{
  "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": 25.46875,
      "epoch": 0.0030721966205837174,
      "grad_norm": 15.347468376159668,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": 0.1301,
      "reward": 0.5321594476699829,
      "reward_std": 0.3012385368347168,
      "reward_std_all": 0.28471407294273376,
      "rewards-max/forensics_sinkhorn_reward": 0.9012109935283661,
      "rewards-max/forensics_strict_edge_reward": 0.011221585165912984,
      "rewards-max/forensics_tiou_reward": 0.7268121540546417,
      "rewards/forensics_sinkhorn_reward": 0.44327498972415924,
      "rewards/forensics_strict_edge_reward": 0.000849602663265614,
      "rewards/forensics_tiou_reward": 0.17821162939071655,
      "step": 1
    },
    {
      "completion_length": 24.25,
      "epoch": 0.006144393241167435,
      "grad_norm": 4.461289882659912,
      "kl": 0.000232696533203125,
      "learning_rate": 9.992307692307693e-07,
      "loss": 0.1465,
      "reward": 0.5799558460712433,
      "reward_std": 0.19386174529790878,
      "reward_std_all": 0.20078768581151962,
      "rewards-max/forensics_sinkhorn_reward": 0.9261378943920135,
      "rewards-max/forensics_strict_edge_reward": 0.032440694980323315,
      "rewards-max/forensics_tiou_reward": 0.8120884895324707,
      "rewards/forensics_sinkhorn_reward": 0.4781336933374405,
      "rewards/forensics_strict_edge_reward": 0.001777687284629792,
      "rewards/forensics_tiou_reward": 0.21776944398880005,
      "step": 2
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.009216589861751152,
      "grad_norm": 5.804114818572998,
      "kl": 0.0008544921875,
      "learning_rate": 9.984615384615384e-07,
      "loss": 0.063,
      "reward": 0.5909421443939209,
      "reward_std": 0.20651352405548096,
      "reward_std_all": 0.20784511417150497,
      "rewards-max/forensics_sinkhorn_reward": 0.8679223954677582,
      "rewards-max/forensics_strict_edge_reward": 0.08590756636112928,
      "rewards-max/forensics_tiou_reward": 0.7243439853191376,
      "rewards/forensics_sinkhorn_reward": 0.4692036956548691,
      "rewards/forensics_strict_edge_reward": 0.0037584786186926067,
      "rewards/forensics_tiou_reward": 0.18137888610363007,
      "step": 3
    },
    {
      "completion_length": 23.984375,
      "epoch": 0.01228878648233487,
      "grad_norm": 5.528610706329346,
      "kl": 0.0021209716796875,
      "learning_rate": 9.976923076923077e-07,
      "loss": 0.154,
      "reward": 0.6152364611625671,
      "reward_std": 0.18409816920757294,
      "reward_std_all": 0.183795727789402,
      "rewards-max/forensics_sinkhorn_reward": 0.8929429352283478,
      "rewards-max/forensics_strict_edge_reward": 0.016381994355469942,
      "rewards-max/forensics_tiou_reward": 0.7406289577484131,
      "rewards/forensics_sinkhorn_reward": 0.5229470729827881,
      "rewards/forensics_strict_edge_reward": 0.0010250777704641223,
      "rewards/forensics_tiou_reward": 0.22659553587436676,
      "step": 4
    },
    {
      "completion_length": 21.40625,
      "epoch": 0.015360983102918587,
      "grad_norm": 5.505784511566162,
      "kl": 0.009552001953125,
      "learning_rate": 9.969230769230768e-07,
      "loss": 0.1813,
      "reward": 0.5578969419002533,
      "reward_std": 0.19853424280881882,
      "reward_std_all": 0.21776363253593445,
      "rewards-max/forensics_sinkhorn_reward": 0.8949594497680664,
      "rewards-max/forensics_strict_edge_reward": 0.08894250530283898,
      "rewards-max/forensics_tiou_reward": 0.7374344170093536,
      "rewards/forensics_sinkhorn_reward": 0.46008576452732086,
      "rewards/forensics_strict_edge_reward": 0.007406071090372279,
      "rewards/forensics_tiou_reward": 0.15482409298419952,
      "step": 5
    },
    {
      "completion_length": 25.703125,
      "epoch": 0.018433179723502304,
      "grad_norm": 6.113433837890625,
      "kl": 0.016632080078125,
      "learning_rate": 9.961538461538461e-07,
      "loss": 0.0722,
      "reward": 0.6195499002933502,
      "reward_std": 0.18410330265760422,
      "reward_std_all": 0.17999836802482605,
      "rewards-max/forensics_sinkhorn_reward": 0.8808777034282684,
      "rewards-max/forensics_strict_edge_reward": 0.21412493474781513,
      "rewards-max/forensics_tiou_reward": 0.8233838975429535,
      "rewards/forensics_sinkhorn_reward": 0.4669080674648285,
      "rewards/forensics_strict_edge_reward": 0.008955805038567632,
      "rewards/forensics_tiou_reward": 0.2158643677830696,
      "step": 6
    },
    {
      "completion_length": 23.640625,
      "epoch": 0.021505376344086023,
      "grad_norm": 5.0991926193237305,
      "kl": 0.024658203125,
      "learning_rate": 9.953846153846154e-07,
      "loss": 0.2067,
      "reward": 0.6048990488052368,
      "reward_std": 0.19542238116264343,
      "reward_std_all": 0.19114382565021515,
      "rewards-max/forensics_sinkhorn_reward": 0.7801648378372192,
      "rewards-max/forensics_strict_edge_reward": 0.23885848622921912,
      "rewards-max/forensics_tiou_reward": 0.610019862651825,
      "rewards/forensics_sinkhorn_reward": 0.4822639226913452,
      "rewards/forensics_strict_edge_reward": 0.013634517464367946,
      "rewards/forensics_tiou_reward": 0.1742360144853592,
      "step": 7
    },
    {
      "completion_length": 24.09375,
      "epoch": 0.02457757296466974,
      "grad_norm": 6.411648750305176,
      "kl": 0.04803466796875,
      "learning_rate": 9.946153846153846e-07,
      "loss": 0.073,
      "reward": 0.5479959547519684,
      "reward_std": 0.18894954770803452,
      "reward_std_all": 0.2174968272447586,
      "rewards-max/forensics_sinkhorn_reward": 0.8723973631858826,
      "rewards-max/forensics_strict_edge_reward": 0.020868475490715355,
      "rewards-max/forensics_tiou_reward": 0.6973647177219391,
      "rewards/forensics_sinkhorn_reward": 0.4341670274734497,
      "rewards/forensics_strict_edge_reward": 0.0008787783353909617,
      "rewards/forensics_tiou_reward": 0.1627625972032547,
      "step": 8
    },
    {
      "completion_length": 24.015625,
      "epoch": 0.027649769585253458,
      "grad_norm": 5.246419906616211,
      "kl": 0.0531005859375,
      "learning_rate": 9.938461538461539e-07,
      "loss": 0.2098,
      "reward": 0.6323288083076477,
      "reward_std": 0.15681426972150803,
      "reward_std_all": 0.15937019139528275,
      "rewards-max/forensics_sinkhorn_reward": 0.9431183934211731,
      "rewards-max/forensics_strict_edge_reward": 0.048581810646282975,
      "rewards-max/forensics_tiou_reward": 0.860964447259903,
      "rewards/forensics_sinkhorn_reward": 0.49806034564971924,
      "rewards/forensics_strict_edge_reward": 0.002078552953662438,
      "rewards/forensics_tiou_reward": 0.18466073274612427,
      "step": 9
    },
    {
      "completion_length": 22.75,
      "epoch": 0.030721966205837174,
      "grad_norm": 4.147330284118652,
      "kl": 0.11328125,
      "learning_rate": 9.93076923076923e-07,
      "loss": 0.1547,
      "reward": 0.664477676153183,
      "reward_std": 0.14819196984171867,
      "reward_std_all": 0.1433270275592804,
      "rewards-max/forensics_sinkhorn_reward": 0.974990725517273,
      "rewards-max/forensics_strict_edge_reward": 0.3304598890244961,
      "rewards-max/forensics_tiou_reward": 0.9316145777702332,
      "rewards/forensics_sinkhorn_reward": 0.5513027310371399,
      "rewards/forensics_strict_edge_reward": 0.014180634170770645,
      "rewards/forensics_tiou_reward": 0.21869638562202454,
      "step": 10
    },
    {
      "completion_length": 24.796875,
      "epoch": 0.03379416282642089,
      "grad_norm": 4.121086120605469,
      "kl": 0.14990234375,
      "learning_rate": 9.923076923076923e-07,
      "loss": 0.2722,
      "reward": 0.6312294006347656,
      "reward_std": 0.1359262689948082,
      "reward_std_all": 0.15061113983392715,
      "rewards-max/forensics_sinkhorn_reward": 0.9322721362113953,
      "rewards-max/forensics_strict_edge_reward": 0.19710708409547806,
      "rewards-max/forensics_tiou_reward": 0.8142786026000977,
      "rewards/forensics_sinkhorn_reward": 0.5293104350566864,
      "rewards/forensics_strict_edge_reward": 0.009823803789913654,
      "rewards/forensics_tiou_reward": 0.23669850081205368,
      "step": 11
    },
    {
      "completion_length": 27.453125,
      "epoch": 0.03686635944700461,
      "grad_norm": 4.977242946624756,
      "kl": 0.18212890625,
      "learning_rate": 9.915384615384616e-07,
      "loss": 0.257,
      "reward": 0.6025407910346985,
      "reward_std": 0.13309375196695328,
      "reward_std_all": 0.15950025618076324,
      "rewards-max/forensics_sinkhorn_reward": 0.8507182598114014,
      "rewards-max/forensics_strict_edge_reward": 0.09616908174939454,
      "rewards-max/forensics_tiou_reward": 0.6099935472011566,
      "rewards/forensics_sinkhorn_reward": 0.47487248480319977,
      "rewards/forensics_strict_edge_reward": 0.004007553943665698,
      "rewards/forensics_tiou_reward": 0.16198722645640373,
      "step": 12
    },
    {
      "completion_length": 25.96875,
      "epoch": 0.039938556067588324,
      "grad_norm": 6.263796806335449,
      "kl": 0.16015625,
      "learning_rate": 9.907692307692307e-07,
      "loss": 0.0639,
      "reward": 0.6124232113361359,
      "reward_std": 0.21073048561811447,
      "reward_std_all": 0.20109017193317413,
      "rewards-max/forensics_sinkhorn_reward": 0.9345386624336243,
      "rewards-max/forensics_strict_edge_reward": 0.2578616881510243,
      "rewards-max/forensics_tiou_reward": 0.821893572807312,
      "rewards/forensics_sinkhorn_reward": 0.4967418909072876,
      "rewards/forensics_strict_edge_reward": 0.011387760307115968,
      "rewards/forensics_tiou_reward": 0.20321230590343475,
      "step": 13
    },
    {
      "completion_length": 28.890625,
      "epoch": 0.043010752688172046,
      "grad_norm": 4.994441032409668,
      "kl": 0.16064453125,
      "learning_rate": 9.9e-07,
      "loss": 0.1238,
      "reward": 0.6265738904476166,
      "reward_std": 0.17253193259239197,
      "reward_std_all": 0.17325910180807114,
      "rewards-max/forensics_sinkhorn_reward": 0.9456848502159119,
      "rewards-max/forensics_strict_edge_reward": 0.020381283028882535,
      "rewards-max/forensics_tiou_reward": 0.8350750505924225,
      "rewards/forensics_sinkhorn_reward": 0.4989750385284424,
      "rewards/forensics_strict_edge_reward": 0.0018117730716982194,
      "rewards/forensics_tiou_reward": 0.2392309531569481,
      "step": 14
    },
    {
      "completion_length": 26.359375,
      "epoch": 0.04608294930875576,
      "grad_norm": 6.938230991363525,
      "kl": 0.205078125,
      "learning_rate": 9.892307692307692e-07,
      "loss": 0.1298,
      "reward": 0.6658391356468201,
      "reward_std": 0.13502147048711777,
      "reward_std_all": 0.1410532146692276,
      "rewards-max/forensics_sinkhorn_reward": 0.9178393483161926,
      "rewards-max/forensics_strict_edge_reward": 0.19193802299560048,
      "rewards-max/forensics_tiou_reward": 0.7894963324069977,
      "rewards/forensics_sinkhorn_reward": 0.5612374842166901,
      "rewards/forensics_strict_edge_reward": 0.008331788623763714,
      "rewards/forensics_tiou_reward": 0.2578892856836319,
      "step": 15
    },
    {
      "completion_length": 24.921875,
      "epoch": 0.04915514592933948,
      "grad_norm": 5.029412746429443,
      "kl": 0.177734375,
      "learning_rate": 9.884615384615385e-07,
      "loss": 0.2386,
      "reward": 0.6694287657737732,
      "reward_std": 0.13935699313879013,
      "reward_std_all": 0.1351357027888298,
      "rewards-max/forensics_sinkhorn_reward": 0.9541502594947815,
      "rewards-max/forensics_strict_edge_reward": 0.2218532245606184,
      "rewards-max/forensics_tiou_reward": 0.8786501884460449,
      "rewards/forensics_sinkhorn_reward": 0.5449525713920593,
      "rewards/forensics_strict_edge_reward": 0.009564209263771772,
      "rewards/forensics_tiou_reward": 0.23034479469060898,
      "step": 16
    },
    {
      "completion_length": 26.34375,
      "epoch": 0.05222734254992319,
      "grad_norm": 4.147587776184082,
      "kl": 0.1865234375,
      "learning_rate": 9.876923076923076e-07,
      "loss": 0.1388,
      "reward": 0.6366811990737915,
      "reward_std": 0.14843393117189407,
      "reward_std_all": 0.1520770937204361,
      "rewards-max/forensics_sinkhorn_reward": 0.7942995727062225,
      "rewards-max/forensics_strict_edge_reward": 0.001121620732405404,
      "rewards-max/forensics_tiou_reward": 0.4698890596628189,
      "rewards/forensics_sinkhorn_reward": 0.5018395185470581,
      "rewards/forensics_strict_edge_reward": 5.0004096382672e-05,
      "rewards/forensics_tiou_reward": 0.15243875235319138,
      "step": 17
    },
    {
      "completion_length": 22.046875,
      "epoch": 0.055299539170506916,
      "grad_norm": 3.833613872528076,
      "kl": 0.171875,
      "learning_rate": 9.86923076923077e-07,
      "loss": 0.1437,
      "reward": 0.6860875487327576,
      "reward_std": 0.13706158101558685,
      "reward_std_all": 0.12970934063196182,
      "rewards-max/forensics_sinkhorn_reward": 0.9284997284412384,
      "rewards-max/forensics_strict_edge_reward": 0.006789884238969535,
      "rewards-max/forensics_tiou_reward": 0.8086600303649902,
      "rewards/forensics_sinkhorn_reward": 0.5854785442352295,
      "rewards/forensics_strict_edge_reward": 0.0002968397348013241,
      "rewards/forensics_tiou_reward": 0.2661939561367035,
      "step": 18
    },
    {
      "completion_length": 21.640625,
      "epoch": 0.05837173579109063,
      "grad_norm": 4.520503044128418,
      "kl": 0.20654296875,
      "learning_rate": 9.86153846153846e-07,
      "loss": 0.1867,
      "reward": 0.6524480581283569,
      "reward_std": 0.10924489796161652,
      "reward_std_all": 0.12764858454465866,
      "rewards-max/forensics_sinkhorn_reward": 0.9039958715438843,
      "rewards-max/forensics_strict_edge_reward": 0.0030994111439213157,
      "rewards-max/forensics_tiou_reward": 0.7379671633243561,
      "rewards/forensics_sinkhorn_reward": 0.5591803789138794,
      "rewards/forensics_strict_edge_reward": 0.00022152256133267656,
      "rewards/forensics_tiou_reward": 0.20863716304302216,
      "step": 19
    },
    {
      "completion_length": 21.796875,
      "epoch": 0.06144393241167435,
      "grad_norm": 4.767131805419922,
      "kl": 0.18603515625,
      "learning_rate": 9.853846153846153e-07,
      "loss": 0.1335,
      "reward": 0.7120663225650787,
      "reward_std": 0.08852923288941383,
      "reward_std_all": 0.09033717587590218,
      "rewards-max/forensics_sinkhorn_reward": 0.9665599167346954,
      "rewards-max/forensics_strict_edge_reward": 0.11534658821262712,
      "rewards-max/forensics_tiou_reward": 0.8948699533939362,
      "rewards/forensics_sinkhorn_reward": 0.6256314516067505,
      "rewards/forensics_strict_edge_reward": 0.004809049890206255,
      "rewards/forensics_tiou_reward": 0.25778525322675705,
      "step": 20
    },
    {
      "completion_length": 21.28125,
      "epoch": 0.06451612903225806,
      "grad_norm": 4.5374674797058105,
      "kl": 0.24365234375,
      "learning_rate": 9.846153846153847e-07,
      "loss": -0.0086,
      "reward": 0.6976388990879059,
      "reward_std": 0.08607712015509605,
      "reward_std_all": 0.09707223623991013,
      "rewards-max/forensics_sinkhorn_reward": 0.9498055577278137,
      "rewards-max/forensics_strict_edge_reward": 0.02188235055655241,
      "rewards-max/forensics_tiou_reward": 0.8547166585922241,
      "rewards/forensics_sinkhorn_reward": 0.6150555908679962,
      "rewards/forensics_strict_edge_reward": 0.000941067497478798,
      "rewards/forensics_tiou_reward": 0.2451043799519539,
      "step": 21
    },
    {
      "completion_length": 21.828125,
      "epoch": 0.06758832565284179,
      "grad_norm": 5.260393142700195,
      "kl": 0.1826171875,
      "learning_rate": 9.838461538461538e-07,
      "loss": 0.1988,
      "reward": 0.7086692750453949,
      "reward_std": 0.09423800185322762,
      "reward_std_all": 0.09576109051704407,
      "rewards-max/forensics_sinkhorn_reward": 0.9115630686283112,
      "rewards-max/forensics_strict_edge_reward": 0.31350379809737206,
      "rewards-max/forensics_tiou_reward": 0.7841100990772247,
      "rewards/forensics_sinkhorn_reward": 0.6030339300632477,
      "rewards/forensics_strict_edge_reward": 0.013278956932481378,
      "rewards/forensics_tiou_reward": 0.2592284381389618,
      "step": 22
    },
    {
      "completion_length": 21.0625,
      "epoch": 0.0706605222734255,
      "grad_norm": 4.581875324249268,
      "kl": 0.23779296875,
      "learning_rate": 9.83076923076923e-07,
      "loss": 0.1143,
      "reward": 0.7386639416217804,
      "reward_std": 0.061016401275992393,
      "reward_std_all": 0.06575866043567657,
      "rewards-max/forensics_sinkhorn_reward": 0.9358212053775787,
      "rewards-max/forensics_strict_edge_reward": 0.0022222248080652207,
      "rewards-max/forensics_tiou_reward": 0.8375610411167145,
      "rewards/forensics_sinkhorn_reward": 0.6542277932167053,
      "rewards/forensics_strict_edge_reward": 0.00011199372238479555,
      "rewards/forensics_tiou_reward": 0.2845210134983063,
      "step": 23
    },
    {
      "completion_length": 22.65625,
      "epoch": 0.07373271889400922,
      "grad_norm": 5.669876575469971,
      "kl": 0.21484375,
      "learning_rate": 9.823076923076922e-07,
      "loss": 0.0877,
      "reward": 0.709322988986969,
      "reward_std": 0.09344140812754631,
      "reward_std_all": 0.09324096143245697,
      "rewards-max/forensics_sinkhorn_reward": 0.848249614238739,
      "rewards-max/forensics_strict_edge_reward": 0.0026692821803635525,
      "rewards-max/forensics_tiou_reward": 0.6036376655101776,
      "rewards/forensics_sinkhorn_reward": 0.5822441577911377,
      "rewards/forensics_strict_edge_reward": 0.00018463024155579433,
      "rewards/forensics_tiou_reward": 0.21692559123039246,
      "step": 24
    },
    {
      "completion_length": 22.546875,
      "epoch": 0.07680491551459294,
      "grad_norm": 4.167121410369873,
      "kl": 0.24609375,
      "learning_rate": 9.815384615384615e-07,
      "loss": 0.2471,
      "reward": 0.618078202009201,
      "reward_std": 0.13819832354784012,
      "reward_std_all": 0.1537420153617859,
      "rewards-max/forensics_sinkhorn_reward": 0.9370299279689789,
      "rewards-max/forensics_strict_edge_reward": 0.01727904785730061,
      "rewards-max/forensics_tiou_reward": 0.8386141657829285,
      "rewards/forensics_sinkhorn_reward": 0.5067234635353088,
      "rewards/forensics_strict_edge_reward": 0.0008052682160268887,
      "rewards/forensics_tiou_reward": 0.15822170674800873,
      "step": 25
    },
    {
      "completion_length": 23.671875,
      "epoch": 0.07987711213517665,
      "grad_norm": 4.477877616882324,
      "kl": 0.18603515625,
      "learning_rate": 9.807692307692306e-07,
      "loss": 0.0749,
      "reward": 0.6976008415222168,
      "reward_std": 0.08004984259605408,
      "reward_std_all": 0.09436658024787903,
      "rewards-max/forensics_sinkhorn_reward": 0.9561503231525421,
      "rewards-max/forensics_strict_edge_reward": 0.2378610223531723,
      "rewards-max/forensics_tiou_reward": 0.8868575096130371,
      "rewards/forensics_sinkhorn_reward": 0.6132515966892242,
      "rewards/forensics_strict_edge_reward": 0.010918715735897422,
      "rewards/forensics_tiou_reward": 0.28085868060588837,
      "step": 26
    },
    {
      "completion_length": 23.9375,
      "epoch": 0.08294930875576037,
      "grad_norm": 4.620000839233398,
      "kl": 0.2421875,
      "learning_rate": 9.8e-07,
      "loss": 0.2608,
      "reward": 0.7084905505180359,
      "reward_std": 0.07262243703007698,
      "reward_std_all": 0.08186757192015648,
      "rewards-max/forensics_sinkhorn_reward": 0.8895104229450226,
      "rewards-max/forensics_strict_edge_reward": 0.014267444144934416,
      "rewards-max/forensics_tiou_reward": 0.7302951514720917,
      "rewards/forensics_sinkhorn_reward": 0.5696227252483368,
      "rewards/forensics_strict_edge_reward": 0.0008797984628472477,
      "rewards/forensics_tiou_reward": 0.23118920624256134,
      "step": 27
    },
    {
      "completion_length": 22.453125,
      "epoch": 0.08602150537634409,
      "grad_norm": 3.8065035343170166,
      "kl": 0.2197265625,
      "learning_rate": 9.792307692307693e-07,
      "loss": 0.1629,
      "reward": 0.6839971244335175,
      "reward_std": 0.12483618035912514,
      "reward_std_all": 0.12142283841967583,
      "rewards-max/forensics_sinkhorn_reward": 0.900942325592041,
      "rewards-max/forensics_strict_edge_reward": 0.007524459089836455,
      "rewards-max/forensics_tiou_reward": 0.7501698732376099,
      "rewards/forensics_sinkhorn_reward": 0.5634796023368835,
      "rewards/forensics_strict_edge_reward": 0.0003584624945460746,
      "rewards/forensics_tiou_reward": 0.21479316800832748,
      "step": 28
    },
    {
      "completion_length": 20.234375,
      "epoch": 0.0890937019969278,
      "grad_norm": 4.596648216247559,
      "kl": 0.2568359375,
      "learning_rate": 9.784615384615384e-07,
      "loss": 0.0821,
      "reward": 0.6973553895950317,
      "reward_std": 0.09406367689371109,
      "reward_std_all": 0.09998373687267303,
      "rewards-max/forensics_sinkhorn_reward": 0.880192905664444,
      "rewards-max/forensics_strict_edge_reward": 0.015678658382967114,
      "rewards-max/forensics_tiou_reward": 0.7164858132600784,
      "rewards/forensics_sinkhorn_reward": 0.6115926504135132,
      "rewards/forensics_strict_edge_reward": 0.0006580039996606502,
      "rewards/forensics_tiou_reward": 0.24123786389827728,
      "step": 29
    },
    {
      "completion_length": 29.421875,
      "epoch": 0.09216589861751152,
      "grad_norm": 4.104807376861572,
      "kl": 0.20166015625,
      "learning_rate": 9.776923076923077e-07,
      "loss": 0.1598,
      "reward": 0.667086660861969,
      "reward_std": 0.11633536964654922,
      "reward_std_all": 0.12207329645752907,
      "rewards-max/forensics_sinkhorn_reward": 0.8823411464691162,
      "rewards-max/forensics_strict_edge_reward": 0.0014267278711486142,
      "rewards-max/forensics_tiou_reward": 0.6711299419403076,
