 0.03437164332717657,
      "rewards-max/forensics_sinkhorn_reward": 0.9849238395690918,
      "rewards-max/forensics_strict_edge_reward": 0.5192448049783707,
      "rewards-max/forensics_tiou_reward": 0.9870820045471191,
      "rewards/forensics_sinkhorn_reward": 0.7596973776817322,
      "rewards/forensics_strict_edge_reward": 0.08018911071121693,
      "rewards/forensics_tiou_reward": 0.6130022406578064,
      "step": 1289
    },
    {
      "completion_length": 24.25,
      "epoch": 3.9585253456221197,
      "grad_norm": 44.39704513549805,
      "kl": 2.3798828125,
      "learning_rate": 8.46153846153846e-09,
      "loss": 0.1878,
      "reward": 0.7676078677177429,
      "reward_std": 0.03785766940563917,
      "reward_std_all": 0.04175455495715141,
      "rewards-max/forensics_sinkhorn_reward": 0.9713242650032043,
      "rewards-max/forensics_strict_edge_reward": 0.4900742173194885,
      "rewards-max/forensics_tiou_reward": 0.9600403904914856,
      "rewards/forensics_sinkhorn_reward": 0.7442829012870789,
      "rewards/forensics_strict_edge_reward": 0.05260126292705536,
      "rewards/forensics_tiou_reward": 0.576500415802002,
      "step": 1290
    },
    {
      "completion_length": 26.96875,
      "epoch": 3.9615975422427034,
      "grad_norm": 9.224637985229492,
      "kl": 0.3544921875,
      "learning_rate": 7.692307692307693e-09,
      "loss": -0.0186,
      "reward": 0.7744863033294678,
      "reward_std": 0.03956282418221235,
      "reward_std_all": 0.03802250977605581,
      "rewards-max/forensics_sinkhorn_reward": 0.9850266873836517,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9889473617076874,
      "rewards/forensics_sinkhorn_reward": 0.7507464289665222,
      "rewards/forensics_strict_edge_reward": 0.12514517828822136,
      "rewards/forensics_tiou_reward": 0.6259333491325378,
      "step": 1291
    },
    {
      "completion_length": 23.734375,
      "epoch": 3.964669738863287,
      "grad_norm": 11.696935653686523,
      "kl": 0.6044921875,
      "learning_rate": 6.923076923076923e-09,
      "loss": 0.0027,
      "reward": 0.7824451923370361,
      "reward_std": 0.04101552069187164,
      "reward_std_all": 0.03742731921374798,
      "rewards-max/forensics_sinkhorn_reward": 0.9843398928642273,
      "rewards-max/forensics_strict_edge_reward": 0.6079998314380646,
      "rewards-max/forensics_tiou_reward": 0.9873385429382324,
      "rewards/forensics_sinkhorn_reward": 0.7891886234283447,
      "rewards/forensics_strict_edge_reward": 0.07560529187321663,
      "rewards/forensics_tiou_reward": 0.6292272210121155,
      "step": 1292
    },
    {
      "completion_length": 24.03125,
      "epoch": 3.967741935483871,
      "grad_norm": 7.4593400955200195,
      "kl": 0.3671875,
      "learning_rate": 6.153846153846154e-09,
      "loss": -0.0887,
      "reward": 0.7783180475234985,
      "reward_std": 0.033957738312892616,
      "reward_std_all": 0.03351995116099715,
      "rewards-max/forensics_sinkhorn_reward": 0.9876517057418823,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9959756135940552,
      "rewards/forensics_sinkhorn_reward": 0.7974980473518372,
      "rewards/forensics_strict_edge_reward": 0.101295106112957,
      "rewards/forensics_tiou_reward": 0.6617343723773956,
      "step": 1293
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.9708141321044548,
      "grad_norm": 9.964742660522461,
      "kl": 0.369140625,
      "learning_rate": 5.384615384615384e-09,
      "loss": -0.0097,
      "reward": 0.7898406088352203,
      "reward_std": 0.01079727231990546,
      "reward_std_all": 0.014261070173233747,
      "rewards-max/forensics_sinkhorn_reward": 0.9781868159770966,
      "rewards-max/forensics_strict_edge_reward": 0.5090594589710236,
      "rewards-max/forensics_tiou_reward": 0.9698370099067688,
      "rewards/forensics_sinkhorn_reward": 0.8222500681877136,
      "rewards/forensics_strict_edge_reward": 0.07568610832095146,
      "rewards/forensics_tiou_reward": 0.6814775466918945,
      "step": 1294
    },
    {
      "completion_length": 22.953125,
      "epoch": 3.9738863287250386,
      "grad_norm": 8.001800537109375,
      "kl": 0.548828125,
      "learning_rate": 4.615384615384615e-09,
      "loss": 0.031,
      "reward": 0.7874009013175964,
      "reward_std": 0.03346974216401577,