      "rewards/forensics_sinkhorn_reward": 0.5225517898797989,
      "rewards/forensics_strict_edge_reward": 8.172656430360803e-05,
      "rewards/forensics_tiou_reward": 0.2521114945411682,
      "step": 30
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.09523809523809523,
      "grad_norm": 4.069824695587158,
      "kl": 0.21044921875,
      "learning_rate": 9.769230769230768e-07,
      "loss": 0.2447,
      "reward": 0.6754919588565826,
      "reward_std": 0.12176153063774109,
      "reward_std_all": 0.12374782934784889,
      "rewards-max/forensics_sinkhorn_reward": 0.873365044593811,
      "rewards-max/forensics_strict_edge_reward": 0.1340041309595108,
      "rewards-max/forensics_tiou_reward": 0.6924420148134232,
      "rewards/forensics_sinkhorn_reward": 0.5542837679386139,
      "rewards/forensics_strict_edge_reward": 0.01258716022130102,
      "rewards/forensics_tiou_reward": 0.2303716093301773,
      "step": 31
    },
    {
      "completion_length": 24.203125,
      "epoch": 0.09831029185867896,
      "grad_norm": 3.4910826683044434,
      "kl": 0.177734375,
      "learning_rate": 9.761538461538461e-07,
      "loss": 0.1787,
      "reward": 0.7193730175495148,
      "reward_std": 0.08374232053756714,
      "reward_std_all": 0.08438125625252724,
      "rewards-max/forensics_sinkhorn_reward": 0.9793916046619415,
      "rewards-max/forensics_strict_edge_reward": 0.185872670263052,
      "rewards-max/forensics_tiou_reward": 0.9442689418792725,
      "rewards/forensics_sinkhorn_reward": 0.6034879088401794,
      "rewards/forensics_strict_edge_reward": 0.008274932391941547,
      "rewards/forensics_tiou_reward": 0.2717587798833847,
      "step": 32
    },
    {
      "completion_length": 28.453125,
      "epoch": 0.10138248847926268,
      "grad_norm": 4.445133686065674,
      "kl": 0.17041015625,
      "learning_rate": 9.753846153846154e-07,
      "loss": 0.0831,
      "reward": 0.6833698749542236,
      "reward_std": 0.10489486157894135,
      "reward_std_all": 0.1082405224442482,
      "rewards-max/forensics_sinkhorn_reward": 0.9260045289993286,
      "rewards-max/forensics_strict_edge_reward": 0.07464364496991038,
      "rewards-max/forensics_tiou_reward": 0.8175894021987915,
      "rewards/forensics_sinkhorn_reward": 0.5669651329517365,
      "rewards/forensics_strict_edge_reward": 0.004610852280166,
      "rewards/forensics_tiou_reward": 0.2601548582315445,
      "step": 33
    },
    {
      "completion_length": 25.984375,
      "epoch": 0.10445468509984639,
      "grad_norm": 4.596963405609131,
      "kl": 0.1904296875,
      "learning_rate": 9.746153846153845e-07,
      "loss": 0.2088,
      "reward": 0.7155901491641998,
      "reward_std": 0.08117703348398209,
      "reward_std_all": 0.08337435498833656,
      "rewards-max/forensics_sinkhorn_reward": 0.9253369569778442,
      "rewards-max/forensics_strict_edge_reward": 0.09432075335644186,
      "rewards-max/forensics_tiou_reward": 0.772551953792572,
      "rewards/forensics_sinkhorn_reward": 0.6045091450214386,
      "rewards/forensics_strict_edge_reward": 0.004463442419364583,
      "rewards/forensics_tiou_reward": 0.27763091027736664,
      "step": 34
    },
    {
      "completion_length": 28.515625,
      "epoch": 0.10752688172043011,
      "grad_norm": 4.6559858322143555,
      "kl": 0.23828125,
      "learning_rate": 9.738461538461539e-07,
      "loss": 0.2142,
      "reward": 0.7140363454818726,
      "reward_std": 0.11314874514937401,
      "reward_std_all": 0.10566703230142593,
      "rewards-max/forensics_sinkhorn_reward": 0.8260165750980377,
      "rewards-max/forensics_strict_edge_reward": 0.014187191380187869,
      "rewards-max/forensics_tiou_reward": 0.603460431098938,
      "rewards/forensics_sinkhorn_reward": 0.5785817503929138,
      "rewards/forensics_strict_edge_reward": 0.000619408117927378,
      "rewards/forensics_tiou_reward": 0.2572922110557556,
      "step": 35
    },
    {
      "completion_length": 25.984375,
      "epoch": 0.11059907834101383,
      "grad_norm": 3.628636360168457,
      "kl": 0.20703125,
      "learning_rate": 9.73076923076923e-07,
      "loss": 0.2349,
      "reward": 0.6608977913856506,
      "reward_std": 0.10994819179177284,
      "reward_std_all": 0.12138861790299416,
      "rewards-max/forensics_sinkhorn_reward": 0.9395059049129486,
      "rewards-max/forensics_strict_edge_reward": 0.0123178674839437,
      "rewards-max/forensics_tiou_reward": 0.8185171484947205,
      "rewards/forensics_sinkhorn_reward": 0.535056084394455,
      "rewards/forensics_strict_edge_reward": 0.0005189900239201961,
      "rewards/forensics_tiou_reward": 0.25179316103458405,
      "step": 36
    },
    {
      "completion_length": 27.28125,
      "epoch": 0.11367127496159754,
      "grad_norm": 4.389624118804932,
      "kl": 0.2392578125,
      "learning_rate": 9.723076923076923e-07,
      "loss": 0.1516,
      "reward": 0.6203176081180573,
      "reward_std": 0.1635601744055748,
      "reward_std_all": 0.16844695061445236,
      "rewards-max/forensics_sinkhorn_reward": 0.9092599451541901,
      "rewards-max/forensics_strict_edge_reward": 0.06744563097890932,
      "rewards-max/forensics_tiou_reward": 0.7419902086257935,
      "rewards/forensics_sinkhorn_reward": 0.48201411962509155,
      "rewards/forensics_strict_edge_reward": 0.002818856332851283,
      "rewards/forensics_tiou_reward": 0.18779651075601578,
      "step": 37
    },
    {
      "completion_length": 24.9375,
      "epoch": 0.11674347158218126,
      "grad_norm": 5.054864883422852,
      "kl": 0.20263671875,
      "learning_rate": 9.715384615384616e-07,
      "loss": 0.319,
      "reward": 0.6382007002830505,
      "reward_std": 0.10327673703432083,
      "reward_std_all": 0.1349373161792755,
      "rewards-max/forensics_sinkhorn_reward": 0.9037927091121674,
      "rewards-max/forensics_strict_edge_reward": 0.011985161732809502,
      "rewards-max/forensics_tiou_reward": 0.7318490147590637,
      "rewards/forensics_sinkhorn_reward": 0.5062143206596375,
      "rewards/forensics_strict_edge_reward": 0.0007787051993091154,
      "rewards/forensics_tiou_reward": 0.25151100754737854,
      "step": 38
    },
    {
      "completion_length": 25.0,
      "epoch": 0.11981566820276497,
      "grad_norm": 6.527235984802246,
      "kl": 0.29345703125,
      "learning_rate": 9.707692307692307e-07,
      "loss": 0.0921,
      "reward": 0.6804450154304504,
      "reward_std": 0.0940854400396347,
      "reward_std_all": 0.1120898649096489,
      "rewards-max/forensics_sinkhorn_reward": 0.9769657254219055,
      "rewards-max/forensics_strict_edge_reward": 0.4544278457760811,
      "rewards-max/forensics_tiou_reward": 0.9454360902309418,
      "rewards/forensics_sinkhorn_reward": 0.5697269439697266,
      "rewards/forensics_strict_edge_reward": 0.021843782626092434,
      "rewards/forensics_tiou_reward": 0.30136288702487946,
      "step": 39
    },
    {
      "completion_length": 27.1875,
      "epoch": 0.1228878648233487,
      "grad_norm": 5.063241958618164,
      "kl": 0.23388671875,
      "learning_rate": 9.7e-07,
      "loss": 0.241,
      "reward": 0.6821983158588409,
      "reward_std": 0.10674767941236496,
      "reward_std_all": 0.11549137532711029,
      "rewards-max/forensics_sinkhorn_reward": 0.9631100296974182,
      "rewards-max/forensics_strict_edge_reward": 0.0074507202953100204,
      "rewards-max/forensics_tiou_reward": 0.8920581340789795,
      "rewards/forensics_sinkhorn_reward": 0.5906620919704437,
      "rewards/forensics_strict_edge_reward": 0.000618617283180356,
      "rewards/forensics_tiou_reward": 0.30324168503284454,
      "step": 40
    },
    {
      "completion_length": 28.078125,
      "epoch": 0.1259600614439324,
      "grad_norm": 4.656695365905762,
      "kl": 0.17529296875,
      "learning_rate": 9.692307692307691e-07,
      "loss": 0.1074,
      "reward": 0.6907217502593994,
      "reward_std": 0.11731478944420815,
      "reward_std_all": 0.11747030913829803,
      "rewards-max/forensics_sinkhorn_reward": 0.8780225515365601,
      "rewards-max/forensics_strict_edge_reward": 0.006422281665436458,
      "rewards-max/forensics_tiou_reward": 0.6986293792724609,
      "rewards/forensics_sinkhorn_reward": 0.5698294937610626,
      "rewards/forensics_strict_edge_reward": 0.0002723863389064718,
      "rewards/forensics_tiou_reward": 0.23813988268375397,
      "step": 41
    },
    {
      "completion_length": 23.3125,
      "epoch": 0.12903225806451613,
      "grad_norm": 4.820652008056641,
      "kl": 0.21533203125,
      "learning_rate": 9.684615384615385e-07,
      "loss": 0.3184,
      "reward": 0.7039644122123718,
      "reward_std": 0.09784587100148201,
      "reward_std_all": 0.10013766214251518,
      "rewards-max/forensics_sinkhorn_reward": 0.9325219690799713,
      "rewards-max/forensics_strict_edge_reward": 0.023314157966524363,
      "rewards-max/forensics_tiou_reward": 0.8177174627780914,
      "rewards/forensics_sinkhorn_reward": 0.602996438741684,
      "rewards/forensics_strict_edge_reward": 0.0013268761686049402,
      "rewards/forensics_tiou_reward": 0.2916017025709152,
      "step": 42
    },
    {
      "completion_length": 25.234375,
      "epoch": 0.13210445468509985,
      "grad_norm": 4.735954284667969,
      "kl": 0.19775390625,
      "learning_rate": 9.676923076923078e-07,
      "loss": 0.0412,
      "reward": 0.7063097357749939,
      "reward_std": 0.10438293218612671,
      "reward_std_all": 0.10275079309940338,
      "rewards-max/forensics_sinkhorn_reward": 0.8777894973754883,
      "rewards-max/forensics_strict_edge_reward": 0.03125722520053387,
      "rewards-max/forensics_tiou_reward": 0.7400456070899963,
      "rewards/forensics_sinkhorn_reward": 0.5870785713195801,
      "rewards/forensics_strict_edge_reward": 0.0016381153836846352,
      "rewards/forensics_tiou_reward": 0.2613539844751358,
      "step": 43
    },
    {
      "completion_length": 23.59375,
      "epoch": 0.13517665130568357,
      "grad_norm": 4.896188259124756,
      "kl": 0.18701171875,
      "learning_rate": 9.669230769230769e-07,
      "loss": 0.1848,
      "reward": 0.6723832190036774,
      "reward_std": 0.12374036759138107,
      "reward_std_all": 0.12632086500525475,
      "rewards-max/forensics_sinkhorn_reward": 0.8991865515708923,
      "rewards-max/forensics_strict_edge_reward": 0.2140682153403759,
      "rewards-max/forensics_tiou_reward": 0.7490293979644775,
      "rewards/forensics_sinkhorn_reward": 0.5726359784603119,
      "rewards/forensics_strict_edge_reward": 0.009102671756409109,
      "rewards/forensics_tiou_reward": 0.20444659888744354,
      "step": 44
    },
    {
      "completion_length": 22.03125,
      "epoch": 0.1382488479262673,
      "grad_norm": 3.950601100921631,
      "kl": 0.19580078125,
      "learning_rate": 9.66153846153846e-07,
      "loss": 0.1254,
      "reward": 0.7365326285362244,
      "reward_std": 0.08342438191175461,
      "reward_std_all": 0.0802253745496273,
      "rewards-max/forensics_sinkhorn_reward": 0.9595674872398376,
      "rewards-max/forensics_strict_edge_reward": 0.10006176913157105,
      "rewards-max/forensics_tiou_reward": 0.8781653046607971,
      "rewards/forensics_sinkhorn_reward": 0.6596432030200958,
      "rewards/forensics_strict_edge_reward": 0.007345247664488852,
      "rewards/forensics_tiou_reward": 0.3182956427335739,
      "step": 45
    },
    {
      "completion_length": 23.234375,
      "epoch": 0.141321044546851,
      "grad_norm": 4.19722318649292,
      "kl": 0.21142578125,
      "learning_rate": 9.653846153846153e-07,
      "loss": 0.1312,
      "reward": 0.6539564430713654,
      "reward_std": 0.1457461304962635,
      "reward_std_all": 0.14591841399669647,
      "rewards-max/forensics_sinkhorn_reward": 0.9071594178676605,
      "rewards-max/forensics_strict_edge_reward": 0.0037045578937977552,
      "rewards-max/forensics_tiou_reward": 0.7541809678077698,
      "rewards/forensics_sinkhorn_reward": 0.5571248531341553,
      "rewards/forensics_strict_edge_reward": 0.00018442699365550652,
      "rewards/forensics_tiou_reward": 0.24519389867782593,
      "step": 46
    },
    {
      "completion_length": 22.796875,
      "epoch": 0.1443932411674347,
      "grad_norm": 4.080806255340576,
      "kl": 0.18896484375,
      "learning_rate": 9.646153846153846e-07,
      "loss": 0.1834,
      "reward": 0.6560020446777344,
      "reward_std": 0.11703315004706383,
      "reward_std_all": 0.12545786798000336,
      "rewards-max/forensics_sinkhorn_reward": 0.9656578302383423,
      "rewards-max/forensics_strict_edge_reward": 0.127821956994012,
      "rewards-max/forensics_tiou_reward": 0.8928276598453522,
      "rewards/forensics_sinkhorn_reward": 0.5631842017173767,
      "rewards/forensics_strict_edge_reward": 0.006289867236773716,
      "rewards/forensics_tiou_reward": 0.23622394353151321,
      "step": 47
    },
    {
      "completion_length": 23.75,
      "epoch": 0.14746543778801843,
      "grad_norm": 3.18664813041687,
      "kl": 0.1611328125,
      "learning_rate": 9.638461538461537e-07,
      "loss": 0.1118,
      "reward": 0.7128113210201263,
      "reward_std": 0.10870461538434029,
      "reward_std_all": 0.1029229387640953,
      "rewards-max/forensics_sinkhorn_reward": 0.9228266179561615,
      "rewards-max/forensics_strict_edge_reward": 0.15920265018939972,
      "rewards-max/forensics_tiou_reward": 0.8016239106655121,
      "rewards/forensics_sinkhorn_reward": 0.5920148491859436,
      "rewards/forensics_strict_edge_reward": 0.007036355324089527,
      "rewards/forensics_tiou_reward": 0.2652731165289879,
      "step": 48
    },
    {
      "completion_length": 21.90625,
      "epoch": 0.15053763440860216,
      "grad_norm": 3.513411283493042,
      "kl": 0.15185546875,
      "learning_rate": 9.63076923076923e-07,
      "loss": 0.1905,
      "reward": 0.6904492676258087,
      "reward_std": 0.12153774499893188,
      "reward_std_all": 0.1171429455280304,
      "rewards-max/forensics_sinkhorn_reward": 0.9519280791282654,
      "rewards-max/forensics_strict_edge_reward": 0.07311294064857066,
      "rewards-max/forensics_tiou_reward": 0.8746467530727386,
      "rewards/forensics_sinkhorn_reward": 0.6081065535545349,
      "rewards/forensics_strict_edge_reward": 0.003625991492299363,
      "rewards/forensics_tiou_reward": 0.2568497210741043,
      "step": 49
    },
    {
      "completion_length": 23.5625,
      "epoch": 0.15360983102918588,
      "grad_norm": 4.5418548583984375,
      "kl": 0.13671875,
      "learning_rate": 9.623076923076922e-07,
      "loss": 0.0936,
      "reward": 0.6741245687007904,
      "reward_std": 0.10734398663043976,
      "reward_std_all": 0.11605984717607498,
      "rewards-max/forensics_sinkhorn_reward": 0.9219567477703094,
      "rewards-max/forensics_strict_edge_reward": 0.001968461598153226,
      "rewards-max/forensics_tiou_reward": 0.798466145992279,
      "rewards/forensics_sinkhorn_reward": 0.5720180571079254,
      "rewards/forensics_strict_edge_reward": 0.00015547491966572125,
      "rewards/forensics_tiou_reward": 0.24795904755592346,
      "step": 50
    },
    {
      "completion_length": 22.9375,
      "epoch": 0.15668202764976957,
      "grad_norm": 3.367051601409912,
      "kl": 0.19921875,
      "learning_rate": 9.615384615384615e-07,
      "loss": 0.1193,
      "reward": 0.6769155561923981,
      "reward_std": 0.08506779000163078,
      "reward_std_all": 0.09974125772714615,
      "rewards-max/forensics_sinkhorn_reward": 0.9526878297328949,
      "rewards-max/forensics_strict_edge_reward": 0.009158344701631904,
      "rewards-max/forensics_tiou_reward": 0.8708608150482178,
      "rewards/forensics_sinkhorn_reward": 0.5771538317203522,
      "rewards/forensics_strict_edge_reward": 0.0007430287395919777,
      "rewards/forensics_tiou_reward": 0.20386920869350433,
      "step": 51
    },
    {
      "completion_length": 22.328125,
      "epoch": 0.1597542242703533,
      "grad_norm": 4.641002178192139,
      "kl": 0.20947265625,
      "learning_rate": 9.607692307692308e-07,
      "loss": 0.2752,
      "reward": 0.6857975125312805,
      "reward_std": 0.0784035176038742,
      "reward_std_all": 0.09501734375953674,
      "rewards-max/forensics_sinkhorn_reward": 0.9475750029087067,
      "rewards-max/forensics_strict_edge_reward": 0.3547293581068516,
      "rewards-max/forensics_tiou_reward": 0.854004293680191,
      "rewards/forensics_sinkhorn_reward": 0.5818311870098114,
      "rewards/forensics_strict_edge_reward": 0.01484647300094366,
      "rewards/forensics_tiou_reward": 0.2379562109708786,
      "step": 52
    },
    {
      "completion_length": 25.234375,
      "epoch": 0.16282642089093702,
      "grad_norm": 4.973934173583984,
      "kl": 0.18017578125,
      "learning_rate": 9.6e-07,
      "loss": 0.2971,
      "reward": 0.6878113150596619,
      "reward_std": 0.10984786972403526,
      "reward_std_all": 0.11105304583907127,
      "rewards-max/forensics_sinkhorn_reward": 0.9624806344509125,
      "rewards-max/forensics_strict_edge_reward": 0.17526422999799252,
      "rewards-max/forensics_tiou_reward": 0.9074965119361877,
      "rewards/forensics_sinkhorn_reward": 0.5746654868125916,
      "rewards/forensics_strict_edge_reward": 0.008068921510130167,
      "rewards/forensics_tiou_reward": 0.2859899401664734,
      "step": 53
    },
    {
      "completion_length": 21.765625,
      "epoch": 0.16589861751152074,
      "grad_norm": 3.85918927192688,
      "kl": 0.22705078125,
      "learning_rate": 9.592307692307692e-07,
      "loss": 0.1325,
      "reward": 0.7200714647769928,
      "reward_std": 0.10432419180870056,
      "reward_std_all": 0.09841955080628395,
      "rewards-max/forensics_sinkhorn_reward": 0.9375846087932587,
      "rewards-max/forensics_strict_edge_reward": 0.16159459279151633,
      "rewards-max/forensics_tiou_reward": 0.8400345146656036,
      "rewards/forensics_sinkhorn_reward": 0.6405290961265564,
      "rewards/forensics_strict_edge_reward": 0.009108987065701513,
      "rewards/forensics_tiou_reward": 0.2854693531990051,
      "step": 54
    },
    {
      "completion_length": 26.34375,
      "epoch": 0.16897081413210446,
      "grad_norm": 3.8372039794921875,
      "kl": 0.1953125,
      "learning_rate": 9.584615384615383e-07,
      "loss": 0.1185,
      "reward": 0.6623416543006897,
      "reward_std": 0.1120363250374794,
      "reward_std_all": 0.12238143011927605,
      "rewards-max/forensics_sinkhorn_reward": 0.8961183130741119,
      "rewards-max/forensics_strict_edge_reward": 0.0020061505492776632,
      "rewards-max/forensics_tiou_reward": 0.7136631608009338,
      "rewards/forensics_sinkhorn_reward": 0.5257352888584137,
      "rewards/forensics_strict_edge_reward": 9.443236922379583e-05,
      "rewards/forensics_tiou_reward": 0.222500741481781,
      "step": 55
    },
    {
      "completion_length": 22.40625,
      "epoch": 0.17204301075268819,
      "grad_norm": 5.725866317749023,
      "kl": 0.22509765625,
      "learning_rate": 9.576923076923077e-07,
      "loss": 0.1493,
      "reward": 0.6793689131736755,
      "reward_std": 0.11882486194372177,
      "reward_std_all": 0.12034083902835846,
      "rewards-max/forensics_sinkhorn_reward": 0.9646992683410645,
      "rewards-max/forensics_strict_edge_reward": 0.19741079211235046,
      "rewards-max/forensics_tiou_reward": 0.9152944684028625,
      "rewards/forensics_sinkhorn_reward": 0.5749840140342712,
      "rewards/forensics_strict_edge_reward": 0.010631137993186712,
      "rewards/forensics_tiou_reward": 0.28461509943008423,
      "step": 56
    },
    {
      "completion_length": 22.828125,
      "epoch": 0.17511520737327188,
      "grad_norm": 4.4096174240112305,
      "kl": 0.203125,
      "learning_rate": 9.569230769230768e-07,
      "loss": 0.2412,
      "reward": 0.6741872131824493,
      "reward_std": 0.08610201254487038,
      "reward_std_all": 0.1045316644012928,
      "rewards-max/forensics_sinkhorn_reward": 0.9214798808097839,
      "rewards-max/forensics_strict_edge_reward": 0.1394176019821316,
      "rewards-max/forensics_tiou_reward": 0.8009498119354248,
      "rewards/forensics_sinkhorn_reward": 0.5926258862018585,
      "rewards/forensics_strict_edge_reward": 0.008393016250920482,
      "rewards/forensics_tiou_reward": 0.2600769102573395,
      "step": 57
    },
    {
      "completion_length": 22.640625,
      "epoch": 0.1781874039938556,
      "grad_norm": 3.778080940246582,
      "kl": 0.15966796875,
      "learning_rate": 9.56153846153846e-07,
      "loss": 0.1375,
      "reward": 0.7263222634792328,
      "reward_std": 0.08609312772750854,
      "reward_std_all": 0.08335037901997566,
      "rewards-max/forensics_sinkhorn_reward": 0.9575528800487518,
      "rewards-max/forensics_strict_edge_reward": 0.030603182647610083,
      "rewards-max/forensics_tiou_reward": 0.8928368985652924,
      "rewards/forensics_sinkhorn_reward": 0.6248072385787964,
      "rewards/forensics_strict_edge_reward": 0.0023621747459401377,
      "rewards/forensics_tiou_reward": 0.3132065683603287,
      "step": 58
    },
    {
      "completion_length": 24.5,
      "epoch": 0.18125960061443933,
      "grad_norm": 5.119016170501709,
      "kl": 0.19580078125,
      "learning_rate": 9.553846153846154e-07,
      "loss": 0.0532,
      "reward": 0.6235556602478027,
      "reward_std": 0.14059922099113464,
      "reward_std_all": 0.15429045259952545,
      "rewards-max/forensics_sinkhorn_reward": 0.9401635825634003,
      "rewards-max/forensics_strict_edge_reward": 0.35838585289457114,
      "rewards-max/forensics_tiou_reward": 0.8598277568817139,
      "rewards/forensics_sinkhorn_reward": 0.5099776983261108,
      "rewards/forensics_strict_edge_reward": 0.028593290200660704,
      "rewards/forensics_tiou_reward": 0.25255508720874786,
      "step": 59
    },
    {
      "completion_length": 25.859375,
      "epoch": 0.18433179723502305,
      "grad_norm": 5.370347499847412,
      "kl": 0.18701171875,
      "learning_rate": 9.546153846153845e-07,
      "loss": -0.0138,
      "reward": 0.6921555995941162,
      "reward_std": 0.11208362877368927,
      "reward_std_all": 0.11070001125335693,
      "rewards-max/forensics_sinkhorn_reward": 0.8775908648967743,
      "rewards-max/forensics_strict_edge_reward": 0.02211292553693056,
      "rewards-max/forensics_tiou_reward": 0.7015091180801392,
      "rewards/forensics_sinkhorn_reward": 0.5639707744121552,
      "rewards/forensics_strict_edge_reward": 0.0010771181259769946,
      "rewards/forensics_tiou_reward": 0.24623467773199081,
      "step": 60
    },
    {
      "completion_length": 25.5625,
      "epoch": 0.18740399385560677,
      "grad_norm": 3.2563533782958984,
      "kl": 0.19580078125,
      "learning_rate": 9.538461538461538e-07,
      "loss": 0.2472,
      "reward": 0.6388257443904877,
      "reward_std": 0.12843214720487595,
      "reward_std_all": 0.14076019078493118,
      "rewards-max/forensics_sinkhorn_reward": 0.859062135219574,
      "rewards-max/forensics_strict_edge_reward": 0.07233620030274324,
      "rewards-max/forensics_tiou_reward": 0.680600568652153,
      "rewards/forensics_sinkhorn_reward": 0.5045346915721893,
      "rewards/forensics_strict_edge_reward": 0.0030730008228090355,
      "rewards/forensics_tiou_reward": 0.18907269835472107,
      "step": 61
    },
    {
      "completion_length": 25.765625,
      "epoch": 0.19047619047619047,
      "grad_norm": 3.398639440536499,
      "kl": 0.2021484375,
      "learning_rate": 9.53076923076923e-07,
      "loss": 0.28,
      "reward": 0.679456889629364,
      "reward_std": 0.10049519315361977,
      "reward_std_all": 0.11017924547195435,
      "rewards-max/forensics_sinkhorn_reward": 0.9883681535720825,
      "rewards-max/forensics_strict_edge_reward": 0.2600954002700746,
      "rewards-max/forensics_tiou_reward": 0.9738560616970062,
      "rewards/forensics_sinkhorn_reward": 0.5715499818325043,
      "rewards/forensics_strict_edge_reward": 0.01173906197072938,
      "rewards/forensics_tiou_reward": 0.26571907103061676,
      "step": 62
    },
    {
      "completion_length": 23.015625,
      "epoch": 0.1935483870967742,
      "grad_norm": 3.502018928527832,
      "kl": 0.1875,
      "learning_rate": 9.523076923076923e-07,
      "loss": 0.0528,
      "reward": 0.6988321840763092,
      "reward_std": 0.1102040521800518,
      "reward_std_all": 0.11047248914837837,
      "rewards-max/forensics_sinkhorn_reward": 0.8961779773235321,
      "rewards-max/forensics_strict_edge_reward": 0.0347833746127435,
      "rewards-max/forensics_tiou_reward": 0.736899346113205,
      "rewards/forensics_sinkhorn_reward": 0.5701726675033569,
      "rewards/forensics_strict_edge_reward": 0.002494998878205479,
      "rewards/forensics_tiou_reward": 0.2450440227985382,
      "step": 63
    },
    {
      "completion_length": 24.921875,
      "epoch": 0.1966205837173579,
      "grad_norm": 3.7708899974823,
      "kl": 0.18603515625,
      "learning_rate": 9.515384615384615e-07,
      "loss": 0.1937,
      "reward": 0.7020441293716431,
      "reward_std": 0.11750103160738945,
      "reward_std_all": 0.11155735701322556,
      "rewards-max/forensics_sinkhorn_reward": 0.9628909230232239,
      "rewards-max/forensics_strict_edge_reward": 0.2722722962498665,
      "rewards-max/forensics_tiou_reward": 0.9032605886459351,
      "rewards/forensics_sinkhorn_reward": 0.608861893415451,
      "rewards/forensics_strict_edge_reward": 0.012760635232552886,
      "rewards/forensics_tiou_reward": 0.35145772993564606,
      "step": 64
    },
    {
      "completion_length": 24.578125,
      "epoch": 0.19969278033794163,
      "grad_norm": 4.151946544647217,
      "kl": 0.16845703125,
      "learning_rate": 9.507692307692308e-07,
      "loss": 0.1033,
      "reward": 0.7553640902042389,
      "reward_std": 0.07566024735569954,
      "reward_std_all": 0.06860055774450302,
      "rewards-max/forensics_sinkhorn_reward": 0.960526704788208,
      "rewards-max/forensics_strict_edge_reward": 0.0811405498534441,
      "rewards-max/forensics_tiou_reward": 0.8928759694099426,
      "rewards/forensics_sinkhorn_reward": 0.6847148835659027,
      "rewards/forensics_strict_edge_reward": 0.004932065261527896,
      "rewards/forensics_tiou_reward": 0.3883485496044159,
      "step": 65
    },
    {
      "completion_length": 25.03125,
      "epoch": 0.20276497695852536,
      "grad_norm": 3.756241798400879,
      "kl": 0.20654296875,
      "learning_rate": 9.499999999999999e-07,
      "loss": 0.1277,
      "reward": 0.6867574155330658,
      "reward_std": 0.1063082329928875,
      "reward_std_all": 0.11204167455434799,
      "rewards-max/forensics_sinkhorn_reward": 0.9473297595977783,
      "rewards-max/forensics_strict_edge_reward": 0.1695053055882454,
      "rewards-max/forensics_tiou_reward": 0.8667500019073486,
      "rewards/forensics_sinkhorn_reward": 0.6093921065330505,
      "rewards/forensics_strict_edge_reward": 0.0071974837919697165,
      "rewards/forensics_tiou_reward": 0.28695280849933624,
      "step": 66
    },
    {
      "completion_length": 24.9375,
      "epoch": 0.20583717357910905,
      "grad_norm": 4.295208930969238,
      "kl": 0.2197265625,
      "learning_rate": 9.492307692307691e-07,
      "loss": 0.0005,
      "reward": 0.6802761852741241,
      "reward_std": 0.11952823773026466,
      "reward_std_all": 0.11919596046209335,
      "rewards-max/forensics_sinkhorn_reward": 0.9336958229541779,
      "rewards-max/forensics_strict_edge_reward": 0.05117631331086159,
      "rewards-max/forensics_tiou_reward": 0.8215879499912262,
      "rewards/forensics_sinkhorn_reward": 0.5458104610443115,
      "rewards/forensics_strict_edge_reward": 0.0025850021047517657,
      "rewards/forensics_tiou_reward": 0.2648279219865799,
      "step": 67
    },
    {
      "completion_length": 26.265625,
      "epoch": 0.20890937019969277,
      "grad_norm": 4.753832817077637,
      "kl": 0.21875,
      "learning_rate": 9.484615384615384e-07,
      "loss": 0.1086,
      "reward": 0.6684581339359283,
      "reward_std": 0.1337982341647148,
      "reward_std_all": 0.13330581039190292,
      "rewards-max/forensics_sinkhorn_reward": 0.9505482316017151,
      "rewards-max/forensics_strict_edge_reward": 0.18701935559511185,
      "rewards-max/forensics_tiou_reward": 0.8736748397350311,
      "rewards/forensics_sinkhorn_reward": 0.5790732800960541,
      "rewards/forensics_strict_edge_reward": 0.009027664782479405,
      "rewards/forensics_tiou_reward": 0.3073820397257805,
      "step": 68
    },
    {
      "completion_length": 22.609375,
      "epoch": 0.2119815668202765,
      "grad_norm": 3.789693593978882,
      "kl": 0.201171875,
      "learning_rate": 9.476923076923077e-07,
      "loss": 0.202,
      "reward": 0.6628182530403137,
      "reward_std": 0.12748783826828003,
      "reward_std_all": 0.13044492900371552,
      "rewards-max/forensics_sinkhorn_reward": 0.9529646933078766,
      "rewards-max/forensics_strict_edge_reward": 0.061083086766302586,
      "rewards-max/forensics_tiou_reward": 0.8704379498958588,
      "rewards/forensics_sinkhorn_reward": 0.5438156127929688,
      "rewards/forensics_strict_edge_reward": 0.0037968187243677676,
      "rewards/forensics_tiou_reward": 0.2460697442293167,
      "step": 69
    },
    {
      "completion_length": 26.703125,
      "epoch": 0.21505376344086022,
      "grad_norm": 4.235161304473877,
      "kl": 0.27392578125,
      "learning_rate": 9.46923076923077e-07,
      "loss": 0.2151,
      "reward": 0.6243164539337158,
      "reward_std": 0.12607774138450623,
      "reward_std_all": 0.1502988114953041,
      "rewards-max/forensics_sinkhorn_reward": 0.9421569406986237,
      "rewards-max/forensics_strict_edge_reward": 0.05726632662117481,
      "rewards-max/forensics_tiou_reward": 0.85442915558815,
      "rewards/forensics_sinkhorn_reward": 0.5348154753446579,
      "rewards/forensics_strict_edge_reward": 0.003585430793464184,
      "rewards/forensics_tiou_reward": 0.2554518058896065,
      "step": 70
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.21812596006144394,
      "grad_norm": 4.07004976272583,
      "kl": 0.19091796875,
      "learning_rate": 9.461538461538461e-07,
      "loss": 0.2679,
      "reward": 0.6605474948883057,
      "reward_std": 0.09482244774699211,
      "reward_std_all": 0.11709873378276825,
      "rewards-max/forensics_sinkhorn_reward": 0.8739146590232849,
      "rewards-max/forensics_strict_edge_reward": 0.0010160261881537735,
      "rewards-max/forensics_tiou_reward": 0.709428608417511,
      "rewards/forensics_sinkhorn_reward": 0.5441160798072815,
      "rewards/forensics_strict_edge_reward": 6.607649629586376e-05,
      "rewards/forensics_tiou_reward": 0.24951928108930588,
      "step": 71
    },
    {
      "completion_length": 22.75,
      "epoch": 0.22119815668202766,
      "grad_norm": 3.7789909839630127,
      "kl": 0.1943359375,
      "learning_rate": 9.453846153846153e-07,
      "loss": 0.1132,
      "reward": 0.6649026870727539,
      "reward_std": 0.10475825518369675,
      "reward_std_all": 0.1199057474732399,
      "rewards-max/forensics_sinkhorn_reward": 0.8253899812698364,
      "rewards-max/forensics_strict_edge_reward": 0.00754847502685152,
      "rewards-max/forensics_tiou_reward": 0.5907166302204132,
      "rewards/forensics_sinkhorn_reward": 0.5395308136940002,
      "rewards/forensics_strict_edge_reward": 0.0005702181697415654,
      "rewards/forensics_tiou_reward": 0.17877256125211716,
      "step": 72
    },
    {
      "completion_length": 23.296875,
      "epoch": 0.22427035330261136,
      "grad_norm": 3.2792880535125732,
      "kl": 0.19189453125,
      "learning_rate": 9.446153846153846e-07,
      "loss": 0.2421,
      "reward": 0.6935276687145233,
      "reward_std": 0.10080065578222275,
      "reward_std_all": 0.10525698214769363,
      "rewards-max/forensics_sinkhorn_reward": 0.9860374331474304,
      "rewards-max/forensics_strict_edge_reward": 0.40730567276477814,
      "rewards-max/forensics_tiou_reward": 0.9619360864162445,
      "rewards/forensics_sinkhorn_reward": 0.6255067884922028,
      "rewards/forensics_strict_edge_reward": 0.02745046280324459,
      "rewards/forensics_tiou_reward": 0.31997424364089966,
      "step": 73
    },
    {
      "completion_length": 24.234375,
      "epoch": 0.22734254992319508,
      "grad_norm": 3.6181061267852783,
      "kl": 0.2158203125,
      "learning_rate": 9.438461538461538e-07,
      "loss": 0.197,
      "reward": 0.7167730033397675,
      "reward_std": 0.07780490070581436,
      "reward_std_all": 0.08471458032727242,
      "rewards-max/forensics_sinkhorn_reward": 0.9583948850631714,
      "rewards-max/forensics_strict_edge_reward": 0.20705574750900269,
      "rewards-max/forensics_tiou_reward": 0.8875834047794342,
      "rewards/forensics_sinkhorn_reward": 0.6321018040180206,
      "rewards/forensics_strict_edge_reward": 0.011896383482962847,
      "rewards/forensics_tiou_reward": 0.2723027765750885,
      "step": 74
    },
    {
      "completion_length": 22.375,
      "epoch": 0.2304147465437788,
      "grad_norm": 4.354221343994141,
      "kl": 0.205078125,
      "learning_rate": 9.43076923076923e-07,
      "loss": 0.1808,
      "reward": 0.6789171695709229,
      "reward_std": 0.12316759303212166,
      "reward_std_all": 0.12321212515234947,
      "rewards-max/forensics_sinkhorn_reward": 0.9184705018997192,
      "rewards-max/forensics_strict_edge_reward": 0.0033012130297720432,
      "rewards-max/forensics_tiou_reward": 0.7320781052112579,
      "rewards/forensics_sinkhorn_reward": 0.5838427841663361,
      "rewards/forensics_strict_edge_reward": 0.00022703855211148039,
      "rewards/forensics_tiou_reward": 0.27660782635211945,
      "step": 75
    },
    {
      "completion_length": 21.90625,
      "epoch": 0.23348694316436253,
      "grad_norm": 5.612739086151123,
      "kl": 0.22705078125,
      "learning_rate": 9.423076923076923e-07,
      "loss": 0.1629,
      "reward": 0.6720679104328156,
      "reward_std": 0.14254407957196236,
      "reward_std_all": 0.1375703178346157,
      "rewards-max/forensics_sinkhorn_reward": 0.9819596111774445,
      "rewards-max/forensics_strict_edge_reward": 0.1131279207766056,
      "rewards-max/forensics_tiou_reward": 0.9431156814098358,
      "rewards/forensics_sinkhorn_reward": 0.591622918844223,
      "rewards/forensics_strict_edge_reward": 0.006813473999500275,
      "rewards/forensics_tiou_reward": 0.2768428921699524,
      "step": 76
    },
    {
      "completion_length": 21.75,
      "epoch": 0.23655913978494625,
      "grad_norm": 4.796081066131592,
      "kl": 0.21337890625,
      "learning_rate": 9.415384615384615e-07,
      "loss": 0.1595,
      "reward": 0.6545740365982056,
      "reward_std": 0.1445176638662815,
      "reward_std_all": 0.14281227812170982,
      "rewards-max/forensics_sinkhorn_reward": 0.940790593624115,
      "rewards-max/forensics_strict_edge_reward": 0.041764519875869155,
      "rewards-max/forensics_tiou_reward": 0.8415098488330841,
      "rewards/forensics_sinkhorn_reward": 0.5497276782989502,
      "rewards/forensics_strict_edge_reward": 0.001921079499879852,
      "rewards/forensics_tiou_reward": 0.18958216905593872,
      "step": 77
    },
    {
      "completion_length": 21.28125,
      "epoch": 0.23963133640552994,
      "grad_norm": 3.4595181941986084,
      "kl": 0.2158203125,
      "learning_rate": 9.407692307692308e-07,
      "loss": 0.07,
      "reward": 0.7282496094703674,
      "reward_std": 0.09596757218241692,
      "reward_std_all": 0.08979488164186478,
      "rewards-max/forensics_sinkhorn_reward": 0.9619913101196289,
      "rewards-max/forensics_strict_edge_reward": 0.18880087370052934,
      "rewards-max/forensics_tiou_reward": 0.8960683047771454,
      "rewards/forensics_sinkhorn_reward": 0.6689808964729309,
      "rewards/forensics_strict_edge_reward": 0.008037681138375774,
      "rewards/forensics_tiou_reward": 0.3439515382051468,
      "step": 78
    },
    {
      "completion_length": 21.84375,
      "epoch": 0.24270353302611367,
      "grad_norm": 4.886840343475342,
      "kl": 0.20068359375,
      "learning_rate": 9.399999999999999e-07,
      "loss": 0.1844,
      "reward": 0.7036336660385132,
      "reward_std": 0.10300270840525627,
      "reward_std_all": 0.10242516547441483,
      "rewards-max/forensics_sinkhorn_reward": 0.9297762513160706,
      "rewards-max/forensics_strict_edge_reward": 0.06573514599585906,
      "rewards-max/forensics_tiou_reward": 0.8126470148563385,
      "rewards/forensics_sinkhorn_reward": 0.6053462326526642,
      "rewards/forensics_strict_edge_reward": 0.003161022897984367,
      "rewards/forensics_tiou_reward": 0.2375946342945099,
      "step": 79
    },
    {
      "completion_length": 21.171875,
      "epoch": 0.2457757296466974,
      "grad_norm": 4.522665500640869,
      "kl": 0.24755859375,
      "learning_rate": 9.392307692307691e-07,
      "loss": 0.1448,
      "reward": 0.7157261371612549,
      "reward_std": 0.09633031859993935,
      "reward_std_all": 0.09352724626660347,
      "rewards-max/forensics_sinkhorn_reward": 0.9521295428276062,
      "rewards-max/forensics_strict_edge_reward": 0.03617061395198107,
      "rewards-max/forensics_tiou_reward": 0.8714019656181335,
      "rewards/forensics_sinkhorn_reward": 0.6648508012294769,
      "rewards/forensics_strict_edge_reward": 0.002994889975525439,
      "rewards/forensics_tiou_reward": 0.35101887583732605,
      "step": 80
    },
    {
      "completion_length": 21.875,
      "epoch": 0.2488479262672811,
      "grad_norm": 4.267309188842773,
      "kl": 0.265625,
      "learning_rate": 9.384615384615384e-07,
      "loss": 0.2045,
      "reward": 0.7083345055580139,
      "reward_std": 0.09555099904537201,
      "reward_std_all": 0.09518783539533615,
      "rewards-max/forensics_sinkhorn_reward": 0.9439051747322083,
      "rewards-max/forensics_strict_edge_reward": 0.09790695831179619,
      "rewards-max/forensics_tiou_reward": 0.8666444420814514,
      "rewards/forensics_sinkhorn_reward": 0.6256648600101471,
      "rewards/forensics_strict_edge_reward": 0.005586386425420642,
      "rewards/forensics_tiou_reward": 0.3146432340145111,
      "step": 81
    },
    {
      "completion_length": 22.9375,
      "epoch": 0.2519201228878648,
      "grad_norm": 3.96553897857666,
      "kl": 0.2607421875,
      "learning_rate": 9.376923076923076e-07,
      "loss": 0.084,
      "reward": 0.7051494419574738,
      "reward_std": 0.09474311396479607,
      "reward_std_all": 0.09473393484950066,
      "rewards-max/forensics_sinkhorn_reward": 0.946824699640274,
      "rewards-max/forensics_strict_edge_reward": 0.2773010693490505,
      "rewards-max/forensics_tiou_reward": 0.8467276096343994,
      "rewards/forensics_sinkhorn_reward": 0.602400153875351,
      "rewards/forensics_strict_edge_reward": 0.011941231263335794,
      "rewards/forensics_tiou_reward": 0.29894913733005524,
      "step": 82
    },
    {
      "completion_length": 22.140625,
      "epoch": 0.25499231950844853,
      "grad_norm": 5.177999973297119,
      "kl": 0.19189453125,
      "learning_rate": 9.36923076923077e-07,
      "loss": 0.1915,
      "reward": 0.6612635254859924,
      "reward_std": 0.09515508264303207,
      "reward_std_all": 0.11638705059885979,
      "rewards-max/forensics_sinkhorn_reward": 0.9283065497875214,
      "rewards-max/forensics_strict_edge_reward": 0.16900974977761507,
      "rewards-max/forensics_tiou_reward": 0.8392939269542694,
      "rewards/forensics_sinkhorn_reward": 0.5633261799812317,
      "rewards/forensics_strict_edge_reward": 0.012205081467982382,
      "rewards/forensics_tiou_reward": 0.2269340679049492,
      "step": 83
    },
    {
      "completion_length": 20.15625,
      "epoch": 0.25806451612903225,
      "grad_norm": 3.5022122859954834,
      "kl": 0.21630859375,
      "learning_rate": 9.361538461538461e-07,
      "loss": 0.1968,
      "reward": 0.7020106017589569,
      "reward_std": 0.08823681622743607,
      "reward_std_all": 0.09473077766597271,
      "rewards-max/forensics_sinkhorn_reward": 0.9073673784732819,
      "rewards-max/forensics_strict_edge_reward": 0.009592261631041765,
      "rewards-max/forensics_tiou_reward": 0.7835521399974823,
      "rewards/forensics_sinkhorn_reward": 0.5900413393974304,
      "rewards/forensics_strict_edge_reward": 0.0006977295852266252,
      "rewards/forensics_tiou_reward": 0.21867496520280838,
      "step": 84
    },
    {
      "completion_length": 23.140625,
      "epoch": 0.261136712749616,
      "grad_norm": 4.534854888916016,
      "kl": 0.25048828125,
      "learning_rate": 9.353846153846153e-07,
      "loss": 0.2151,
      "reward": 0.6993209421634674,
      "reward_std": 0.07394920662045479,
      "reward_std_all": 0.08942687883973122,
      "rewards-max/forensics_sinkhorn_reward": 0.9096974432468414,
      "rewards-max/forensics_strict_edge_reward": 0.06324205547571182,
      "rewards-max/forensics_tiou_reward": 0.7948300540447235,
      "rewards/forensics_sinkhorn_reward": 0.6057525873184204,
      "rewards/forensics_strict_edge_reward": 0.004140285309404135,
      "rewards/forensics_tiou_reward": 0.26952384412288666,
      "step": 85
    },
    {
      "completion_length": 22.953125,
      "epoch": 0.2642089093701997,
      "grad_norm": 5.263971328735352,
      "kl": 0.2841796875,
      "learning_rate": 9.346153846153846e-07,
      "loss": 0.161,
      "reward": 0.7274176180362701,
      "reward_std": 0.08045116998255253,
      "reward_std_all": 0.07950194738805294,
      "rewards-max/forensics_sinkhorn_reward": 0.9311032891273499,
      "rewards-max/forensics_strict_edge_reward": 0.0871574841439724,
      "rewards-max/forensics_tiou_reward": 0.8368279933929443,
      "rewards/forensics_sinkhorn_reward": 0.6522811055183411,
      "rewards/forensics_strict_edge_reward": 0.004616252263076603,
      "rewards/forensics_tiou_reward": 0.32812371850013733,
      "step": 86
    },
    {
      "completion_length": 23.296875,
      "epoch": 0.2672811059907834,
      "grad_norm": 4.682419300079346,
      "kl": 0.22314453125,
      "learning_rate": 9.338461538461538e-07,
      "loss": 0.0523,
      "reward": 0.7131103575229645,
      "reward_std": 0.08766103535890579,
      "reward_std_all": 0.09007611498236656,
      "rewards-max/forensics_sinkhorn_reward": 0.9878047108650208,
      "rewards-max/forensics_strict_edge_reward": 0.22173970192670822,
      "rewards-max/forensics_tiou_reward": 0.9705213606357574,
      "rewards/forensics_sinkhorn_reward": 0.6255195736885071,
      "rewards/forensics_strict_edge_reward": 0.009840480925049633,
      "rewards/forensics_tiou_reward": 0.34357085824012756,
      "step": 87
    },
    {
      "completion_length": 21.9375,
      "epoch": 0.27035330261136714,
      "grad_norm": 6.46309757232666,
      "kl": 0.447265625,
      "learning_rate": 9.33076923076923e-07,
      "loss": 0.1281,
      "reward": 0.7071106135845184,
      "reward_std": 0.11256527900695801,
      "reward_std_all": 0.10637815669178963,
      "rewards-max/forensics_sinkhorn_reward": 0.9839564263820648,
      "rewards-max/forensics_strict_edge_reward": 0.25728432572213933,
      "rewards-max/forensics_tiou_reward": 0.9588529169559479,
      "rewards/forensics_sinkhorn_reward": 0.6402434706687927,
      "rewards/forensics_strict_edge_reward": 0.017942002228664933,
      "rewards/forensics_tiou_reward": 0.35953471064567566,
      "step": 88
    },
    {
      "completion_length": 22.390625,
      "epoch": 0.27342549923195086,
      "grad_norm": 5.783779144287109,
      "kl": 0.22265625,
      "learning_rate": 9.323076923076922e-07,
      "loss": 0.1768,
      "reward": 0.6984972059726715,
      "reward_std": 0.08007017895579338,
      "reward_std_all": 0.08921826258301735,
      "rewards-max/forensics_sinkhorn_reward": 0.9347757995128632,
      "rewards-max/forensics_strict_edge_reward": 0.027071746066212654,
      "rewards-max/forensics_tiou_reward": 0.8269099295139313,
      "rewards/forensics_sinkhorn_reward": 0.5875550508499146,
      "rewards/forensics_strict_edge_reward": 0.001240427780430764,
      "rewards/forensics_tiou_reward": 0.23200254142284393,
      "step": 89
    },
    {
      "completion_length": 22.859375,
      "epoch": 0.2764976958525346,
      "grad_norm": 5.734184741973877,
      "kl": 0.22265625,
      "learning_rate": 9.315384615384615e-07,
      "loss": 0.0842,
      "reward": 0.7314207255840302,
      "reward_std": 0.0667300894856453,