      "reward_std_all": 0.030509180389344692,
      "rewards-max/forensics_sinkhorn_reward": 0.9869449138641357,
      "rewards-max/forensics_strict_edge_reward": 0.742465078830719,
      "rewards-max/forensics_tiou_reward": 0.9944006204605103,
      "rewards/forensics_sinkhorn_reward": 0.8167812824249268,
      "rewards/forensics_strict_edge_reward": 0.09653853625059128,
      "rewards/forensics_tiou_reward": 0.6742887496948242,
      "step": 1295
    },
    {
      "completion_length": 24.0625,
      "epoch": 3.976958525345622,
      "grad_norm": 10.73339557647705,
      "kl": 0.412109375,
      "learning_rate": 3.846153846153846e-09,
      "loss": -0.052,
      "reward": 0.7906866371631622,
      "reward_std": 0.024632351472973824,
      "reward_std_all": 0.02367506269365549,
      "rewards-max/forensics_sinkhorn_reward": 0.9846073389053345,
      "rewards-max/forensics_strict_edge_reward": 0.5612282156944275,
      "rewards-max/forensics_tiou_reward": 0.9873645007610321,
      "rewards/forensics_sinkhorn_reward": 0.7793530225753784,
      "rewards/forensics_strict_edge_reward": 0.07837698375806212,
      "rewards/forensics_tiou_reward": 0.577019214630127,
      "step": 1296
    },
    {
      "completion_length": 27.625,
      "epoch": 3.980030721966206,
      "grad_norm": 7.758701801300049,
      "kl": 0.70703125,
      "learning_rate": 3.076923076923077e-09,
      "loss": 0.0271,
      "reward": 0.7797825038433075,
      "reward_std": 0.028150424361228943,
      "reward_std_all": 0.02867866586893797,
      "rewards-max/forensics_sinkhorn_reward": 0.9743399918079376,
      "rewards-max/forensics_strict_edge_reward": 0.7234414666891098,
      "rewards-max/forensics_tiou_reward": 0.9625023007392883,
      "rewards/forensics_sinkhorn_reward": 0.7758975028991699,
      "rewards/forensics_strict_edge_reward": 0.05606902576982975,
      "rewards/forensics_tiou_reward": 0.6055862307548523,
      "step": 1297
    },
    {
      "completion_length": 23.734375,
      "epoch": 3.9831029185867894,
      "grad_norm": 8.032296180725098,
      "kl": 0.3935546875,
      "learning_rate": 2.3076923076923076e-09,
      "loss": -0.0286,
      "reward": 0.7863436937332153,
      "reward_std": 0.03452318673953414,
      "reward_std_all": 0.032021467573940754,
      "rewards-max/forensics_sinkhorn_reward": 0.9817761182785034,
      "rewards-max/forensics_strict_edge_reward": 0.5972513854503632,
      "rewards-max/forensics_tiou_reward": 0.9827597439289093,
      "rewards/forensics_sinkhorn_reward": 0.790315717458725,
      "rewards/forensics_strict_edge_reward": 0.07731349393725395,
      "rewards/forensics_tiou_reward": 0.6672176420688629,
      "step": 1298
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.986175115207373,
      "grad_norm": 8.598669052124023,
      "kl": 0.4921875,
      "learning_rate": 1.5384615384615385e-09,
      "loss": 0.095,
      "reward": 0.7799384593963623,
      "reward_std": 0.04733921028673649,
      "reward_std_all": 0.042524101212620735,
      "rewards-max/forensics_sinkhorn_reward": 0.9889691770076752,
      "rewards-max/forensics_strict_edge_reward": 0.9836080372333527,
      "rewards-max/forensics_tiou_reward": 0.9999292194843292,
      "rewards/forensics_sinkhorn_reward": 0.82268026471138,
      "rewards/forensics_strict_edge_reward": 0.10904386639595032,
      "rewards/forensics_tiou_reward": 0.7208040654659271,
      "step": 1299
    },
    {
      "completion_length": 23.171875,
      "epoch": 3.989247311827957,
      "grad_norm": 8.225883483886719,
      "kl": 0.4365234375,
      "learning_rate": 7.692307692307692e-10,
      "loss": 0.0174,
      "reward": 0.7662175893783569,
      "reward_std": 0.039202967658638954,
      "reward_std_all": 0.04052509926259518,
      "rewards-max/forensics_sinkhorn_reward": 0.9868296384811401,
      "rewards-max/forensics_strict_edge_reward": 0.8443589508533478,
      "rewards-max/forensics_tiou_reward": 0.9943958222866058,
      "rewards/forensics_sinkhorn_reward": 0.7885161638259888,
      "rewards/forensics_strict_edge_reward": 0.0953594446182251,
      "rewards/forensics_tiou_reward": 0.6223123967647552,
      "step": 1300
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 1300,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 1300,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}