      "reward_std_all": 0.06867684237658978,
      "rewards-max/forensics_sinkhorn_reward": 0.9193016886711121,
      "rewards-max/forensics_strict_edge_reward": 0.47527681291103363,
      "rewards-max/forensics_tiou_reward": 0.8040453791618347,
      "rewards/forensics_sinkhorn_reward": 0.6321836411952972,
      "rewards/forensics_strict_edge_reward": 0.024705429561436176,
      "rewards/forensics_tiou_reward": 0.28743402659893036,
      "step": 90
    },
    {
      "completion_length": 21.75,
      "epoch": 0.27956989247311825,
      "grad_norm": 4.371704578399658,
      "kl": 0.20068359375,
      "learning_rate": 9.307692307692308e-07,
      "loss": 0.2509,
      "reward": 0.7060385644435883,
      "reward_std": 0.09754272177815437,
      "reward_std_all": 0.09799978882074356,
      "rewards-max/forensics_sinkhorn_reward": 0.9925464689731598,
      "rewards-max/forensics_strict_edge_reward": 0.38867069967091084,
      "rewards-max/forensics_tiou_reward": 0.982019692659378,
      "rewards/forensics_sinkhorn_reward": 0.6240027844905853,
      "rewards/forensics_strict_edge_reward": 0.017556319944560528,
      "rewards/forensics_tiou_reward": 0.3097108453512192,
      "step": 91
    },
    {
      "completion_length": 23.25,
      "epoch": 0.282642089093702,
      "grad_norm": 3.392103433609009,
      "kl": 0.2138671875,
      "learning_rate": 9.3e-07,
      "loss": 0.0039,
      "reward": 0.6785430014133453,
      "reward_std": 0.10889589600265026,
      "reward_std_all": 0.11161499097943306,
      "rewards-max/forensics_sinkhorn_reward": 0.9758457839488983,
      "rewards-max/forensics_strict_edge_reward": 0.3054015785455704,
      "rewards-max/forensics_tiou_reward": 0.9314217269420624,
      "rewards/forensics_sinkhorn_reward": 0.5776605606079102,
      "rewards/forensics_strict_edge_reward": 0.02239783084951341,
      "rewards/forensics_tiou_reward": 0.2694685161113739,
      "step": 92
    },
    {
      "completion_length": 27.0,
      "epoch": 0.2857142857142857,
      "grad_norm": 5.067103385925293,
      "kl": 0.24072265625,
      "learning_rate": 9.292307692307692e-07,
      "loss": 0.0784,
      "reward": 0.6620716452598572,
      "reward_std": 0.11369789764285088,
      "reward_std_all": 0.12107479572296143,
      "rewards-max/forensics_sinkhorn_reward": 0.8903214931488037,
      "rewards-max/forensics_strict_edge_reward": 0.032955541188130155,
      "rewards-max/forensics_tiou_reward": 0.7481644451618195,
      "rewards/forensics_sinkhorn_reward": 0.5040369629859924,
      "rewards/forensics_strict_edge_reward": 0.0019644908388727345,
      "rewards/forensics_tiou_reward": 0.2577449902892113,
      "step": 93
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.2887864823348694,
      "grad_norm": 4.97788667678833,
      "kl": 0.23193359375,
      "learning_rate": 9.284615384615384e-07,
      "loss": 0.2283,
      "reward": 0.6466630101203918,
      "reward_std": 0.1553238481283188,
      "reward_std_all": 0.15212185680866241,
      "rewards-max/forensics_sinkhorn_reward": 0.9404212534427643,
      "rewards-max/forensics_strict_edge_reward": 0.14612476155161858,
      "rewards-max/forensics_tiou_reward": 0.8592219352722168,
      "rewards/forensics_sinkhorn_reward": 0.5256589502096176,
      "rewards/forensics_strict_edge_reward": 0.010722059290856123,
      "rewards/forensics_tiou_reward": 0.24575212597846985,
      "step": 94
    },
    {
      "completion_length": 26.4375,
      "epoch": 0.29185867895545314,
      "grad_norm": 4.471096038818359,
      "kl": 0.2001953125,
      "learning_rate": 9.276923076923076e-07,
      "loss": 0.1261,
      "reward": 0.6510410904884338,
      "reward_std": 0.13532878458499908,
      "reward_std_all": 0.14270759373903275,
      "rewards-max/forensics_sinkhorn_reward": 0.9223261177539825,
      "rewards-max/forensics_strict_edge_reward": 0.0173903233371675,
      "rewards-max/forensics_tiou_reward": 0.8163892328739166,
      "rewards/forensics_sinkhorn_reward": 0.5347152054309845,
      "rewards/forensics_strict_edge_reward": 0.0007400436152238399,
      "rewards/forensics_tiou_reward": 0.24349398910999298,
      "step": 95
    },
    {
      "completion_length": 24.015625,
      "epoch": 0.29493087557603687,
      "grad_norm": 4.713464260101318,
      "kl": 0.16748046875,
      "learning_rate": 9.26923076923077e-07,
      "loss": 0.1077,
      "reward": 0.6916749477386475,
      "reward_std": 0.09073468297719955,
      "reward_std_all": 0.0962236188352108,
      "rewards-max/forensics_sinkhorn_reward": 0.9156582355499268,
      "rewards-max/forensics_strict_edge_reward": 0.04722312014200725,
      "rewards-max/forensics_tiou_reward": 0.7893075942993164,
      "rewards/forensics_sinkhorn_reward": 0.5780527293682098,
      "rewards/forensics_strict_edge_reward": 0.0026481303430045955,
      "rewards/forensics_tiou_reward": 0.2366192787885666,
      "step": 96
    },
    {
      "completion_length": 25.140625,
      "epoch": 0.2980030721966206,
      "grad_norm": 3.873727798461914,
      "kl": 0.22216796875,
      "learning_rate": 9.261538461538461e-07,
      "loss": 0.0957,
      "reward": 0.7001480460166931,
      "reward_std": 0.11561942100524902,
      "reward_std_all": 0.11145566403865814,
      "rewards-max/forensics_sinkhorn_reward": 0.9413643181324005,
      "rewards-max/forensics_strict_edge_reward": 0.02925721462816,
      "rewards-max/forensics_tiou_reward": 0.8508453965187073,
      "rewards/forensics_sinkhorn_reward": 0.5997774302959442,
      "rewards/forensics_strict_edge_reward": 0.0020760871120728552,
      "rewards/forensics_tiou_reward": 0.3352777361869812,
      "step": 97
    },
    {
      "completion_length": 26.09375,
      "epoch": 0.3010752688172043,
      "grad_norm": 3.832230806350708,
      "kl": 0.201171875,
      "learning_rate": 9.253846153846153e-07,
      "loss": 0.2279,
      "reward": 0.7102619111537933,
      "reward_std": 0.06317484751343727,
      "reward_std_all": 0.07985207624733448,
      "rewards-max/forensics_sinkhorn_reward": 0.9025106728076935,
      "rewards-max/forensics_strict_edge_reward": 0.07588224415667355,
      "rewards-max/forensics_tiou_reward": 0.7569214999675751,
      "rewards/forensics_sinkhorn_reward": 0.5685238242149353,
      "rewards/forensics_strict_edge_reward": 0.005016790302761365,
      "rewards/forensics_tiou_reward": 0.28007563948631287,
      "step": 98
    },
    {
      "completion_length": 26.3125,
      "epoch": 0.30414746543778803,
      "grad_norm": 4.282777309417725,
      "kl": 0.1767578125,
      "learning_rate": 9.246153846153846e-07,
      "loss": 0.2667,
      "reward": 0.6609078943729401,
      "reward_std": 0.0996792446821928,
      "reward_std_all": 0.11593426018953323,
      "rewards-max/forensics_sinkhorn_reward": 0.9505952894687653,
      "rewards-max/forensics_strict_edge_reward": 0.12879132118541747,
      "rewards-max/forensics_tiou_reward": 0.8716882467269897,
      "rewards/forensics_sinkhorn_reward": 0.5036909133195877,
      "rewards/forensics_strict_edge_reward": 0.008115282874314289,
      "rewards/forensics_tiou_reward": 0.24626708030700684,
      "step": 99
    },
    {
      "completion_length": 23.046875,
      "epoch": 0.30721966205837176,
      "grad_norm": 3.651859760284424,
      "kl": 0.20703125,
      "learning_rate": 9.238461538461538e-07,
      "loss": 0.0851,
      "reward": 0.6674145460128784,
      "reward_std": 0.10269656777381897,
      "reward_std_all": 0.11705900728702545,
      "rewards-max/forensics_sinkhorn_reward": 0.921154648065567,
      "rewards-max/forensics_strict_edge_reward": 0.01749390526674688,
      "rewards-max/forensics_tiou_reward": 0.8003347218036652,
      "rewards/forensics_sinkhorn_reward": 0.5700124502182007,
      "rewards/forensics_strict_edge_reward": 0.0008481373661197722,
      "rewards/forensics_tiou_reward": 0.23463299870491028,
      "step": 100
    },
    {
      "completion_length": 23.25,
      "epoch": 0.3102918586789555,
      "grad_norm": 5.758857727050781,
      "kl": 0.177734375,
      "learning_rate": 9.230769230769231e-07,
      "loss": 0.1755,
      "reward": 0.6749509274959564,
      "reward_std": 0.11829860508441925,
      "reward_std_all": 0.12091788277029991,
      "rewards-max/forensics_sinkhorn_reward": 0.8699263036251068,
      "rewards-max/forensics_strict_edge_reward": 0.013249407173134387,
      "rewards-max/forensics_tiou_reward": 0.6778396964073181,
      "rewards/forensics_sinkhorn_reward": 0.5671199560165405,
      "rewards/forensics_strict_edge_reward": 0.0006598432519240305,
      "rewards/forensics_tiou_reward": 0.24152100086212158,
      "step": 101
    },
    {
      "completion_length": 22.703125,
      "epoch": 0.31336405529953915,
      "grad_norm": 4.306581497192383,
      "kl": 0.23095703125,
      "learning_rate": 9.223076923076922e-07,
      "loss": 0.0615,
      "reward": 0.6690616607666016,
      "reward_std": 0.13039033859968185,
      "reward_std_all": 0.12883027642965317,
      "rewards-max/forensics_sinkhorn_reward": 0.9585892856121063,
      "rewards-max/forensics_strict_edge_reward": 0.450752479955554,
      "rewards-max/forensics_tiou_reward": 0.8728136718273163,
      "rewards/forensics_sinkhorn_reward": 0.5817264318466187,
      "rewards/forensics_strict_edge_reward": 0.018891837098635733,
      "rewards/forensics_tiou_reward": 0.3456335812807083,
      "step": 102
    },
    {
      "completion_length": 26.296875,
      "epoch": 0.31643625192012287,
      "grad_norm": 5.305150032043457,
      "kl": 0.19921875,
      "learning_rate": 9.215384615384614e-07,
      "loss": 0.0943,
      "reward": 0.7009948790073395,
      "reward_std": 0.11132040992379189,
      "reward_std_all": 0.10893324390053749,
      "rewards-max/forensics_sinkhorn_reward": 0.8894485235214233,
      "rewards-max/forensics_strict_edge_reward": 0.0914009790867567,
      "rewards-max/forensics_tiou_reward": 0.7505085468292236,
      "rewards/forensics_sinkhorn_reward": 0.5567324459552765,
      "rewards/forensics_strict_edge_reward": 0.004187697311863303,
      "rewards/forensics_tiou_reward": 0.2624644711613655,
      "step": 103
    },
    {
      "completion_length": 24.75,
      "epoch": 0.3195084485407066,
      "grad_norm": 3.941246271133423,
      "kl": 0.21923828125,
      "learning_rate": 9.207692307692308e-07,
      "loss": 0.0175,
      "reward": 0.6502668559551239,
      "reward_std": 0.13082334399223328,
      "reward_std_all": 0.13847079128026962,
      "rewards-max/forensics_sinkhorn_reward": 0.940136730670929,
      "rewards-max/forensics_strict_edge_reward": 0.1285964889520983,
      "rewards-max/forensics_tiou_reward": 0.8449200689792633,
      "rewards/forensics_sinkhorn_reward": 0.5641140341758728,
      "rewards/forensics_strict_edge_reward": 0.0053809519204151,
      "rewards/forensics_tiou_reward": 0.2704722434282303,
      "step": 104
    },
    {
      "completion_length": 23.390625,
      "epoch": 0.3225806451612903,
      "grad_norm": 4.266923427581787,
      "kl": 0.2353515625,
      "learning_rate": 9.2e-07,
      "loss": 0.1586,
      "reward": 0.6936835944652557,
      "reward_std": 0.09648536518216133,
      "reward_std_all": 0.10405761376023293,
      "rewards-max/forensics_sinkhorn_reward": 0.9576727747917175,
      "rewards-max/forensics_strict_edge_reward": 0.06222156621515751,
      "rewards-max/forensics_tiou_reward": 0.8732912540435791,
      "rewards/forensics_sinkhorn_reward": 0.6078499555587769,
      "rewards/forensics_strict_edge_reward": 0.0036241684574633837,
      "rewards/forensics_tiou_reward": 0.3111257404088974,
      "step": 105
    },
    {
      "completion_length": 26.5,
      "epoch": 0.32565284178187404,
      "grad_norm": 3.670259714126587,
      "kl": 0.1962890625,
      "learning_rate": 9.192307692307692e-07,
      "loss": 0.1796,
      "reward": 0.686922162771225,
      "reward_std": 0.09596589207649231,
      "reward_std_all": 0.1053490862250328,
      "rewards-max/forensics_sinkhorn_reward": 0.9388539493083954,
      "rewards-max/forensics_strict_edge_reward": 0.09727144031785429,
      "rewards-max/forensics_tiou_reward": 0.8435119092464447,
      "rewards/forensics_sinkhorn_reward": 0.5451177954673767,
      "rewards/forensics_strict_edge_reward": 0.004163810954196379,
      "rewards/forensics_tiou_reward": 0.25957778096199036,
      "step": 106
    },
    {
      "completion_length": 23.9375,
      "epoch": 0.32872503840245776,
      "grad_norm": 3.848907709121704,
      "kl": 0.1904296875,
      "learning_rate": 9.184615384615384e-07,
      "loss": 0.2429,
      "reward": 0.697106271982193,
      "reward_std": 0.11788028478622437,
      "reward_std_all": 0.11314428970217705,
      "rewards-max/forensics_sinkhorn_reward": 0.9946532547473907,
      "rewards-max/forensics_strict_edge_reward": 0.4935835525393486,
      "rewards-max/forensics_tiou_reward": 0.9900981783866882,
      "rewards/forensics_sinkhorn_reward": 0.6205025911331177,
      "rewards/forensics_strict_edge_reward": 0.026960923336446285,
      "rewards/forensics_tiou_reward": 0.3781028687953949,
      "step": 107
    },
    {
      "completion_length": 21.609375,
      "epoch": 0.3317972350230415,
      "grad_norm": 4.191844940185547,
      "kl": 0.2021484375,
      "learning_rate": 9.176923076923076e-07,
      "loss": 0.1765,
      "reward": 0.6963502764701843,
      "reward_std": 0.09073744714260101,
      "reward_std_all": 0.09794282913208008,
      "rewards-max/forensics_sinkhorn_reward": 0.9407356083393097,
      "rewards-max/forensics_strict_edge_reward": 0.04968561977148056,
      "rewards-max/forensics_tiou_reward": 0.8556377291679382,
      "rewards/forensics_sinkhorn_reward": 0.6096320450305939,
      "rewards/forensics_strict_edge_reward": 0.003534223244059831,
      "rewards/forensics_tiou_reward": 0.295095831155777,
      "step": 108
    },
    {
      "completion_length": 24.796875,
      "epoch": 0.3348694316436252,
      "grad_norm": 4.119208812713623,
      "kl": 0.20361328125,
      "learning_rate": 9.169230769230769e-07,
      "loss": 0.2461,
      "reward": 0.6996461749076843,
      "reward_std": 0.10732059180736542,
      "reward_std_all": 0.10664438083767891,
      "rewards-max/forensics_sinkhorn_reward": 0.89900141954422,
      "rewards-max/forensics_strict_edge_reward": 0.00578644999768585,
      "rewards-max/forensics_tiou_reward": 0.7420389354228973,
      "rewards/forensics_sinkhorn_reward": 0.5603993237018585,
      "rewards/forensics_strict_edge_reward": 0.00045732919170404784,
      "rewards/forensics_tiou_reward": 0.2663929983973503,
      "step": 109
    },
    {
      "completion_length": 22.34375,
      "epoch": 0.3379416282642089,
      "grad_norm": 4.429539680480957,
      "kl": 0.205078125,
      "learning_rate": 9.161538461538462e-07,
      "loss": 0.1595,
      "reward": 0.7196446061134338,
      "reward_std": 0.08532222360372543,
      "reward_std_all": 0.08693917468190193,
      "rewards-max/forensics_sinkhorn_reward": 0.9067962467670441,
      "rewards-max/forensics_strict_edge_reward": 0.04022823052946478,
      "rewards-max/forensics_tiou_reward": 0.7575763165950775,
      "rewards/forensics_sinkhorn_reward": 0.632832944393158,
      "rewards/forensics_strict_edge_reward": 0.0018733221950242296,
      "rewards/forensics_tiou_reward": 0.2654494047164917,
      "step": 110
    },
    {
      "completion_length": 21.453125,
      "epoch": 0.34101382488479265,
      "grad_norm": 3.9681291580200195,
      "kl": 0.2216796875,
      "learning_rate": 9.153846153846153e-07,
      "loss": 0.1482,
      "reward": 0.6843064725399017,
      "reward_std": 0.09540785104036331,
      "reward_std_all": 0.10736753046512604,
      "rewards-max/forensics_sinkhorn_reward": 0.9833578169345856,
      "rewards-max/forensics_strict_edge_reward": 0.33621274679899216,
      "rewards-max/forensics_tiou_reward": 0.9619598686695099,
      "rewards/forensics_sinkhorn_reward": 0.6026304066181183,
      "rewards/forensics_strict_edge_reward": 0.0218857666477561,
      "rewards/forensics_tiou_reward": 0.3199964910745621,
      "step": 111
    },
    {
      "completion_length": 22.03125,
      "epoch": 0.34408602150537637,
      "grad_norm": 4.757893085479736,
      "kl": 0.23193359375,
      "learning_rate": 9.146153846153846e-07,
      "loss": 0.1268,
      "reward": 0.696610152721405,
      "reward_std": 0.09412697330117226,
      "reward_std_all": 0.1028609424829483,
      "rewards-max/forensics_sinkhorn_reward": 0.945885568857193,
      "rewards-max/forensics_strict_edge_reward": 0.12476137652993202,
      "rewards-max/forensics_tiou_reward": 0.8604589104652405,
      "rewards/forensics_sinkhorn_reward": 0.6355997323989868,
      "rewards/forensics_strict_edge_reward": 0.009776128455996513,
      "rewards/forensics_tiou_reward": 0.3830427825450897,
      "step": 112
    },
    {
      "completion_length": 21.90625,
      "epoch": 0.34715821812596004,
      "grad_norm": 4.336268424987793,
      "kl": 0.22509765625,
      "learning_rate": 9.138461538461538e-07,
      "loss": 0.2115,
      "reward": 0.7084046900272369,
      "reward_std": 0.1011243611574173,
      "reward_std_all": 0.10060903057456017,
      "rewards-max/forensics_sinkhorn_reward": 0.9806437492370605,
      "rewards-max/forensics_strict_edge_reward": 0.3722097232239321,
      "rewards-max/forensics_tiou_reward": 0.9500768482685089,
      "rewards/forensics_sinkhorn_reward": 0.6272992491722107,
      "rewards/forensics_strict_edge_reward": 0.015802662703208625,
      "rewards/forensics_tiou_reward": 0.3167574927210808,
      "step": 113
    },
    {
      "completion_length": 22.421875,
      "epoch": 0.35023041474654376,
      "grad_norm": 4.98586368560791,
      "kl": 0.20654296875,
      "learning_rate": 9.130769230769231e-07,
      "loss": 0.0418,
      "reward": 0.685937225818634,
      "reward_std": 0.13640501350164413,
      "reward_std_all": 0.12937798351049423,
      "rewards-max/forensics_sinkhorn_reward": 0.9092879295349121,
      "rewards-max/forensics_strict_edge_reward": 0.20888755097985268,
      "rewards-max/forensics_tiou_reward": 0.7876718938350677,
      "rewards/forensics_sinkhorn_reward": 0.5623048543930054,
      "rewards/forensics_strict_edge_reward": 0.0134406762663275,
      "rewards/forensics_tiou_reward": 0.24273161590099335,
      "step": 114
    },
    {
      "completion_length": 24.34375,
      "epoch": 0.3533026113671275,
      "grad_norm": 4.993419647216797,
      "kl": 0.20703125,
      "learning_rate": 9.123076923076922e-07,
      "loss": 0.209,
      "reward": 0.7216192185878754,
      "reward_std": 0.09492982551455498,
      "reward_std_all": 0.09039242565631866,
      "rewards-max/forensics_sinkhorn_reward": 0.8701175451278687,
      "rewards-max/forensics_strict_edge_reward": 0.28077322617173195,
      "rewards-max/forensics_tiou_reward": 0.6901290714740753,
      "rewards/forensics_sinkhorn_reward": 0.6264376044273376,
      "rewards/forensics_strict_edge_reward": 0.014547927072271705,
      "rewards/forensics_tiou_reward": 0.28580842912197113,
      "step": 115
    },
    {
      "completion_length": 22.828125,
      "epoch": 0.3563748079877112,
      "grad_norm": 4.193148136138916,
      "kl": 0.23974609375,
      "learning_rate": 9.115384615384614e-07,
      "loss": 0.0641,
      "reward": 0.7196631133556366,
      "reward_std": 0.04420343227684498,
      "reward_std_all": 0.0661297608166933,
      "rewards-max/forensics_sinkhorn_reward": 0.9358420968055725,
      "rewards-max/forensics_strict_edge_reward": 0.0057453550655282015,
      "rewards-max/forensics_tiou_reward": 0.8413376808166504,
      "rewards/forensics_sinkhorn_reward": 0.6327272057533264,
      "rewards/forensics_strict_edge_reward": 0.00029437838747981004,
      "rewards/forensics_tiou_reward": 0.2853798121213913,
      "step": 116
    },
    {
      "completion_length": 22.703125,
      "epoch": 0.35944700460829493,
      "grad_norm": 7.356918811798096,
      "kl": 0.4931640625,
      "learning_rate": 9.107692307692308e-07,
      "loss": 0.0528,
      "reward": 0.6704687476158142,
      "reward_std": 0.13053467124700546,
      "reward_std_all": 0.1353789083659649,
      "rewards-max/forensics_sinkhorn_reward": 0.8863795101642609,
      "rewards-max/forensics_strict_edge_reward": 0.0709561612457037,
      "rewards-max/forensics_tiou_reward": 0.7317667901515961,
      "rewards/forensics_sinkhorn_reward": 0.5513264834880829,
      "rewards/forensics_strict_edge_reward": 0.004277402244042605,
      "rewards/forensics_tiou_reward": 0.2335566133260727,
      "step": 117
    },
    {
      "completion_length": 24.828125,
      "epoch": 0.36251920122887865,
      "grad_norm": 3.9320228099823,
      "kl": 0.2744140625,
      "learning_rate": 9.1e-07,
      "loss": 0.2043,
      "reward": 0.6722441017627716,
      "reward_std": 0.10549745336174965,
      "reward_std_all": 0.1161915734410286,
      "rewards-max/forensics_sinkhorn_reward": 0.9503358006477356,
      "rewards-max/forensics_strict_edge_reward": 0.08269815426319838,
      "rewards-max/forensics_tiou_reward": 0.8839049339294434,
      "rewards/forensics_sinkhorn_reward": 0.5577518939971924,
      "rewards/forensics_strict_edge_reward": 0.005159110311069526,
      "rewards/forensics_tiou_reward": 0.2497859075665474,
      "step": 118
    },
    {
      "completion_length": 22.328125,
      "epoch": 0.3655913978494624,
      "grad_norm": 4.499390125274658,
      "kl": 0.251953125,
      "learning_rate": 9.092307692307692e-07,
      "loss": 0.112,
      "reward": 0.7428581714630127,
      "reward_std": 0.06675820238888264,
      "reward_std_all": 0.06550135277211666,
      "rewards-max/forensics_sinkhorn_reward": 0.9533773958683014,
      "rewards-max/forensics_strict_edge_reward": 0.08149041049182415,
      "rewards-max/forensics_tiou_reward": 0.8820644915103912,
      "rewards/forensics_sinkhorn_reward": 0.6383150219917297,
      "rewards/forensics_strict_edge_reward": 0.003813597315456718,
      "rewards/forensics_tiou_reward": 0.31794168055057526,
      "step": 119
    },
    {
      "completion_length": 24.171875,
      "epoch": 0.3686635944700461,
      "grad_norm": 4.109659194946289,
      "kl": 0.2294921875,
      "learning_rate": 9.084615384615384e-07,
      "loss": 0.0462,
      "reward": 0.6653639674186707,
      "reward_std": 0.150541253387928,
      "reward_std_all": 0.14494439214468002,
      "rewards-max/forensics_sinkhorn_reward": 0.9251636862754822,
      "rewards-max/forensics_strict_edge_reward": 0.008015240178281147,
      "rewards-max/forensics_tiou_reward": 0.7941099405288696,
      "rewards/forensics_sinkhorn_reward": 0.5431583821773529,
      "rewards/forensics_strict_edge_reward": 0.00044731838955769376,
      "rewards/forensics_tiou_reward": 0.2943018823862076,
      "step": 120
    },
    {
      "completion_length": 23.765625,
      "epoch": 0.3717357910906298,
      "grad_norm": 4.576853275299072,
      "kl": 0.27490234375,
      "learning_rate": 9.076923076923076e-07,
      "loss": 0.0973,
      "reward": 0.687908262014389,
      "reward_std": 0.10273952037096024,
      "reward_std_all": 0.10803429409861565,
      "rewards-max/forensics_sinkhorn_reward": 0.9183385670185089,
      "rewards-max/forensics_strict_edge_reward": 0.23336449405178428,
      "rewards-max/forensics_tiou_reward": 0.7928915023803711,
      "rewards/forensics_sinkhorn_reward": 0.5412402749061584,
      "rewards/forensics_strict_edge_reward": 0.014364477465278469,
      "rewards/forensics_tiou_reward": 0.257422111928463,
      "step": 121
    },
    {
      "completion_length": 23.90625,
      "epoch": 0.37480798771121354,
      "grad_norm": 3.8593838214874268,
      "kl": 0.2392578125,
      "learning_rate": 9.069230769230769e-07,
      "loss": 0.0859,
      "reward": 0.7451426386833191,
      "reward_std": 0.04950149543583393,
      "reward_std_all": 0.05695851147174835,
      "rewards-max/forensics_sinkhorn_reward": 0.9636361300945282,
      "rewards-max/forensics_strict_edge_reward": 0.5030358955264091,
      "rewards-max/forensics_tiou_reward": 0.9181647300720215,
      "rewards/forensics_sinkhorn_reward": 0.6559692025184631,
      "rewards/forensics_strict_edge_reward": 0.041041008196771145,
      "rewards/forensics_tiou_reward": 0.3703630864620209,
      "step": 122
    },
    {
      "completion_length": 25.453125,
      "epoch": 0.3778801843317972,
      "grad_norm": 4.542162895202637,
      "kl": 0.2490234375,
      "learning_rate": 9.061538461538461e-07,
      "loss": 0.1985,
      "reward": 0.6868870258331299,
      "reward_std": 0.11774998903274536,
      "reward_std_all": 0.12058917433023453,
      "rewards-max/forensics_sinkhorn_reward": 0.9397499561309814,
      "rewards-max/forensics_strict_edge_reward": 0.306136317551136,
      "rewards-max/forensics_tiou_reward": 0.8533807992935181,
      "rewards/forensics_sinkhorn_reward": 0.5968291163444519,
      "rewards/forensics_strict_edge_reward": 0.016539046075195074,
      "rewards/forensics_tiou_reward": 0.33223406970500946,
      "step": 123
    },
    {
      "completion_length": 22.625,
      "epoch": 0.38095238095238093,
      "grad_norm": 4.517123222351074,
      "kl": 0.2421875,
      "learning_rate": 9.053846153846153e-07,
      "loss": 0.1112,
      "reward": 0.7160563170909882,
      "reward_std": 0.08195282891392708,
      "reward_std_all": 0.08720916137099266,
      "rewards-max/forensics_sinkhorn_reward": 0.987419992685318,
      "rewards-max/forensics_strict_edge_reward": 0.43434032052755356,
      "rewards-max/forensics_tiou_reward": 0.9688415825366974,
      "rewards/forensics_sinkhorn_reward": 0.6367010474205017,
      "rewards/forensics_strict_edge_reward": 0.019371022004634142,
      "rewards/forensics_tiou_reward": 0.3580985367298126,
      "step": 124
    },
    {
      "completion_length": 28.84375,
      "epoch": 0.38402457757296465,
      "grad_norm": 3.759065866470337,
      "kl": 0.234375,
      "learning_rate": 9.046153846153846e-07,
      "loss": 0.1557,
      "reward": 0.7020866274833679,
      "reward_std": 0.0922927800565958,
      "reward_std_all": 0.09774931333959103,
      "rewards-max/forensics_sinkhorn_reward": 0.9170592725276947,
      "rewards-max/forensics_strict_edge_reward": 0.1342186463298276,
      "rewards-max/forensics_tiou_reward": 0.7828560471534729,
      "rewards/forensics_sinkhorn_reward": 0.5510289072990417,
      "rewards/forensics_strict_edge_reward": 0.008892470970749855,
      "rewards/forensics_tiou_reward": 0.3218236565589905,
      "step": 125
    },
    {
      "completion_length": 23.4375,
      "epoch": 0.3870967741935484,
      "grad_norm": 4.462498664855957,
      "kl": 0.25244140625,
      "learning_rate": 9.038461538461538e-07,
      "loss": 0.1216,
      "reward": 0.6799530684947968,
      "reward_std": 0.11339934542775154,
      "reward_std_all": 0.11674759536981583,
      "rewards-max/forensics_sinkhorn_reward": 0.8510110974311829,
      "rewards-max/forensics_strict_edge_reward": 0.11542116949567571,
      "rewards-max/forensics_tiou_reward": 0.6631093621253967,
      "rewards/forensics_sinkhorn_reward": 0.5672641396522522,
      "rewards/forensics_strict_edge_reward": 0.004979556251782924,
      "rewards/forensics_tiou_reward": 0.25060373544692993,
      "step": 126
    },
    {
      "completion_length": 24.015625,
      "epoch": 0.3901689708141321,
      "grad_norm": 4.037590980529785,
      "kl": 0.2578125,
      "learning_rate": 9.030769230769231e-07,
      "loss": 0.1328,
      "reward": 0.7168701589107513,
      "reward_std": 0.1071094498038292,
      "reward_std_all": 0.10039391368627548,
      "rewards-max/forensics_sinkhorn_reward": 0.9773807227611542,
      "rewards-max/forensics_strict_edge_reward": 0.1457085356087191,
      "rewards-max/forensics_tiou_reward": 0.9418978691101074,
      "rewards/forensics_sinkhorn_reward": 0.6365708708763123,
      "rewards/forensics_strict_edge_reward": 0.010555490205661044,
      "rewards/forensics_tiou_reward": 0.32350872457027435,
      "step": 127
    },
    {
      "completion_length": 23.53125,
      "epoch": 0.3932411674347158,
      "grad_norm": 5.402480125427246,
      "kl": 0.2705078125,
      "learning_rate": 9.023076923076922e-07,
      "loss": 0.1443,
      "reward": 0.7196123003959656,
      "reward_std": 0.09300026297569275,
      "reward_std_all": 0.08966567739844322,
      "rewards-max/forensics_sinkhorn_reward": 0.9723354578018188,
      "rewards-max/forensics_strict_edge_reward": 0.12943623587489128,
      "rewards-max/forensics_tiou_reward": 0.9259933233261108,
      "rewards/forensics_sinkhorn_reward": 0.6553561091423035,
      "rewards/forensics_strict_edge_reward": 0.012309343321248889,
      "rewards/forensics_tiou_reward": 0.3730026185512543,
      "step": 128
    },
    {
      "completion_length": 22.9375,
      "epoch": 0.39631336405529954,
      "grad_norm": 3.8285226821899414,
      "kl": 0.349609375,
      "learning_rate": 9.015384615384614e-07,
      "loss": 0.0544,
      "reward": 0.7000433802604675,
      "reward_std": 0.07900070585310459,
      "reward_std_all": 0.09184487536549568,
      "rewards-max/forensics_sinkhorn_reward": 0.8383587896823883,
      "rewards-max/forensics_strict_edge_reward": 0.25064099044539034,
      "rewards-max/forensics_tiou_reward": 0.6331994235515594,
      "rewards/forensics_sinkhorn_reward": 0.597316712141037,
      "rewards/forensics_strict_edge_reward": 0.011212840159714688,
      "rewards/forensics_tiou_reward": 0.26716138422489166,
      "step": 129
    },
    {
      "completion_length": 22.828125,
      "epoch": 0.39938556067588327,
      "grad_norm": 4.4456400871276855,
      "kl": 0.26953125,
      "learning_rate": 9.007692307692307e-07,
      "loss": 0.0424,
      "reward": 0.6828520596027374,
      "reward_std": 0.09351414814591408,
      "reward_std_all": 0.10638725012540817,
      "rewards-max/forensics_sinkhorn_reward": 0.8766602575778961,
      "rewards-max/forensics_strict_edge_reward": 0.025485567981377244,
      "rewards-max/forensics_tiou_reward": 0.6985976696014404,
      "rewards/forensics_sinkhorn_reward": 0.5731269419193268,
      "rewards/forensics_strict_edge_reward": 0.0018630536342243431,
      "rewards/forensics_tiou_reward": 0.2420588582754135,
      "step": 130
    },
    {
      "completion_length": 24.546875,
      "epoch": 0.402457757296467,
      "grad_norm": 4.56884765625,
      "kl": 0.24755859375,
      "learning_rate": 9e-07,
      "loss": 0.1358,
      "reward": 0.7248361110687256,
      "reward_std": 0.05778038874268532,
      "reward_std_all": 0.06923318468034267,
      "rewards-max/forensics_sinkhorn_reward": 0.9501520991325378,
      "rewards-max/forensics_strict_edge_reward": 0.4689457108033821,
      "rewards-max/forensics_tiou_reward": 0.8578184843063354,
      "rewards/forensics_sinkhorn_reward": 0.6212911605834961,
      "rewards/forensics_strict_edge_reward": 0.019669226407131646,
      "rewards/forensics_tiou_reward": 0.3344137817621231,
      "step": 131
    },
    {
      "completion_length": 24.015625,
      "epoch": 0.4055299539170507,
      "grad_norm": 5.617086410522461,
      "kl": 0.24560546875,
      "learning_rate": 8.992307692307693e-07,
      "loss": 0.22,
      "reward": 0.6154589354991913,
      "reward_std": 0.1573341339826584,
      "reward_std_all": 0.17037751525640488,
      "rewards-max/forensics_sinkhorn_reward": 0.94552943110466,
      "rewards-max/forensics_strict_edge_reward": 0.16226347163319588,
      "rewards-max/forensics_tiou_reward": 0.8603616952896118,
      "rewards/forensics_sinkhorn_reward": 0.5157815515995026,
      "rewards/forensics_strict_edge_reward": 0.011111680301837623,
      "rewards/forensics_tiou_reward": 0.25274065136909485,
      "step": 132
    },
    {
      "completion_length": 26.421875,
      "epoch": 0.40860215053763443,
      "grad_norm": 4.30948543548584,
      "kl": 0.18994140625,
      "learning_rate": 8.984615384615384e-07,
      "loss": 0.1913,
      "reward": 0.6428159475326538,
      "reward_std": 0.10600744560360909,
      "reward_std_all": 0.12937723100185394,
      "rewards-max/forensics_sinkhorn_reward": 0.9266071021556854,
      "rewards-max/forensics_strict_edge_reward": 0.2688929031137377,
      "rewards-max/forensics_tiou_reward": 0.8290423452854156,
      "rewards/forensics_sinkhorn_reward": 0.5401527285575867,
      "rewards/forensics_strict_edge_reward": 0.013628619475639425,
      "rewards/forensics_tiou_reward": 0.25042008608579636,
      "step": 133
    },
    {
      "completion_length": 24.296875,
      "epoch": 0.4116743471582181,
      "grad_norm": 4.44610071182251,
      "kl": 0.2080078125,
      "learning_rate": 8.976923076923076e-07,
      "loss": 0.0719,
      "reward": 0.7058161199092865,
      "reward_std": 0.12398567423224449,
      "reward_std_all": 0.11568132042884827,
      "rewards-max/forensics_sinkhorn_reward": 0.947058230638504,
      "rewards-max/forensics_strict_edge_reward": 0.13320930674672127,
      "rewards-max/forensics_tiou_reward": 0.860633909702301,
      "rewards/forensics_sinkhorn_reward": 0.608464241027832,
      "rewards/forensics_strict_edge_reward": 0.007455366081558168,
      "rewards/forensics_tiou_reward": 0.32705311477184296,
      "step": 134
    },
    {
      "completion_length": 22.453125,
      "epoch": 0.4147465437788018,
      "grad_norm": 4.7316412925720215,
      "kl": 0.23974609375,
      "learning_rate": 8.969230769230769e-07,
      "loss": 0.156,
      "reward": 0.6707422137260437,
      "reward_std": 0.07622857578098774,
      "reward_std_all": 0.10293320193886757,
      "rewards-max/forensics_sinkhorn_reward": 0.96274134516716,
      "rewards-max/forensics_strict_edge_reward": 0.1383604295551777,
      "rewards-max/forensics_tiou_reward": 0.9098219573497772,
      "rewards/forensics_sinkhorn_reward": 0.6067312955856323,
      "rewards/forensics_strict_edge_reward": 0.006948464782908559,
      "rewards/forensics_tiou_reward": 0.34147903323173523,
      "step": 135
    },
    {
      "completion_length": 25.140625,
      "epoch": 0.41781874039938555,
      "grad_norm": 3.7157399654388428,
      "kl": 0.23193359375,
      "learning_rate": 8.961538461538461e-07,
      "loss": 0.1952,
      "reward": 0.6918184459209442,
      "reward_std": 0.103420939296484,
      "reward_std_all": 0.10832478106021881,
      "rewards-max/forensics_sinkhorn_reward": 0.9140854179859161,
      "rewards-max/forensics_strict_edge_reward": 0.11565752979367971,
      "rewards-max/forensics_tiou_reward": 0.7978079319000244,
      "rewards/forensics_sinkhorn_reward": 0.5882527232170105,
      "rewards/forensics_strict_edge_reward": 0.00496724116965197,
      "rewards/forensics_tiou_reward": 0.3183724880218506,
      "step": 136
    },
    {
      "completion_length": 23.328125,
      "epoch": 0.42089093701996927,
      "grad_norm": 4.806955337524414,
      "kl": 0.2197265625,
      "learning_rate": 8.953846153846152e-07,
      "loss": 0.1011,
      "reward": 0.6985369026660919,
      "reward_std": 0.12122054770588875,
      "reward_std_all": 0.11615178361535072,
      "rewards-max/forensics_sinkhorn_reward": 0.9687458872795105,
      "rewards-max/forensics_strict_edge_reward": 0.04717296632588841,
      "rewards-max/forensics_tiou_reward": 0.8986740112304688,
      "rewards/forensics_sinkhorn_reward": 0.6085841655731201,
      "rewards/forensics_strict_edge_reward": 0.0034020572620647727,
      "rewards/forensics_tiou_reward": 0.33463431894779205,
      "step": 137
    },
    {
      "completion_length": 23.203125,
      "epoch": 0.423963133640553,
      "grad_norm": 3.943673610687256,
      "kl": 0.22314453125,
      "learning_rate": 8.946153846153846e-07,
      "loss": 0.1366,
      "reward": 0.6322506666183472,
      "reward_std": 0.08917439728975296,
      "reward_std_all": 0.12745489180088043,
      "rewards-max/forensics_sinkhorn_reward": 0.9417715072631836,
      "rewards-max/forensics_strict_edge_reward": 0.022110791062004864,
      "rewards-max/forensics_tiou_reward": 0.8608191013336182,
      "rewards/forensics_sinkhorn_reward": 0.5357138812541962,
      "rewards/forensics_strict_edge_reward": 0.0009653703309595585,
      "rewards/forensics_tiou_reward": 0.22606366127729416,
      "step": 138
    },
    {
      "completion_length": 23.640625,
      "epoch": 0.4270353302611367,
      "grad_norm": 4.481675624847412,
      "kl": 0.20703125,
      "learning_rate": 8.938461538461538e-07,
      "loss": 0.1738,
      "reward": 0.6940003335475922,
      "reward_std": 0.05614432319998741,
      "reward_std_all": 0.08237510547041893,
      "rewards-max/forensics_sinkhorn_reward": 0.9601886868476868,
      "rewards-max/forensics_strict_edge_reward": 0.3033798933029175,
      "rewards-max/forensics_tiou_reward": 0.8876310586929321,
      "rewards/forensics_sinkhorn_reward": 0.6066888272762299,
      "rewards/forensics_strict_edge_reward": 0.015464129857718945,
      "rewards/forensics_tiou_reward": 0.3220823407173157,
      "step": 139
    },
    {
      "completion_length": 23.890625,
      "epoch": 0.43010752688172044,
      "grad_norm": 5.5614752769470215,
      "kl": 0.26708984375,
      "learning_rate": 8.930769230769231e-07,
      "loss": 0.171,
      "reward": 0.6966561079025269,
      "reward_std": 0.10907620564103127,
      "reward_std_all": 0.10680127516388893,
      "rewards-max/forensics_sinkhorn_reward": 0.9597806036472321,
      "rewards-max/forensics_strict_edge_reward": 0.297326996922493,
      "rewards-max/forensics_tiou_reward": 0.8957400918006897,
      "rewards/forensics_sinkhorn_reward": 0.6105474531650543,
      "rewards/forensics_strict_edge_reward": 0.0199437215924263,
      "rewards/forensics_tiou_reward": 0.3221277892589569,
      "step": 140
    },
    {
      "completion_length": 23.40625,
      "epoch": 0.43317972350230416,
      "grad_norm": 4.380975246429443,
      "kl": 0.2353515625,
      "learning_rate": 8.923076923076923e-07,
      "loss": 0.1691,
      "reward": 0.6950605511665344,
      "reward_std": 0.08884712681174278,
      "reward_std_all": 0.09721125662326813,
      "rewards-max/forensics_sinkhorn_reward": 0.9782198667526245,
      "rewards-max/forensics_strict_edge_reward": 0.04806019621901214,
      "rewards-max/forensics_tiou_reward": 0.9331243634223938,
      "rewards/forensics_sinkhorn_reward": 0.5920431017875671,
      "rewards/forensics_strict_edge_reward": 0.0028479790780693293,
      "rewards/forensics_tiou_reward": 0.32815253734588623,
      "step": 141
    },
    {
      "completion_length": 23.21875,
      "epoch": 0.4362519201228879,
      "grad_norm": 4.639754295349121,
      "kl": 0.26416015625,
      "learning_rate": 8.915384615384614e-07,
      "loss": 0.0911,
      "reward": 0.7425923943519592,
      "reward_std": 0.07787330076098442,
      "reward_std_all": 0.07368751429021358,
      "rewards-max/forensics_sinkhorn_reward": 0.9823223352432251,
      "rewards-max/forensics_strict_edge_reward": 0.36977964639663696,
      "rewards-max/forensics_tiou_reward": 0.9535254836082458,
      "rewards/forensics_sinkhorn_reward": 0.6752426326274872,
      "rewards/forensics_strict_edge_reward": 0.01594983274117112,
      "rewards/forensics_tiou_reward": 0.37758007645606995,
      "step": 142
    },
    {
      "completion_length": 23.453125,
      "epoch": 0.4393241167434716,
      "grad_norm": 4.190910816192627,
      "kl": 0.23046875,
      "learning_rate": 8.907692307692307e-07,
      "loss": 0.242,
      "reward": 0.7108581364154816,
      "reward_std": 0.07340349815785885,
      "reward_std_all": 0.08348913863301277,
      "rewards-max/forensics_sinkhorn_reward": 0.9465740919113159,
      "rewards-max/forensics_strict_edge_reward": 0.3836973309516907,
      "rewards-max/forensics_tiou_reward": 0.8733769655227661,
      "rewards/forensics_sinkhorn_reward": 0.602129727602005,
      "rewards/forensics_strict_edge_reward": 0.02461184933781624,
      "rewards/forensics_tiou_reward": 0.3137497007846832,
      "step": 143
    },
    {
      "completion_length": 22.03125,
      "epoch": 0.4423963133640553,
      "grad_norm": 4.92557954788208,
      "kl": 0.21630859375,
      "learning_rate": 8.9e-07,
      "loss": 0.0059,
      "reward": 0.7382743954658508,
      "reward_std": 0.06565627083182335,
      "reward_std_all": 0.06610023975372314,
      "rewards-max/forensics_sinkhorn_reward": 0.9732181131839752,
      "rewards-max/forensics_strict_edge_reward": 0.2678813338279724,
      "rewards-max/forensics_tiou_reward": 0.9264658689498901,
      "rewards/forensics_sinkhorn_reward": 0.6533842086791992,
      "rewards/forensics_strict_edge_reward": 0.0230034738779068,
      "rewards/forensics_tiou_reward": 0.33305831253528595,
      "step": 144
    },
    {
      "completion_length": 25.09375,
      "epoch": 0.445468509984639,
      "grad_norm": 4.436679840087891,
      "kl": 0.26220703125,
      "learning_rate": 8.892307692307693e-07,
      "loss": 0.1319,
      "reward": 0.7067385315895081,
      "reward_std": 0.10623078793287277,
      "reward_std_all": 0.10308923944830894,
      "rewards-max/forensics_sinkhorn_reward": 0.9815090000629425,
      "rewards-max/forensics_strict_edge_reward": 0.07771316915750504,
      "rewards-max/forensics_tiou_reward": 0.9559069275856018,
      "rewards/forensics_sinkhorn_reward": 0.6443244814872742,
      "rewards/forensics_strict_edge_reward": 0.0056918051559478045,
      "rewards/forensics_tiou_reward": 0.3938632309436798,
      "step": 145
    },
    {
      "completion_length": 24.140625,
      "epoch": 0.4485407066052227,
      "grad_norm": 3.6208548545837402,
      "kl": 0.20361328125,
      "learning_rate": 8.884615384615384e-07,
      "loss": 0.0731,
      "reward": 0.6957501471042633,
      "reward_std": 0.07268349453806877,
      "reward_std_all": 0.09121785685420036,
      "rewards-max/forensics_sinkhorn_reward": 0.9783385396003723,
      "rewards-max/forensics_strict_edge_reward": 0.28010551631450653,
      "rewards-max/forensics_tiou_reward": 0.9451811611652374,
      "rewards/forensics_sinkhorn_reward": 0.5998196303844452,
      "rewards/forensics_strict_edge_reward": 0.02538702543824911,
      "rewards/forensics_tiou_reward": 0.33268043398857117,
      "step": 146
    },
    {
      "completion_length": 21.3125,
      "epoch": 0.45161290322580644,
      "grad_norm": 5.834578990936279,
      "kl": 0.21826171875,
      "learning_rate": 8.876923076923076e-07,
      "loss": 0.1465,
      "reward": 0.7179538607597351,
      "reward_std": 0.0769749041646719,
      "reward_std_all": 0.08218056708574295,
      "rewards-max/forensics_sinkhorn_reward": 0.9626390337944031,
      "rewards-max/forensics_strict_edge_reward": 0.0852483157068491,
      "rewards-max/forensics_tiou_reward": 0.9096929728984833,
      "rewards/forensics_sinkhorn_reward": 0.6471675634384155,
      "rewards/forensics_strict_edge_reward": 0.005664842145051807,
      "rewards/forensics_tiou_reward": 0.3458715081214905,
      "step": 147
    },
    {
      "completion_length": 23.84375,
      "epoch": 0.45468509984639016,
      "grad_norm": 4.493556976318359,
      "kl": 0.23974609375,
      "learning_rate": 8.869230769230769e-07,
      "loss": 0.1746,
      "reward": 0.718451201915741,
      "reward_std": 0.08239182084798813,
      "reward_std_all": 0.08669078350067139,
      "rewards-max/forensics_sinkhorn_reward": 0.9707462787628174,
      "rewards-max/forensics_strict_edge_reward": 0.4043564051389694,
      "rewards-max/forensics_tiou_reward": 0.9333576858043671,
      "rewards/forensics_sinkhorn_reward": 0.62010857462883,
      "rewards/forensics_strict_edge_reward": 0.017938916571438313,
      "rewards/forensics_tiou_reward": 0.32624559104442596,
      "step": 148
    },
    {
      "completion_length": 23.375,
      "epoch": 0.4577572964669739,
      "grad_norm": 4.366865158081055,
      "kl": 0.2353515625,
      "learning_rate": 8.861538461538461e-07,
      "loss": 0.0955,
      "reward": 0.7322458922863007,
      "reward_std": 0.07764365151524544,
      "reward_std_all": 0.07579059898853302,
      "rewards-max/forensics_sinkhorn_reward": 0.9900864660739899,
      "rewards-max/forensics_strict_edge_reward": 0.2683865576982498,
      "rewards-max/forensics_tiou_reward": 0.9783221483230591,
      "rewards/forensics_sinkhorn_reward": 0.6344064772129059,
      "rewards/forensics_strict_edge_reward": 0.017346458043903112,
      "rewards/forensics_tiou_reward": 0.3654249459505081,
      "step": 149
    },
    {
      "completion_length": 24.078125,
      "epoch": 0.4608294930875576,
      "grad_norm": 3.9880828857421875,
      "kl": 0.224609375,
      "learning_rate": 8.853846153846154e-07,
      "loss": 0.1572,
      "reward": 0.7250038683414459,
      "reward_std": 0.07541700452566147,
      "reward_std_all": 0.07721010223031044,
      "rewards-max/forensics_sinkhorn_reward": 0.9819718599319458,
      "rewards-max/forensics_strict_edge_reward": 0.09718754142522812,
      "rewards-max/forensics_tiou_reward": 0.9580458700656891,
      "rewards/forensics_sinkhorn_reward": 0.6279492080211639,
      "rewards/forensics_strict_edge_reward": 0.007427396951243281,
      "rewards/forensics_tiou_reward": 0.35544082522392273,
      "step": 150
    },
    {
      "completion_length": 22.890625,
      "epoch": 0.46390168970814133,
      "grad_norm": 3.8852884769439697,
      "kl": 0.2265625,
      "learning_rate": 8.846153846153846e-07,
      "loss": 0.0358,
      "reward": 0.6989790499210358,
      "reward_std": 0.06481604278087616,
      "reward_std_all": 0.08446284011006355,
      "rewards-max/forensics_sinkhorn_reward": 0.8903006911277771,
      "rewards-max/forensics_strict_edge_reward": 0.33930318569764495,
      "rewards-max/forensics_tiou_reward": 0.7874097228050232,
      "rewards/forensics_sinkhorn_reward": 0.6075411736965179,
      "rewards/forensics_strict_edge_reward": 0.015471833845367655,
      "rewards/forensics_tiou_reward": 0.29260119795799255,
      "step": 151
    },
    {
      "completion_length": 21.765625,
      "epoch": 0.46697388632872505,
      "grad_norm": 4.790369510650635,
      "kl": 0.224609375,
      "learning_rate": 8.838461538461538e-07,
      "loss": 0.169,
      "reward": 0.7444172203540802,
      "reward_std": 0.06740911863744259,
      "reward_std_all": 0.06541113182902336,
      "rewards-max/forensics_sinkhorn_reward": 0.9676911234855652,
      "rewards-max/forensics_strict_edge_reward": 0.3065662086009979,
      "rewards-max/forensics_tiou_reward": 0.9150961637496948,
      "rewards/forensics_sinkhorn_reward": 0.6536272764205933,
      "rewards/forensics_strict_edge_reward": 0.01818910613656044,
      "rewards/forensics_tiou_reward": 0.35784509778022766,
      "step": 152
    },
    {
      "completion_length": 22.515625,
      "epoch": 0.4700460829493088,
      "grad_norm": 4.1206207275390625,
      "kl": 0.30517578125,
      "learning_rate": 8.830769230769231e-07,
      "loss": 0.1144,
      "reward": 0.7220687568187714,
      "reward_std": 0.09270121809095144,
      "reward_std_all": 0.0890442542731762,
      "rewards-max/forensics_sinkhorn_reward": 0.9831302464008331,
      "rewards-max/forensics_strict_edge_reward": 0.35029198229312897,
      "rewards-max/forensics_tiou_reward": 0.9580205380916595,
      "rewards/forensics_sinkhorn_reward": 0.6339473724365234,
      "rewards/forensics_strict_edge_reward": 0.03007927443832159,
      "rewards/forensics_tiou_reward": 0.3888643980026245,
      "step": 153
    },
    {
      "completion_length": 22.109375,
      "epoch": 0.4731182795698925,
      "grad_norm": 4.4416823387146,
      "kl": 0.2666015625,
      "learning_rate": 8.823076923076923e-07,
      "loss": 0.0935,
      "reward": 0.7247555553913116,
      "reward_std": 0.07036376371979713,
      "reward_std_all": 0.07449117675423622,
      "rewards-max/forensics_sinkhorn_reward": 0.9029477834701538,
      "rewards-max/forensics_strict_edge_reward": 0.0986202274216339,
      "rewards-max/forensics_tiou_reward": 0.768364667892456,
      "rewards/forensics_sinkhorn_reward": 0.6296150088310242,
      "rewards/forensics_strict_edge_reward": 0.011133480824355502,
      "rewards/forensics_tiou_reward": 0.357718825340271,
      "step": 154
    },
    {
      "completion_length": 23.5,
      "epoch": 0.47619047619047616,
      "grad_norm": 4.343717098236084,
      "kl": 0.22900390625,
      "learning_rate": 8.815384615384614e-07,
      "loss": 0.1293,
      "reward": 0.7506407797336578,
      "reward_std": 0.05753517895936966,
      "reward_std_all": 0.056940166279673576,
      "rewards-max/forensics_sinkhorn_reward": 0.9717722833156586,
      "rewards-max/forensics_strict_edge_reward": 0.4927365332841873,
      "rewards-max/forensics_tiou_reward": 0.9181863963603973,
      "rewards/forensics_sinkhorn_reward": 0.6670053005218506,
      "rewards/forensics_strict_edge_reward": 0.03171866573393345,
      "rewards/forensics_tiou_reward": 0.37393203377723694,
      "step": 155
    },
    {
      "completion_length": 22.15625,
      "epoch": 0.4792626728110599,
      "grad_norm": 5.994975566864014,
      "kl": 0.2705078125,
      "learning_rate": 8.807692307692307e-07,
      "loss": 0.194,
      "reward": 0.6592794954776764,
      "reward_std": 0.08895191177725792,
      "reward_std_all": 0.11657635495066643,
      "rewards-max/forensics_sinkhorn_reward": 0.9565739035606384,
      "rewards-max/forensics_strict_edge_reward": 0.17340171337127686,
      "rewards-max/forensics_tiou_reward": 0.8565451800823212,
      "rewards/forensics_sinkhorn_reward": 0.5836188793182373,
      "rewards/forensics_strict_edge_reward": 0.012786499777575955,
      "rewards/forensics_tiou_reward": 0.2854501083493233,
      "step": 156
    },
    {
      "completion_length": 23.453125,
      "epoch": 0.4823348694316436,
      "grad_norm": 4.1767425537109375,
      "kl": 0.20166015625,
      "learning_rate": 8.799999999999999e-07,
      "loss": -0.0004,
      "reward": 0.7369997799396515,
      "reward_std": 0.07462213188409805,
      "reward_std_all": 0.0732230395078659,
      "rewards-max/forensics_sinkhorn_reward": 0.9823203086853027,
      "rewards-max/forensics_strict_edge_reward": 0.4641416519880295,
      "rewards-max/forensics_tiou_reward": 0.9575460255146027,
      "rewards/forensics_sinkhorn_reward": 0.6681732535362244,
      "rewards/forensics_strict_edge_reward": 0.03217455931007862,
      "rewards/forensics_tiou_reward": 0.44298669695854187,
      "step": 157
    },
    {
      "completion_length": 25.375,
      "epoch": 0.48540706605222733,
      "grad_norm": 5.338266849517822,
      "kl": 0.318359375,
      "learning_rate": 8.792307692307693e-07,
      "loss": 0.1092,
      "reward": 0.7090220749378204,
      "reward_std": 0.09170323610305786,
      "reward_std_all": 0.09395529702305794,
      "rewards-max/forensics_sinkhorn_reward": 0.9487894177436829,
      "rewards-max/forensics_strict_edge_reward": 0.19250920647755265,
      "rewards-max/forensics_tiou_reward": 0.8634896278381348,
      "rewards/forensics_sinkhorn_reward": 0.6033979654312134,
      "rewards/forensics_strict_edge_reward": 0.017525175178889185,
      "rewards/forensics_tiou_reward": 0.3323701322078705,
      "step": 158
    },
    {
      "completion_length": 24.8125,
      "epoch": 0.48847926267281105,
      "grad_norm": 4.802734375,
      "kl": 0.2783203125,
      "learning_rate": 8.784615384615385e-07,
      "loss": 0.203,
      "reward": 0.6980876624584198,
      "reward_std": 0.09890275076031685,
      "reward_std_all": 0.10101747140288353,
      "rewards-max/forensics_sinkhorn_reward": 0.9721401035785675,
      "rewards-max/forensics_strict_edge_reward": 0.33664941787719727,
      "rewards-max/forensics_tiou_reward": 0.9325580894947052,
      "rewards/forensics_sinkhorn_reward": 0.5902624726295471,
      "rewards/forensics_strict_edge_reward": 0.019311350770294666,
      "rewards/forensics_tiou_reward": 0.34504224359989166,
      "step": 159
    },
    {
      "completion_length": 24.1875,
      "epoch": 0.4915514592933948,
      "grad_norm": 4.681209087371826,
      "kl": 0.31298828125,
      "learning_rate": 8.776923076923076e-07,
      "loss": 0.2028,
      "reward": 0.694551408290863,
      "reward_std": 0.10716735944151878,
      "reward_std_all": 0.11182663589715958,
      "rewards-max/forensics_sinkhorn_reward": 0.9433127641677856,
      "rewards-max/forensics_strict_edge_reward": 0.3139570940693375,
      "rewards-max/forensics_tiou_reward": 0.8594982326030731,
      "rewards/forensics_sinkhorn_reward": 0.5871033370494843,
      "rewards/forensics_strict_edge_reward": 0.014245882088289363,
      "rewards/forensics_tiou_reward": 0.2969244420528412,
      "step": 160
    },
    {
      "completion_length": 23.828125,
      "epoch": 0.4946236559139785,
      "grad_norm": 4.630343914031982,
      "kl": 0.21875,
      "learning_rate": 8.769230769230769e-07,
      "loss": 0.211,
      "reward": 0.6970683336257935,
      "reward_std": 0.09248972684144974,
      "reward_std_all": 0.102482620626688,
      "rewards-max/forensics_sinkhorn_reward": 0.895233154296875,
      "rewards-max/forensics_strict_edge_reward": 0.034262655302882195,
      "rewards-max/forensics_tiou_reward": 0.7327714562416077,
      "rewards/forensics_sinkhorn_reward": 0.58126100897789,
      "rewards/forensics_strict_edge_reward": 0.001987723633646965,
      "rewards/forensics_tiou_reward": 0.3113214075565338,
      "step": 161
    },
    {
      "completion_length": 24.421875,
      "epoch": 0.4976958525345622,
      "grad_norm": 6.10003137588501,
      "kl": 0.265625,
      "learning_rate": 8.761538461538461e-07,
      "loss": 0.2193,
      "reward": 0.7148627638816833,
      "reward_std": 0.09669182822108269,
      "reward_std_all": 0.09407962113618851,
      "rewards-max/forensics_sinkhorn_reward": 0.9779216945171356,
      "rewards-max/forensics_strict_edge_reward": 0.5410424992442131,
      "rewards-max/forensics_tiou_reward": 0.9463993310928345,
      "rewards/forensics_sinkhorn_reward": 0.6036443412303925,
      "rewards/forensics_strict_edge_reward": 0.026242372579872608,
      "rewards/forensics_tiou_reward": 0.3802211582660675,
      "step": 162
    },
    {
      "completion_length": 23.140625,
      "epoch": 0.500768049155146,
      "grad_norm": 5.4705047607421875,
      "kl": 0.23828125,
      "learning_rate": 8.753846153846154e-07,
      "loss": 0.149,
      "reward": 0.663455992937088,
      "reward_std": 0.08132778853178024,
      "reward_std_all": 0.10576692968606949,
      "rewards-max/forensics_sinkhorn_reward": 0.8815431296825409,
      "rewards-max/forensics_strict_edge_reward": 0.012365056900307536,
      "rewards-max/forensics_tiou_reward": 0.6991561651229858,
      "rewards/forensics_sinkhorn_reward": 0.5597411394119263,
      "rewards/forensics_strict_edge_reward": 0.0005780719075119123,
      "rewards/forensics_tiou_reward": 0.21893252432346344,
      "step": 163
    },
    {
      "completion_length": 23.640625,
      "epoch": 0.5038402457757296,
      "grad_norm": 6.615388870239258,
      "kl": 0.390625,
      "learning_rate": 8.746153846153845e-07,
      "loss": 0.0222,
      "reward": 0.7178822755813599,
      "reward_std": 0.0951346606016159,
      "reward_std_all": 0.09349283203482628,
      "rewards-max/forensics_sinkhorn_reward": 0.8977520763874054,
      "rewards-max/forensics_strict_edge_reward": 0.06404319999273866,
      "rewards-max/forensics_tiou_reward": 0.7276570796966553,
      "rewards/forensics_sinkhorn_reward": 0.6223095059394836,
      "rewards/forensics_strict_edge_reward": 0.00321434905345086,
      "rewards/forensics_tiou_reward": 0.2845940887928009,
      "step": 164
    },
    {
      "completion_length": 23.4375,
      "epoch": 0.5069124423963134,
      "grad_norm": 5.520002841949463,
      "kl": 0.263671875,
      "learning_rate": 8.738461538461538e-07,
      "loss": 0.1128,
      "reward": 0.7092176973819733,
      "reward_std": 0.09039541520178318,
      "reward_std_all": 0.09068349003791809,
      "rewards-max/forensics_sinkhorn_reward": 0.9809052646160126,
      "rewards-max/forensics_strict_edge_reward": 0.19213168788701296,
      "rewards-max/forensics_tiou_reward": 0.9500034749507904,
      "rewards/forensics_sinkhorn_reward": 0.6461016237735748,
      "rewards/forensics_strict_edge_reward": 0.01926516811363399,
      "rewards/forensics_tiou_reward": 0.4246707856655121,
      "step": 165
    },
    {
      "completion_length": 24.5,
      "epoch": 0.5099846390168971,
      "grad_norm": 6.423213005065918,
      "kl": 0.2529296875,
      "learning_rate": 8.730769230769231e-07,
      "loss": 0.191,
      "reward": 0.70113605260849,
      "reward_std": 0.06328256614506245,
      "reward_std_all": 0.08109109848737717,
      "rewards-max/forensics_sinkhorn_reward": 0.9759975075721741,
      "rewards-max/forensics_strict_edge_reward": 0.42822691053152084,
      "rewards-max/forensics_tiou_reward": 0.9412090182304382,
      "rewards/forensics_sinkhorn_reward": 0.6042493879795074,
      "rewards/forensics_strict_edge_reward": 0.02756571676582098,
      "rewards/forensics_tiou_reward": 0.32871347665786743,
      "step": 166
    },
    {
      "completion_length": 22.703125,
      "epoch": 0.5130568356374808,
      "grad_norm": 5.74674654006958,
      "kl": 0.44921875,
      "learning_rate": 8.723076923076923e-07,
      "loss": 0.0544,
      "reward": 0.7336578071117401,
      "reward_std": 0.0898175835609436,
      "reward_std_all": 0.08395221456885338,
      "rewards-max/forensics_sinkhorn_reward": 0.9704026579856873,
      "rewards-max/forensics_strict_edge_reward": 0.35358040407299995,
      "rewards-max/forensics_tiou_reward": 0.913949728012085,
      "rewards/forensics_sinkhorn_reward": 0.6623841524124146,
      "rewards/forensics_strict_edge_reward": 0.015631193993613124,
      "rewards/forensics_tiou_reward": 0.39219848811626434,
      "step": 167
    },
    {
      "completion_length": 24.046875,
      "epoch": 0.5161290322580645,
      "grad_norm": 4.628568649291992,
      "kl": 0.2880859375,
      "learning_rate": 8.715384615384614e-07,
      "loss": 0.1932,
      "reward": 0.6890405416488647,
      "reward_std": 0.06537915393710136,
      "reward_std_all": 0.088071184232831,
      "rewards-max/forensics_sinkhorn_reward": 0.9859272241592407,
      "rewards-max/forensics_strict_edge_reward": 0.089108906686306,
      "rewards-max/forensics_tiou_reward": 0.9635766744613647,
      "rewards/forensics_sinkhorn_reward": 0.6103743016719818,
      "rewards/forensics_strict_edge_reward": 0.0055583142675459385,
      "rewards/forensics_tiou_reward": 0.34986262023448944,
      "step": 168
    },
    {
      "completion_length": 23.0625,
      "epoch": 0.5192012288786483,
      "grad_norm": 4.627744197845459,
      "kl": 0.275390625,
      "learning_rate": 8.707692307692307e-07,
      "loss": 0.0126,
      "reward": 0.7414408326148987,
      "reward_std": 0.05311833042651415,
      "reward_std_all": 0.06041766796261072,
      "rewards-max/forensics_sinkhorn_reward": 0.8878768384456635,
      "rewards-max/forensics_strict_edge_reward": 0.10799140675226226,
      "rewards-max/forensics_tiou_reward": 0.7533958852291107,
      "rewards/forensics_sinkhorn_reward": 0.634843647480011,
      "rewards/forensics_strict_edge_reward": 0.0049491662884975085,
      "rewards/forensics_tiou_reward": 0.30276811122894287,
      "step": 169
    },
    {
      "completion_length": 24.5,
      "epoch": 0.522273425499232,
      "grad_norm": 4.899086952209473,
      "kl": 0.2802734375,
      "learning_rate": 8.699999999999999e-07,
      "loss": 0.178,
      "reward": 0.7416225969791412,
      "reward_std": 0.029588835313916206,
      "reward_std_all": 0.04714537225663662,
      "rewards-max/forensics_sinkhorn_reward": 0.9603225886821747,
      "rewards-max/forensics_strict_edge_reward": 0.0881829485297203,
      "rewards-max/forensics_tiou_reward": 0.8918703198432922,
      "rewards/forensics_sinkhorn_reward": 0.6447297036647797,
      "rewards/forensics_strict_edge_reward": 0.004016506252810359,
      "rewards/forensics_tiou_reward": 0.3347989469766617,
      "step": 170
    },
    {
      "completion_length": 25.40625,
      "epoch": 0.5253456221198156,
      "grad_norm": 4.307405471801758,
      "kl": 0.248046875,
      "learning_rate": 8.692307692307692e-07,
      "loss": 0.0615,
      "reward": 0.6554782688617706,
      "reward_std": 0.09174076095223427,
      "reward_std_all": 0.11497733741998672,
      "rewards-max/forensics_sinkhorn_reward": 0.9336195290088654,
      "rewards-max/forensics_strict_edge_reward": 0.46804800629615784,
      "rewards-max/forensics_tiou_reward": 0.839166909456253,
      "rewards/forensics_sinkhorn_reward": 0.5447595566511154,
      "rewards/forensics_strict_edge_reward": 0.019667600747197866,
      "rewards/forensics_tiou_reward": 0.2705545127391815,
      "step": 171
    },
    {
      "completion_length": 22.59375,
      "epoch": 0.5284178187403994,
      "grad_norm": 5.038961410522461,
      "kl": 0.291015625,
      "learning_rate": 8.684615384615385e-07,
      "loss": 0.0602,
      "reward": 0.7450750172138214,
      "reward_std": 0.06112802401185036,
      "reward_std_all": 0.06085706874728203,
      "rewards-max/forensics_sinkhorn_reward": 0.9599727094173431,
      "rewards-max/forensics_strict_edge_reward": 0.38316231514909305,
      "rewards-max/forensics_tiou_reward": 0.8989139199256897,
      "rewards/forensics_sinkhorn_reward": 0.6748868823051453,
      "rewards/forensics_strict_edge_reward": 0.022243506778977462,
      "rewards/forensics_tiou_reward": 0.3838537037372589,
      "step": 172
    },
    {
      "completion_length": 22.171875,
      "epoch": 0.5314900153609831,
      "grad_norm": 4.085600852966309,
      "kl": 0.263671875,
      "learning_rate": 8.676923076923076e-07,
      "loss": 0.115,
      "reward": 0.7364097535610199,
      "reward_std": 0.049608493223786354,
      "reward_std_all": 0.06239877920597792,
      "rewards-max/forensics_sinkhorn_reward": 0.9582299888134003,
      "rewards-max/forensics_strict_edge_reward": 0.3959764545033977,
      "rewards-max/forensics_tiou_reward": 0.893132746219635,
      "rewards/forensics_sinkhorn_reward": 0.6614158749580383,
      "rewards/forensics_strict_edge_reward": 0.030358231265836366,
      "rewards/forensics_tiou_reward": 0.3966827839612961,
      "step": 173
    },
    {
      "completion_length": 23.9375,
      "epoch": 0.5345622119815668,
      "grad_norm": 4.961109638214111,
      "kl": 0.23974609375,
      "learning_rate": 8.669230769230769e-07,
      "loss": 0.0295,
      "reward": 0.7481094002723694,
      "reward_std": 0.0610380657017231,
      "reward_std_all": 0.06200805678963661,
      "rewards-max/forensics_sinkhorn_reward": 0.9343296587467194,
      "rewards-max/forensics_strict_edge_reward": 0.3200395703315735,
      "rewards-max/forensics_tiou_reward": 0.8375925421714783,
      "rewards/forensics_sinkhorn_reward": 0.6603038907051086,
      "rewards/forensics_strict_edge_reward": 0.02426843298599124,
      "rewards/forensics_tiou_reward": 0.3900272399187088,
      "step": 174
    },
    {
      "completion_length": 24.28125,
      "epoch": 0.5376344086021505,
      "grad_norm": 4.787155628204346,
      "kl": 0.31640625,
      "learning_rate": 8.661538461538461e-07,
      "loss": -0.004,
      "reward": 0.706861674785614,
      "reward_std": 0.08733034506440163,
      "reward_std_all": 0.09066092222929001,
      "rewards-max/forensics_sinkhorn_reward": 0.9430786371231079,
      "rewards-max/forensics_strict_edge_reward": 0.24864380806684494,
      "rewards-max/forensics_tiou_reward": 0.8698292672634125,
      "rewards/forensics_sinkhorn_reward": 0.5983567833900452,
      "rewards/forensics_strict_edge_reward": 0.011640618788078427,
      "rewards/forensics_tiou_reward": 0.33407701551914215,
      "step": 175
    },
    {
      "completion_length": 23.078125,
      "epoch": 0.5407066052227343,
      "grad_norm": 8.43593692779541,
      "kl": 0.5048828125,
      "learning_rate": 8.653846153846154e-07,
      "loss": 0.1252,
      "reward": 0.7108941674232483,
      "reward_std": 0.07670816034078598,
      "reward_std_all": 0.08302289247512817,
      "rewards-max/forensics_sinkhorn_reward": 0.9637303054332733,
      "rewards-max/forensics_strict_edge_reward": 0.03272011844092049,
      "rewards-max/forensics_tiou_reward": 0.9082456529140472,
      "rewards/forensics_sinkhorn_reward": 0.6151198744773865,
      "rewards/forensics_strict_edge_reward": 0.0014330411568153067,
      "rewards/forensics_tiou_reward": 0.30754612386226654,
      "step": 176
    },
    {
      "completion_length": 22.28125,
      "epoch": 0.543778801843318,
      "grad_norm": 4.73450231552124,
      "kl": 0.26318359375,
      "learning_rate": 8.646153846153845e-07,
      "loss": 0.0837,
      "reward": 0.7177549302577972,
      "reward_std": 0.1045987717807293,
      "reward_std_all": 0.10052647814154625,
      "rewards-max/forensics_sinkhorn_reward": 0.9829886257648468,
      "rewards-max/forensics_strict_edge_reward": 0.1441603284329176,
      "rewards-max/forensics_tiou_reward": 0.9506175518035889,
      "rewards/forensics_sinkhorn_reward": 0.652069091796875,
      "rewards/forensics_strict_edge_reward": 0.006060234154574573,
      "rewards/forensics_tiou_reward": 0.39363856613636017,
      "step": 177
    },
    {
      "completion_length": 22.359375,
      "epoch": 0.5468509984639017,
      "grad_norm": 5.091055393218994,
      "kl": 0.3330078125,
      "learning_rate": 8.638461538461537e-07,
      "loss": 0.057,
      "reward": 0.7063981592655182,
      "reward_std": 0.10278325900435448,
      "reward_std_all": 0.10133341699838638,
      "rewards-max/forensics_sinkhorn_reward": 0.95417919754982,
      "rewards-max/forensics_strict_edge_reward": 0.2674546092748642,
      "rewards-max/forensics_tiou_reward": 0.8930916488170624,
      "rewards/forensics_sinkhorn_reward": 0.6345798373222351,
      "rewards/forensics_strict_edge_reward": 0.011686042649671435,
      "rewards/forensics_tiou_reward": 0.31082727015018463,
      "step": 178
    },
    {
      "completion_length": 25.09375,
      "epoch": 0.5499231950844854,
      "grad_norm": 5.571622848510742,
      "kl": 0.236328125,
      "learning_rate": 8.630769230769231e-07,
      "loss": 0.0713,
      "reward": 0.7499645948410034,
      "reward_std": 0.06392805464565754,
      "reward_std_all": 0.06173321418464184,
      "rewards-max/forensics_sinkhorn_reward": 0.9157631099224091,
      "rewards-max/forensics_strict_edge_reward": 0.04631636571139097,
      "rewards-max/forensics_tiou_reward": 0.7906602323055267,
      "rewards/forensics_sinkhorn_reward": 0.6623247861862183,
      "rewards/forensics_strict_edge_reward": 0.0029573237989097834,
      "rewards/forensics_tiou_reward": 0.3658461272716522,
      "step": 179
    },
    {
      "completion_length": 24.140625,
      "epoch": 0.5529953917050692,
      "grad_norm": 5.9720001220703125,
      "kl": 0.2421875,
      "learning_rate": 8.623076923076923e-07,
      "loss": 0.0912,
      "reward": 0.7199124395847321,
      "reward_std": 0.10897059738636017,
      "reward_std_all": 0.10113061964511871,
      "rewards-max/forensics_sinkhorn_reward": 0.9251597225666046,
      "rewards-max/forensics_strict_edge_reward": 0.2220008848235011,
      "rewards-max/forensics_tiou_reward": 0.82280033826828,
      "rewards/forensics_sinkhorn_reward": 0.6462219655513763,
      "rewards/forensics_strict_edge_reward": 0.015333844698034227,
      "rewards/forensics_tiou_reward": 0.36515237390995026,
      "step": 180
    },
    {
      "completion_length": 23.078125,
      "epoch": 0.5560675883256528,
      "grad_norm": 5.037367820739746,
      "kl": 0.240234375,
      "learning_rate": 8.615384615384616e-07,
      "loss": -0.0173,
      "reward": 0.7622413635253906,
      "reward_std": 0.043410683050751686,
      "reward_std_all": 0.043952884152531624,
      "rewards-max/forensics_sinkhorn_reward": 0.9822677075862885,
      "rewards-max/forensics_strict_edge_reward": 0.2202000506222248,
      "rewards-max/forensics_tiou_reward": 0.9550268650054932,
      "rewards/forensics_sinkhorn_reward": 0.7059381306171417,
      "rewards/forensics_strict_edge_reward": 0.015585934277623892,
      "rewards/forensics_tiou_reward": 0.43913714587688446,
      "step": 181
    },
    {
      "completion_length": 26.484375,
      "epoch": 0.5591397849462365,
      "grad_norm": 4.368110656738281,
      "kl": 0.3095703125,
      "learning_rate": 8.607692307692307e-07,
      "loss": 0.219,
      "reward": 0.7238655388355255,
      "reward_std": 0.06050465628504753,
      "reward_std_all": 0.06949567422270775,
      "rewards-max/forensics_sinkhorn_reward": 0.9770047068595886,
      "rewards-max/forensics_strict_edge_reward": 0.5653922781348228,
      "rewards-max/forensics_tiou_reward": 0.948167085647583,
      "rewards/forensics_sinkhorn_reward": 0.615501344203949,
      "rewards/forensics_strict_edge_reward": 0.030046301893889904,
      "rewards/forensics_tiou_reward": 0.4245452880859375,
      "step": 182
    },
    {
      "completion_length": 23.6875,
      "epoch": 0.5622119815668203,
      "grad_norm": 4.508033275604248,
      "kl": 0.322265625,
      "learning_rate": 8.599999999999999e-07,
      "loss": 0.1002,
      "reward": 0.7044855952262878,
      "reward_std": 0.050986869260668755,
      "reward_std_all": 0.07321051508188248,
      "rewards-max/forensics_sinkhorn_reward": 0.9705662429332733,
      "rewards-max/forensics_strict_edge_reward": 0.18767917901277542,
      "rewards-max/forensics_tiou_reward": 0.9300405383110046,
      "rewards/forensics_sinkhorn_reward": 0.6212525367736816,
      "rewards/forensics_strict_edge_reward": 0.009312197100371122,
      "rewards/forensics_tiou_reward": 0.32615791261196136,
      "step": 183
    },
    {
      "completion_length": 24.78125,
      "epoch": 0.565284178187404,
      "grad_norm": 4.6658549308776855,
      "kl": 0.32958984375,
      "learning_rate": 8.592307692307692e-07,
      "loss": 0.0312,
      "reward": 0.7303117215633392,
      "reward_std": 0.04111719690263271,
      "reward_std_all": 0.05826252233237028,
      "rewards-max/forensics_sinkhorn_reward": 0.9526202976703644,
      "rewards-max/forensics_strict_edge_reward": 0.1757325679063797,
      "rewards-max/forensics_tiou_reward": 0.8829300701618195,
      "rewards/forensics_sinkhorn_reward": 0.6552910506725311,
      "rewards/forensics_strict_edge_reward": 0.012624438386410475,
      "rewards/forensics_tiou_reward": 0.3272808790206909,
      "step": 184
    },
    {
      "completion_length": 22.171875,
      "epoch": 0.5683563748079877,
      "grad_norm": 5.65582275390625,
      "kl": 0.42578125,
      "learning_rate": 8.584615384615385e-07,
      "loss": 0.2691,
      "reward": 0.7227618992328644,
      "reward_std": 0.049334101378917694,
      "reward_std_all": 0.06633567810058594,
      "rewards-max/forensics_sinkhorn_reward": 0.952976793050766,
      "rewards-max/forensics_strict_edge_reward": 0.21110916882753372,
      "rewards-max/forensics_tiou_reward": 0.8885067701339722,
      "rewards/forensics_sinkhorn_reward": 0.6160563826560974,
      "rewards/forensics_strict_edge_reward": 0.010619430802762508,
      "rewards/forensics_tiou_reward": 0.3087422698736191,
      "step": 185
    },
    {
      "completion_length": 24.53125,
      "epoch": 0.5714285714285714,
      "grad_norm": 6.130036354064941,
      "kl": 0.31787109375,
      "learning_rate": 8.576923076923076e-07,
      "loss": 0.126,
      "reward": 0.7524901330471039,
      "reward_std": 0.05767356604337692,
      "reward_std_all": 0.05853908881545067,
      "rewards-max/forensics_sinkhorn_reward": 0.9280444979667664,
      "rewards-max/forensics_strict_edge_reward": 0.13239384430926293,
      "rewards-max/forensics_tiou_reward": 0.8122301697731018,
      "rewards/forensics_sinkhorn_reward": 0.6695063710212708,
      "rewards/forensics_strict_edge_reward": 0.01132608910847921,
      "rewards/forensics_tiou_reward": 0.3714744448661804,
      "step": 186
    },
    {
      "completion_length": 27.84375,
      "epoch": 0.5745007680491552,
      "grad_norm": 5.709922790527344,
      "kl": 0.28173828125,
      "learning_rate": 8.569230769230769e-07,
      "loss": 0.1728,
      "reward": 0.6879629492759705,
      "reward_std": 0.12412824481725693,
      "reward_std_all": 0.11920666322112083,
      "rewards-max/forensics_sinkhorn_reward": 0.9150972962379456,
      "rewards-max/forensics_strict_edge_reward": 0.14146363362669945,
      "rewards-max/forensics_tiou_reward": 0.7923214733600616,
      "rewards/forensics_sinkhorn_reward": 0.5507871210575104,
      "rewards/forensics_strict_edge_reward": 0.012752039590850472,
      "rewards/forensics_tiou_reward": 0.31660979986190796,
      "step": 187
    },
    {
      "completion_length": 23.84375,
      "epoch": 0.5775729646697388,
      "grad_norm": 6.198731422424316,
      "kl": 0.26220703125,
      "learning_rate": 8.561538461538461e-07,
      "loss": 0.107,
      "reward": 0.7006376385688782,
      "reward_std": 0.08722444623708725,
      "reward_std_all": 0.09232393652200699,
      "rewards-max/forensics_sinkhorn_reward": 0.9085017740726471,
      "rewards-max/forensics_strict_edge_reward": 0.08044840767979622,
      "rewards-max/forensics_tiou_reward": 0.7904905378818512,
      "rewards/forensics_sinkhorn_reward": 0.6028601825237274,
      "rewards/forensics_strict_edge_reward": 0.005257208424154669,
      "rewards/forensics_tiou_reward": 0.3267920911312103,
      "step": 188
    },
    {
      "completion_length": 22.703125,
      "epoch": 0.5806451612903226,
      "grad_norm": 4.81500768661499,
      "kl": 0.27880859375,
      "learning_rate": 8.553846153846154e-07,
      "loss": 0.0619,
      "reward": 0.7121055126190186,
      "reward_std": 0.08752785623073578,
      "reward_std_all": 0.09529275447130203,
      "rewards-max/forensics_sinkhorn_reward": 0.9571750462055206,
      "rewards-max/forensics_strict_edge_reward": 0.09453566299634986,
      "rewards-max/forensics_tiou_reward": 0.8759787380695343,
      "rewards/forensics_sinkhorn_reward": 0.6510333716869354,
      "rewards/forensics_strict_edge_reward": 0.011176184097621444,
      "rewards/forensics_tiou_reward": 0.3926633447408676,
      "step": 189
    },
    {
      "completion_length": 27.828125,
      "epoch": 0.5837173579109063,
      "grad_norm": 3.9658114910125732,
      "kl": 0.2841796875,
      "learning_rate": 8.546153846153846e-07,
      "loss": 0.1713,
      "reward": 0.6946030259132385,
      "reward_std": 0.09576333314180374,
      "reward_std_all": 0.10399878025054932,
      "rewards-max/forensics_sinkhorn_reward": 0.9357289969921112,
      "rewards-max/forensics_strict_edge_reward": 0.19938001688569784,
      "rewards-max/forensics_tiou_reward": 0.8525427877902985,
      "rewards/forensics_sinkhorn_reward": 0.6101097464561462,
      "rewards/forensics_strict_edge_reward": 0.013285932480357587,
      "rewards/forensics_tiou_reward": 0.3560955971479416,
      "step": 190
    },
    {
      "completion_length": 25.4375,
      "epoch": 0.5867895545314901,
      "grad_norm": 4.652796268463135,
      "kl": 0.326171875,
      "learning_rate": 8.538461538461537e-07,
      "loss": 0.1258,
      "reward": 0.7199756801128387,
      "reward_std": 0.09097590856254101,
      "reward_std_all": 0.08852023631334305,
      "rewards-max/forensics_sinkhorn_reward": 0.9215164184570312,
      "rewards-max/forensics_strict_edge_reward": 0.29868117719888687,
      "rewards-max/forensics_tiou_reward": 0.8172651529312134,
      "rewards/forensics_sinkhorn_reward": 0.5783332586288452,
      "rewards/forensics_strict_edge_reward": 0.013112624874338508,
      "rewards/forensics_tiou_reward": 0.33134888112545013,
      "step": 191
    },
    {
      "completion_length": 25.765625,
      "epoch": 0.5898617511520737,
      "grad_norm": 5.960731029510498,
      "kl": 0.3046875,
      "learning_rate": 8.530769230769231e-07,
      "loss": 0.0774,
      "reward": 0.7306247055530548,
      "reward_std": 0.056585599668323994,
      "reward_std_all": 0.06788578908890486,
      "rewards-max/forensics_sinkhorn_reward": 0.980354517698288,
      "rewards-max/forensics_strict_edge_reward": 0.5919938534498215,
      "rewards-max/forensics_tiou_reward": 0.9500716328620911,
      "rewards/forensics_sinkhorn_reward": 0.6539855897426605,
      "rewards/forensics_strict_edge_reward": 0.04008418321609497,
      "rewards/forensics_tiou_reward": 0.478763222694397,
      "step": 192
    },
    {
      "completion_length": 26.859375,
      "epoch": 0.5929339477726574,
      "grad_norm": 4.617890357971191,
      "kl": 0.322265625,
      "learning_rate": 8.523076923076923e-07,
      "loss": 0.2283,
      "reward": 0.7144529819488525,
      "reward_std": 0.09051537327468395,
      "reward_std_all": 0.08911399357020855,
      "rewards-max/forensics_sinkhorn_reward": 0.9852429330348969,
      "rewards-max/forensics_strict_edge_reward": 0.4506318047642708,
      "rewards-max/forensics_tiou_reward": 0.9672177135944366,
      "rewards/forensics_sinkhorn_reward": 0.6215175092220306,
      "rewards/forensics_strict_edge_reward": 0.039354690350592136,
      "rewards/forensics_tiou_reward": 0.3861679285764694,
      "step": 193
    },
    {
      "completion_length": 28.265625,
      "epoch": 0.5960061443932412,
      "grad_norm": 4.224941730499268,
      "kl": 0.23828125,
      "learning_rate": 8.515384615384616e-07,
      "loss": 0.1153,
      "reward": 0.7039482295513153,
      "reward_std": 0.06967560201883316,
      "reward_std_all": 0.0865241140127182,
      "rewards-max/forensics_sinkhorn_reward": 0.9493037462234497,
      "rewards-max/forensics_strict_edge_reward": 0.15220985002815723,
      "rewards-max/forensics_tiou_reward": 0.883304625749588,
      "rewards/forensics_sinkhorn_reward": 0.5905168950557709,
      "rewards/forensics_strict_edge_reward": 0.007781284104567021,
      "rewards/forensics_tiou_reward": 0.36886104941368103,
      "step": 194
    },
    {
      "completion_length": 25.09375,
      "epoch": 0.5990783410138248,
      "grad_norm": 4.235208511352539,
      "kl": 0.302734375,
      "learning_rate": 8.507692307692307e-07,
      "loss": 0.1025,
      "reward": 0.7054599225521088,
      "reward_std": 0.0490981750190258,
      "reward_std_all": 0.07175290957093239,
      "rewards-max/forensics_sinkhorn_reward": 0.9901129901409149,
      "rewards-max/forensics_strict_edge_reward": 0.7474130392074585,
      "rewards-max/forensics_tiou_reward": 0.9807636141777039,
      "rewards/forensics_sinkhorn_reward": 0.6122033596038818,
      "rewards/forensics_strict_edge_reward": 0.034690458327531815,
      "rewards/forensics_tiou_reward": 0.38794639706611633,
      "step": 195
    },
    {
      "completion_length": 25.140625,
      "epoch": 0.6021505376344086,
      "grad_norm": 5.147428512573242,
      "kl": 0.28564453125,
      "learning_rate": 8.499999999999999e-07,
      "loss": -0.011,
      "reward": 0.6922307014465332,
      "reward_std": 0.08765874430537224,
      "reward_std_all": 0.09766906127333641,
      "rewards-max/forensics_sinkhorn_reward": 0.9278749227523804,
      "rewards-max/forensics_strict_edge_reward": 0.10329780448228121,
      "rewards-max/forensics_tiou_reward": 0.8414944708347321,
      "rewards/forensics_sinkhorn_reward": 0.5685358345508575,
      "rewards/forensics_strict_edge_reward": 0.005188131355680525,
      "rewards/forensics_tiou_reward": 0.2775881588459015,
      "step": 196
    },
    {
      "completion_length": 30.78125,
      "epoch": 0.6052227342549923,
      "grad_norm": 4.771385192871094,
      "kl": 0.24853515625,
      "learning_rate": 8.492307692307692e-07,
      "loss": 0.0426,
      "reward": 0.6510176062583923,
      "reward_std": 0.08268619701266289,
      "reward_std_all": 0.1149444580078125,
      "rewards-max/forensics_sinkhorn_reward": 0.9584707021713257,
      "rewards-max/forensics_strict_edge_reward": 0.15008831396698952,
      "rewards-max/forensics_tiou_reward": 0.9033401310443878,
      "rewards/forensics_sinkhorn_reward": 0.5523868203163147,
      "rewards/forensics_strict_edge_reward": 0.010788967367261648,
      "rewards/forensics_tiou_reward": 0.32183049619197845,
      "step": 197
    },
    {
      "completion_length": 27.828125,
      "epoch": 0.6082949308755761,
      "grad_norm": 5.632797718048096,
      "kl": 0.28564453125,
      "learning_rate": 8.484615384615384e-07,
      "loss": 0.0941,
      "reward": 0.7223730385303497,
      "reward_std": 0.064632099121809,
      "reward_std_all": 0.07316534221172333,
      "rewards-max/forensics_sinkhorn_reward": 0.9346083998680115,
      "rewards-max/forensics_strict_edge_reward": 0.023255647160112858,
      "rewards-max/forensics_tiou_reward": 0.8228476941585541,
      "rewards/forensics_sinkhorn_reward": 0.6306180059909821,
      "rewards/forensics_strict_edge_reward": 0.0014322507777251303,
      "rewards/forensics_tiou_reward": 0.35411714017391205,
      "step": 198
    },
    {
      "completion_length": 24.0,
      "epoch": 0.6113671274961597,
      "grad_norm": 5.271913528442383,
      "kl": 0.3115234375,
      "learning_rate": 8.476923076923077e-07,
      "loss": 0.1188,
      "reward": 0.7144116759300232,
      "reward_std": 0.06819891184568405,
      "reward_std_all": 0.08385755866765976,
      "rewards-max/forensics_sinkhorn_reward": 0.9385958313941956,
      "rewards-max/forensics_strict_edge_reward": 0.12677174620330334,
      "rewards-max/forensics_tiou_reward": 0.8504230082035065,
      "rewards/forensics_sinkhorn_reward": 0.648966372013092,
      "rewards/forensics_strict_edge_reward": 0.007266335655003786,
      "rewards/forensics_tiou_reward": 0.367504745721817,
      "step": 199
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.6144393241167435,
      "grad_norm": 5.647727012634277,
      "kl": 0.2958984375,
      "learning_rate": 8.469230769230769e-07,
      "loss": 0.0585,
      "reward": 0.7057760953903198,
      "reward_std": 0.06656449288129807,
      "reward_std_all": 0.08354915678501129,
      "rewards-max/forensics_sinkhorn_reward": 0.9472911655902863,
      "rewards-max/forensics_strict_edge_reward": 0.2789415828883648,
      "rewards-max/forensics_tiou_reward": 0.8693147003650665,
      "rewards/forensics_sinkhorn_reward": 0.6107226610183716,
      "rewards/forensics_strict_edge_reward": 0.017536553787067533,
      "rewards/forensics_tiou_reward": 0.38530193269252777,
      "step": 200
    },
    {
      "completion_length": 23.09375,
      "epoch": 0.6175115207373272,
      "grad_norm": 6.414347171783447,
      "kl": 0.23583984375,
      "learning_rate": 8.461538461538461e-07,
      "loss": 0.1211,
      "reward": 0.686218798160553,
      "reward_std": 0.08439266681671143,
      "reward_std_all": 0.10065164044499397,
      "rewards-max/forensics_sinkhorn_reward": 0.9778960943222046,
      "rewards-max/forensics_strict_edge_reward": 0.31765206158161163,
      "rewards-max/forensics_tiou_reward": 0.9531524181365967,
      "rewards/forensics_sinkhorn_reward": 0.616859495639801,
      "rewards/forensics_strict_edge_reward": 0.014253450557589531,
      "rewards/forensics_tiou_reward": 0.37730932235717773,
      "step": 201
    },
    {
      "completion_length": 26.71875,
      "epoch": 0.620583717357911,
      "grad_norm": 4.560348033905029,
      "kl": 0.23876953125,
      "learning_rate": 8.453846153846154e-07,
      "loss": -0.052,
      "reward": 0.695436030626297,
      "reward_std": 0.11244922131299973,
      "reward_std_all": 0.11409617215394974,
      "rewards-max/forensics_sinkhorn_reward": 0.9723645150661469,
      "rewards-max/forensics_strict_edge_reward": 0.014233759604394436,
      "rewards-max/forensics_tiou_reward": 0.9334133267402649,
      "rewards/forensics_sinkhorn_reward": 0.5663662850856781,
      "rewards/forensics_strict_edge_reward": 0.0006397166143869981,
      "rewards/forensics_tiou_reward": 0.32052721083164215,
      "step": 202
    },
    {
      "completion_length": 24.625,
      "epoch": 0.6236559139784946,
      "grad_norm": 4.061238765716553,
      "kl": 0.2939453125,
      "learning_rate": 8.446153846153846e-07,
      "loss": 0.1501,
      "reward": 0.7423223555088043,
      "reward_std": 0.06096148118376732,
      "reward_std_all": 0.06166371516883373,
      "rewards-max/forensics_sinkhorn_reward": 0.9456736147403717,
      "rewards-max/forensics_strict_edge_reward": 0.35156793892383575,
      "rewards-max/forensics_tiou_reward": 0.8645065426826477,
      "rewards/forensics_sinkhorn_reward": 0.6463193595409393,
      "rewards/forensics_strict_edge_reward": 0.02465287782251835,
      "rewards/forensics_tiou_reward": 0.32935675978660583,
      "step": 203
    },
    {
      "completion_length": 23.375,
      "epoch": 0.6267281105990783,
      "grad_norm": 4.707395076751709,
      "kl": 0.259765625,
      "learning_rate": 8.438461538461537e-07,
      "loss": 0.0046,
      "reward": 0.7348921895027161,
      "reward_std": 0.057114774361252785,
      "reward_std_all": 0.06115440092980862,
      "rewards-max/forensics_sinkhorn_reward": 0.9684253334999084,
      "rewards-max/forensics_strict_edge_reward": 0.4305874779820442,
      "rewards-max/forensics_tiou_reward": 0.9183161854743958,
      "rewards/forensics_sinkhorn_reward": 0.6144805550575256,
      "rewards/forensics_strict_edge_reward": 0.02216621721163392,
      "rewards/forensics_tiou_reward": 0.37192384898662567,
      "step": 204
    },
    {
      "completion_length": 25.671875,
      "epoch": 0.6298003072196621,
      "grad_norm": 4.605508327484131,
      "kl": 0.23974609375,
      "learning_rate": 8.43076923076923e-07,
      "loss": 0.1177,
      "reward": 0.7054294049739838,
      "reward_std": 0.08001209050416946,
      "reward_std_all": 0.09199441596865654,
      "rewards-max/forensics_sinkhorn_reward": 0.9268761575222015,
      "rewards-max/forensics_strict_edge_reward": 0.15874878549948335,
      "rewards-max/forensics_tiou_reward": 0.8191050291061401,
      "rewards/forensics_sinkhorn_reward": 0.6103891134262085,
      "rewards/forensics_strict_edge_reward": 0.010549723629083019,
      "rewards/forensics_tiou_reward": 0.3480709344148636,
      "step": 205
    },
    {
      "completion_length": 23.59375,
      "epoch": 0.6328725038402457,
      "grad_norm": 5.3309197425842285,
      "kl": 0.26953125,
      "learning_rate": 8.423076923076923e-07,
      "loss": 0.2356,
      "reward": 0.7041538059711456,
      "reward_std": 0.09999942034482956,
      "reward_std_all": 0.10063492506742477,
      "rewards-max/forensics_sinkhorn_reward": 0.9757456481456757,
      "rewards-max/forensics_strict_edge_reward": 0.28496674448251724,
      "rewards-max/forensics_tiou_reward": 0.9359012544155121,
      "rewards/forensics_sinkhorn_reward": 0.6099700033664703,
      "rewards/forensics_strict_edge_reward": 0.014591429382562637,
      "rewards/forensics_tiou_reward": 0.3640151917934418,
      "step": 206
    },
    {
      "completion_length": 24.8125,
      "epoch": 0.6359447004608295,
      "grad_norm": 4.394705295562744,
      "kl": 0.3173828125,
      "learning_rate": 8.415384615384616e-07,
      "loss": 0.084,
      "reward": 0.76295405626297,
      "reward_std": 0.0610829871147871,
      "reward_std_all": 0.05632794462144375,
      "rewards-max/forensics_sinkhorn_reward": 0.9572460055351257,
      "rewards-max/forensics_strict_edge_reward": 0.28684211522340775,
      "rewards-max/forensics_tiou_reward": 0.8972707688808441,
      "rewards/forensics_sinkhorn_reward": 0.7100242972373962,
      "rewards/forensics_strict_edge_reward": 0.016980656888335943,
      "rewards/forensics_tiou_reward": 0.42106279730796814,
      "step": 207
    },
    {
      "completion_length": 23.046875,
      "epoch": 0.6390168970814132,
      "grad_norm": 3.83941650390625,
      "kl": 0.25341796875,
      "learning_rate": 8.407692307692308e-07,
      "loss": 0.3012,
      "reward": 0.6938992142677307,
      "reward_std": 0.09009211510419846,
      "reward_std_all": 0.09949922189116478,
      "rewards-max/forensics_sinkhorn_reward": 0.9262366592884064,
      "rewards-max/forensics_strict_edge_reward": 0.07103661424480379,
      "rewards-max/forensics_tiou_reward": 0.7971981763839722,
      "rewards/forensics_sinkhorn_reward": 0.6129581928253174,
      "rewards/forensics_strict_edge_reward": 0.004458696086658165,
      "rewards/forensics_tiou_reward": 0.3013678938150406,
      "step": 208
    },
    {
      "completion_length": 24.46875,
      "epoch": 0.642089093701997,
      "grad_norm": 4.462911605834961,
      "kl": 0.3427734375,
      "learning_rate": 8.399999999999999e-07,
      "loss": 0.1853,
      "reward": 0.6615142822265625,
      "reward_std": 0.04949467070400715,
      "reward_std_all": 0.09616347402334213,
      "rewards-max/forensics_sinkhorn_reward": 0.9113674163818359,
      "rewards-max/forensics_strict_edge_reward": 0.24066102923825383,
      "rewards-max/forensics_tiou_reward": 0.7846504151821136,
      "rewards/forensics_sinkhorn_reward": 0.552445650100708,
      "rewards/forensics_strict_edge_reward": 0.012238151510246098,
      "rewards/forensics_tiou_reward": 0.2513049691915512,
      "step": 209
    },
    {
      "completion_length": 24.625,
      "epoch": 0.6451612903225806,
      "grad_norm": 5.67304801940918,
      "kl": 0.23046875,
      "learning_rate": 8.392307692307692e-07,
      "loss": 0.0392,
      "reward": 0.7350676655769348,
      "reward_std": 0.06914057023823261,
      "reward_std_all": 0.0727262943983078,
      "rewards-max/forensics_sinkhorn_reward": 0.8890627920627594,
      "rewards-max/forensics_strict_edge_reward": 0.07540529780089855,
      "rewards-max/forensics_tiou_reward": 0.7428399920463562,
      "rewards/forensics_sinkhorn_reward": 0.6249628365039825,
      "rewards/forensics_strict_edge_reward": 0.004166140221059322,
      "rewards/forensics_tiou_reward": 0.3434494137763977,
      "step": 210
    },
    {
      "completion_length": 22.5,
      "epoch": 0.6482334869431644,
      "grad_norm": 5.18406343460083,
      "kl": 0.3076171875,
      "learning_rate": 8.384615384615384e-07,
      "loss": 0.1779,
      "reward": 0.697892040014267,
      "reward_std": 0.08789899945259094,
      "reward_std_all": 0.09636945649981499,
      "rewards-max/forensics_sinkhorn_reward": 0.9736196100711823,
      "rewards-max/forensics_strict_edge_reward": 0.3001514635980129,
      "rewards-max/forensics_tiou_reward": 0.937962144613266,
      "rewards/forensics_sinkhorn_reward": 0.6302169859409332,
      "rewards/forensics_strict_edge_reward": 0.014449699549004436,
      "rewards/forensics_tiou_reward": 0.333346962928772,
      "step": 211
    },
    {
      "completion_length": 22.171875,
      "epoch": 0.6513056835637481,
      "grad_norm": 4.9826836585998535,
      "kl": 0.2998046875,
      "learning_rate": 8.376923076923076e-07,
      "loss": 0.1623,
      "reward": 0.7153072953224182,
      "reward_std": 0.06860922649502754,
      "reward_std_all": 0.07953687757253647,
      "rewards-max/forensics_sinkhorn_reward": 0.97196826338768,
      "rewards-max/forensics_strict_edge_reward": 0.4855904970318079,
      "rewards-max/forensics_tiou_reward": 0.9260562658309937,
      "rewards/forensics_sinkhorn_reward": 0.6362643539905548,
      "rewards/forensics_strict_edge_reward": 0.0226800482487306,
      "rewards/forensics_tiou_reward": 0.36447444558143616,
      "step": 212
    },
    {
      "completion_length": 22.921875,
      "epoch": 0.6543778801843319,
      "grad_norm": 5.722536087036133,
      "kl": 0.275390625,
      "learning_rate": 8.369230769230769e-07,
      "loss": 0.0905,
      "reward": 0.7493269443511963,
      "reward_std": 0.07371676340699196,
      "reward_std_all": 0.0690363347530365,
      "rewards-max/forensics_sinkhorn_reward": 0.9869012534618378,
      "rewards-max/forensics_strict_edge_reward": 0.4827862083911896,
      "rewards-max/forensics_tiou_reward": 0.9716439247131348,
      "rewards/forensics_sinkhorn_reward": 0.6426869034767151,
      "rewards/forensics_strict_edge_reward": 0.027482176199555397,
      "rewards/forensics_tiou_reward": 0.3723777234554291,
      "step": 213
    },
    {
      "completion_length": 23.203125,
      "epoch": 0.6574500768049155,
      "grad_norm": 5.58182954788208,
      "kl": 0.3662109375,
      "learning_rate": 8.361538461538461e-07,
      "loss": 0.0668,
      "reward": 0.7104011178016663,
      "reward_std": 0.07399225234985352,
      "reward_std_all": 0.0817926749587059,
      "rewards-max/forensics_sinkhorn_reward": 0.9303759932518005,
      "rewards-max/forensics_strict_edge_reward": 0.23612900217995048,
      "rewards-max/forensics_tiou_reward": 0.8318465650081635,
      "rewards/forensics_sinkhorn_reward": 0.6330931782722473,
      "rewards/forensics_strict_edge_reward": 0.013329755456652492,
      "rewards/forensics_tiou_reward": 0.34581299126148224,
      "step": 214
    },
    {
      "completion_length": 23.375,
      "epoch": 0.6605222734254992,
      "grad_norm": 4.0944318771362305,
      "kl": 0.23876953125,
      "learning_rate": 8.353846153846154e-07,
      "loss": 0.0286,
      "reward": 0.7583065629005432,
      "reward_std": 0.056749939918518066,
      "reward_std_all": 0.05449729412794113,
      "rewards-max/forensics_sinkhorn_reward": 0.9696190655231476,
      "rewards-max/forensics_strict_edge_reward": 0.1380850486457348,
      "rewards-max/forensics_tiou_reward": 0.9281174838542938,
      "rewards/forensics_sinkhorn_reward": 0.7059367299079895,
      "rewards/forensics_strict_edge_reward": 0.011194286867976189,
      "rewards/forensics_tiou_reward": 0.44149354100227356,
      "step": 215
    },
    {
      "completion_length": 23.390625,
      "epoch": 0.663594470046083,
      "grad_norm": 4.135667324066162,
      "kl": 0.28515625,
      "learning_rate": 8.346153846153846e-07,
      "loss": 0.1366,
      "reward": 0.7373992204666138,
      "reward_std": 0.07754047587513924,
      "reward_std_all": 0.07539278268814087,
      "rewards-max/forensics_sinkhorn_reward": 0.9585040509700775,
      "rewards-max/forensics_strict_edge_reward": 0.29567399015650153,
      "rewards-max/forensics_tiou_reward": 0.8921259343624115,
      "rewards/forensics_sinkhorn_reward": 0.6707073748111725,
      "rewards/forensics_strict_edge_reward": 0.014993820383097045,
      "rewards/forensics_tiou_reward": 0.3872769922018051,
      "step": 216
    },
    {
      "completion_length": 23.75,
      "epoch": 0.6666666666666666,
      "grad_norm": 4.071014881134033,
      "kl": 0.267578125,
      "learning_rate": 8.338461538461537e-07,
      "loss": -0.064,
      "reward": 0.776132345199585,
      "reward_std": 0.03774607367813587,
      "reward_std_all": 0.03680235706269741,
      "rewards-max/forensics_sinkhorn_reward": 0.968011736869812,
      "rewards-max/forensics_strict_edge_reward": 0.3037954159080982,
      "rewards-max/forensics_tiou_reward": 0.918921709060669,
      "rewards/forensics_sinkhorn_reward": 0.7128761410713196,
      "rewards/forensics_strict_edge_reward": 0.027537682093679905,
      "rewards/forensics_tiou_reward": 0.45124876499176025,
      "step": 217
    },
    {
      "completion_length": 25.921875,
      "epoch": 0.6697388632872504,
      "grad_norm": 5.128841400146484,
      "kl": 0.23583984375,
      "learning_rate": 8.33076923076923e-07,
      "loss": -0.0276,
      "reward": 0.7192718088626862,
      "reward_std": 0.06400086358189583,
      "reward_std_all": 0.07673540338873863,
      "rewards-max/forensics_sinkhorn_reward": 0.9740767180919647,
      "rewards-max/forensics_strict_edge_reward": 0.40878118574619293,
      "rewards-max/forensics_tiou_reward": 0.9377051293849945,
      "rewards/forensics_sinkhorn_reward": 0.6476978063583374,
      "rewards/forensics_strict_edge_reward": 0.03507748991250992,
      "rewards/forensics_tiou_reward": 0.4407687485218048,
      "step": 218
    },
    {
      "completion_length": 21.765625,
      "epoch": 0.6728110599078341,
      "grad_norm": 5.21058988571167,
      "kl": 0.21728515625,
      "learning_rate": 8.323076923076922e-07,
      "loss": 0.0724,
      "reward": 0.7479207813739777,
      "reward_std": 0.05656038783490658,
      "reward_std_all": 0.056976692751049995,
      "rewards-max/forensics_sinkhorn_reward": 0.980575293302536,
      "rewards-max/forensics_strict_edge_reward": 0.3467242270708084,
      "rewards-max/forensics_tiou_reward": 0.9504920840263367,
      "rewards/forensics_sinkhorn_reward": 0.6732054650783539,
      "rewards/forensics_strict_edge_reward": 0.023113390430808067,
      "rewards/forensics_tiou_reward": 0.3815457671880722,
      "step": 219
    },
    {
      "completion_length": 25.03125,
      "epoch": 0.6758832565284179,
      "grad_norm": 5.261828422546387,
      "kl": 0.26953125,
      "learning_rate": 8.315384615384616e-07,
      "loss": 0.3038,
      "reward": 0.6740906834602356,
      "reward_std": 0.094681017100811,
      "reward_std_all": 0.10713550448417664,
      "rewards-max/forensics_sinkhorn_reward": 0.9123677015304565,
      "rewards-max/forensics_strict_edge_reward": 0.21626371517777443,
      "rewards-max/forensics_tiou_reward": 0.7857179641723633,
      "rewards/forensics_sinkhorn_reward": 0.5668279826641083,
      "rewards/forensics_strict_edge_reward": 0.012196376919746399,
      "rewards/forensics_tiou_reward": 0.2796141430735588,
      "step": 220
    },
    {
      "completion_length": 22.203125,
      "epoch": 0.6789554531490015,
      "grad_norm": 4.877140045166016,
      "kl": 0.3701171875,
      "learning_rate": 8.307692307692308e-07,
      "loss": 0.2034,
      "reward": 0.6781183183193207,
      "reward_std": 0.08745963871479034,
      "reward_std_all": 0.10274216532707214,
      "rewards-max/forensics_sinkhorn_reward": 0.9595261216163635,
      "rewards-max/forensics_strict_edge_reward": 0.2517708130180836,
      "rewards-max/forensics_tiou_reward": 0.9127359986305237,
      "rewards/forensics_sinkhorn_reward": 0.5793988406658173,
      "rewards/forensics_strict_edge_reward": 0.013340732082724571,
      "rewards/forensics_tiou_reward": 0.33248306810855865,
      "step": 221
    },
    {
      "completion_length": 23.203125,
      "epoch": 0.6820276497695853,
      "grad_norm": 4.798935890197754,
      "kl": 0.2490234375,
      "learning_rate": 8.299999999999999e-07,
      "loss": 0.0832,
      "reward": 0.6942126750946045,
      "reward_std": 0.08113963901996613,
      "reward_std_all": 0.09420065954327583,
      "rewards-max/forensics_sinkhorn_reward": 0.9617737531661987,
      "rewards-max/forensics_strict_edge_reward": 0.043975391425192356,
      "rewards-max/forensics_tiou_reward": 0.896599531173706,
      "rewards/forensics_sinkhorn_reward": 0.5878118872642517,
      "rewards/forensics_strict_edge_reward": 0.002786738274153322,
      "rewards/forensics_tiou_reward": 0.30762647092342377,
      "step": 222
    },
    {
      "completion_length": 23.0625,
      "epoch": 0.685099846390169,
      "grad_norm": 5.141852378845215,
      "kl": 0.3408203125,
      "learning_rate": 8.292307692307692e-07,
      "loss": 0.1128,
      "reward": 0.7374865412712097,
      "reward_std": 0.09146550297737122,
      "reward_std_all": 0.08427507802844048,
      "rewards-max/forensics_sinkhorn_reward": 0.9664525389671326,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9246311783790588,
      "rewards/forensics_sinkhorn_reward": 0.6732359826564789,
      "rewards/forensics_strict_edge_reward": 0.028538056649267673,
      "rewards/forensics_tiou_reward": 0.4256783723831177,
      "step": 223
    },
    {
      "completion_length": 23.265625,
      "epoch": 0.6881720430107527,
      "grad_norm": 5.62467098236084,
      "kl": 0.27880859375,
      "learning_rate": 8.284615384615384e-07,
      "loss": 0.1458,
      "reward": 0.660152941942215,
      "reward_std": 0.06557061709463596,
      "reward_std_all": 0.10152288898825645,
      "rewards-max/forensics_sinkhorn_reward": 0.9493666589260101,
      "rewards-max/forensics_strict_edge_reward": 0.18761547282338142,
      "rewards-max/forensics_tiou_reward": 0.8732470571994781,
      "rewards/forensics_sinkhorn_reward": 0.5597235858440399,
      "rewards/forensics_strict_edge_reward": 0.014696097001433372,
      "rewards/forensics_tiou_reward": 0.25802184641361237,
      "step": 224
    },
    {
      "completion_length": 25.046875,
      "epoch": 0.6912442396313364,
      "grad_norm": 7.575687885284424,
      "kl": 0.33544921875,
      "learning_rate": 8.276923076923077e-07,
      "loss": 0.2607,
      "reward": 0.6758771240711212,
      "reward_std": 0.11473435908555984,
      "reward_std_all": 0.11865881830453873,
      "rewards-max/forensics_sinkhorn_reward": 0.9636828601360321,
      "rewards-max/forensics_strict_edge_reward": 0.19372887630015612,
      "rewards-max/forensics_tiou_reward": 0.9019130766391754,
      "rewards/forensics_sinkhorn_reward": 0.5861241817474365,
      "rewards/forensics_strict_edge_reward": 0.01435717154527083,
      "rewards/forensics_tiou_reward": 0.3426591381430626,
      "step": 225
    },
    {
      "completion_length": 24.84375,
      "epoch": 0.6943164362519201,
      "grad_norm": 4.453864574432373,
      "kl": 0.2822265625,
      "learning_rate": 8.269230769230768e-07,
      "loss": 0.1662,
      "reward": 0.6620627045631409,
      "reward_std": 0.11683877184987068,
      "reward_std_all": 0.13248973339796066,
      "rewards-max/forensics_sinkhorn_reward": 0.9283438920974731,
      "rewards-max/forensics_strict_edge_reward": 0.2881809980608523,
      "rewards-max/forensics_tiou_reward": 0.8607122898101807,
      "rewards/forensics_sinkhorn_reward": 0.545634001493454,
      "rewards/forensics_strict_edge_reward": 0.015618178469594568,
      "rewards/forensics_tiou_reward": 0.296885684132576,
      "step": 226
    },
    {
      "completion_length": 22.1875,
      "epoch": 0.6973886328725039,
      "grad_norm": 4.822687149047852,
      "kl": 0.279296875,
      "learning_rate": 8.261538461538461e-07,
      "loss": 0.1221,
      "reward": 0.7451145350933075,
      "reward_std": 0.0756710534915328,
      "reward_std_all": 0.07152100652456284,
      "rewards-max/forensics_sinkhorn_reward": 0.9637899994850159,
      "rewards-max/forensics_strict_edge_reward": 0.2731504626572132,
      "rewards-max/forensics_tiou_reward": 0.914526104927063,
      "rewards/forensics_sinkhorn_reward": 0.6653495728969574,
      "rewards/forensics_strict_edge_reward": 0.013744731433689594,
      "rewards/forensics_tiou_reward": 0.3986578732728958,
      "step": 227
    },
    {
      "completion_length": 22.59375,
      "epoch": 0.7004608294930875,
      "grad_norm": 5.621265888214111,
      "kl": 0.20654296875,
      "learning_rate": 8.253846153846154e-07,
      "loss": 0.074,
      "reward": 0.705272376537323,
      "reward_std": 0.05321476049721241,
      "reward_std_all": 0.0749137420207262,
      "rewards-max/forensics_sinkhorn_reward": 0.9685590863227844,
      "rewards-max/forensics_strict_edge_reward": 0.38126203045248985,
      "rewards-max/forensics_tiou_reward": 0.928464263677597,
      "rewards/forensics_sinkhorn_reward": 0.6181788146495819,
      "rewards/forensics_strict_edge_reward": 0.0283584640128538,
      "rewards/forensics_tiou_reward": 0.35082603991031647,
      "step": 228
    },
    {
      "completion_length": 23.0625,
      "epoch": 0.7035330261136713,
      "grad_norm": 4.879824161529541,
      "kl": 0.30859375,
      "learning_rate": 8.246153846153846e-07,
      "loss": 0.0548,
      "reward": 0.7358291447162628,
      "reward_std": 0.05787100829184055,
      "reward_std_all": 0.06508222036063671,
      "rewards-max/forensics_sinkhorn_reward": 0.9735794961452484,
      "rewards-max/forensics_strict_edge_reward": 0.25296870060265064,
      "rewards-max/forensics_tiou_reward": 0.9352807998657227,
      "rewards/forensics_sinkhorn_reward": 0.6641064286231995,
      "rewards/forensics_strict_edge_reward": 0.028370089596137404,
      "rewards/forensics_tiou_reward": 0.3930293768644333,
      "step": 229
    },
    {
      "completion_length": 22.796875,
      "epoch": 0.706605222734255,
      "grad_norm": 5.935654163360596,
      "kl": 0.25341796875,
      "learning_rate": 8.238461538461538e-07,
      "loss": 0.1804,
      "reward": 0.6846174895763397,
      "reward_std": 0.10279392451047897,
      "reward_std_all": 0.11027819663286209,
      "rewards-max/forensics_sinkhorn_reward": 0.9762606620788574,
      "rewards-max/forensics_strict_edge_reward": 0.09442757442593575,
      "rewards-max/forensics_tiou_reward": 0.9398332834243774,
      "rewards/forensics_sinkhorn_reward": 0.59345543384552,
      "rewards/forensics_strict_edge_reward": 0.004393811424961314,
      "rewards/forensics_tiou_reward": 0.30011267960071564,
      "step": 230
    },
    {
      "completion_length": 23.546875,
      "epoch": 0.7096774193548387,
      "grad_norm": 4.120591640472412,
      "kl": 0.3076171875,
      "learning_rate": 8.23076923076923e-07,
      "loss": 0.1041,
      "reward": 0.669065535068512,
      "reward_std": 0.09172996506094933,
      "reward_std_all": 0.11195265129208565,
      "rewards-max/forensics_sinkhorn_reward": 0.9519438445568085,
      "rewards-max/forensics_strict_edge_reward": 0.2318475991487503,
      "rewards-max/forensics_tiou_reward": 0.8887537717819214,
      "rewards/forensics_sinkhorn_reward": 0.6004369258880615,
      "rewards/forensics_strict_edge_reward": 0.011847992893308401,
      "rewards/forensics_tiou_reward": 0.3535170406103134,
      "step": 231
    },
    {
      "completion_length": 22.53125,
      "epoch": 0.7127496159754224,
      "grad_norm": 5.315178871154785,
      "kl": 0.3046875,
      "learning_rate": 8.223076923076922e-07,
      "loss": 0.2353,
      "reward": 0.7107355892658234,
      "reward_std": 0.08630114048719406,
      "reward_std_all": 0.08812831714749336,
      "rewards-max/forensics_sinkhorn_reward": 0.9757205545902252,
      "rewards-max/forensics_strict_edge_reward": 0.2858170047402382,
      "rewards-max/forensics_tiou_reward": 0.9439631998538971,
      "rewards/forensics_sinkhorn_reward": 0.6357911229133606,
      "rewards/forensics_strict_edge_reward": 0.020272855646908283,
      "rewards/forensics_tiou_reward": 0.39062416553497314,
      "step": 232
    },
    {
      "completion_length": 22.4375,
      "epoch": 0.7158218125960062,
      "grad_norm": 5.581255912780762,
      "kl": 0.19580078125,
      "learning_rate": 8.215384615384616e-07,
      "loss": 0.1143,
      "reward": 0.7496844828128815,
      "reward_std": 0.04474393930286169,
      "reward_std_all": 0.0511882696300745,
      "rewards-max/forensics_sinkhorn_reward": 0.9908562004566193,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9822283387184143,
      "rewards/forensics_sinkhorn_reward": 0.6650187969207764,
      "rewards/forensics_strict_edge_reward": 0.03266157582402229,
      "rewards/forensics_tiou_reward": 0.35682737827301025,
      "step": 233
    },
    {
      "completion_length": 22.8125,
      "epoch": 0.7188940092165899,
      "grad_norm": 4.495147228240967,
      "kl": 0.30078125,
      "learning_rate": 8.207692307692308e-07,
      "loss": 0.0498,
      "reward": 0.7109193801879883,
      "reward_std": 0.09569581970572472,
      "reward_std_all": 0.09502103179693222,
      "rewards-max/forensics_sinkhorn_reward": 0.9411650896072388,
      "rewards-max/forensics_strict_edge_reward": 0.31722866743803024,
      "rewards-max/forensics_tiou_reward": 0.8497290313243866,
      "rewards/forensics_sinkhorn_reward": 0.644424557685852,
      "rewards/forensics_strict_edge_reward": 0.021013138815760612,
      "rewards/forensics_tiou_reward": 0.35184594988822937,
      "step": 234
    },
    {
      "completion_length": 25.5,
      "epoch": 0.7219662058371735,
      "grad_norm": 5.653595924377441,
      "kl": 0.26416015625,
      "learning_rate": 8.199999999999999e-07,
      "loss": -0.0075,
      "reward": 0.7226854562759399,
      "reward_std": 0.055336516350507736,
      "reward_std_all": 0.07139194756746292,
      "rewards-max/forensics_sinkhorn_reward": 0.9340968728065491,
      "rewards-max/forensics_strict_edge_reward": 0.3661799980327487,
      "rewards-max/forensics_tiou_reward": 0.8207715153694153,
      "rewards/forensics_sinkhorn_reward": 0.6344276964664459,
      "rewards/forensics_strict_edge_reward": 0.017159008042654023,
      "rewards/forensics_tiou_reward": 0.3304154872894287,
      "step": 235
    },
    {
      "completion_length": 21.921875,
      "epoch": 0.7250384024577573,
      "grad_norm": 4.412408828735352,
      "kl": 0.294921875,
      "learning_rate": 8.192307692307692e-07,
      "loss": 0.0361,
      "reward": 0.7251697778701782,
      "reward_std": 0.07445124164223671,
      "reward_std_all": 0.07579655572772026,
      "rewards-max/forensics_sinkhorn_reward": 0.8829619586467743,
      "rewards-max/forensics_strict_edge_reward": 0.08890411397442222,
      "rewards-max/forensics_tiou_reward": 0.7086092233657837,
      "rewards/forensics_sinkhorn_reward": 0.5987952649593353,
      "rewards/forensics_strict_edge_reward": 0.00497619180532638,
      "rewards/forensics_tiou_reward": 0.29529307782649994,
      "step": 236
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.728110599078341,
      "grad_norm": 4.901804447174072,
      "kl": 0.345703125,
      "learning_rate": 8.184615384615384e-07,
      "loss": 0.1088,
      "reward": 0.7541301548480988,
      "reward_std": 0.04868800565600395,
      "reward_std_all": 0.04999152198433876,
      "rewards-max/forensics_sinkhorn_reward": 0.9550349414348602,
      "rewards-max/forensics_strict_edge_reward": 0.19446904212236404,
      "rewards-max/forensics_tiou_reward": 0.8884150087833405,
      "rewards/forensics_sinkhorn_reward": 0.6141410768032074,
      "rewards/forensics_strict_edge_reward": 0.017603048589080572,
      "rewards/forensics_tiou_reward": 0.37311646342277527,
      "step": 237
    },
    {
      "completion_length": 21.96875,
      "epoch": 0.7311827956989247,
      "grad_norm": 4.4662370681762695,
      "kl": 0.3193359375,
      "learning_rate": 8.176923076923077e-07,
      "loss": 0.0733,
      "reward": 0.7580374479293823,
      "reward_std": 0.04811575077474117,
      "reward_std_all": 0.048453956842422485,
      "rewards-max/forensics_sinkhorn_reward": 0.942901611328125,
      "rewards-max/forensics_strict_edge_reward": 0.32127323746681213,
      "rewards-max/forensics_tiou_reward": 0.8548619449138641,
      "rewards/forensics_sinkhorn_reward": 0.6526435017585754,
      "rewards/forensics_strict_edge_reward": 0.021789672784507275,
      "rewards/forensics_tiou_reward": 0.34758342802524567,
      "step": 238
    },
    {
      "completion_length": 26.28125,
      "epoch": 0.7342549923195084,
      "grad_norm": 4.525164604187012,
      "kl": 0.341796875,
      "learning_rate": 8.169230769230769e-07,
      "loss": 0.2567,
      "reward": 0.7141441702842712,
      "reward_std": 0.11362102627754211,
      "reward_std_all": 0.10598810762166977,
      "rewards-max/forensics_sinkhorn_reward": 0.9774658381938934,
      "rewards-max/forensics_strict_edge_reward": 0.21022531017661095,
      "rewards-max/forensics_tiou_reward": 0.9405053853988647,
      "rewards/forensics_sinkhorn_reward": 0.6441550254821777,
      "rewards/forensics_strict_edge_reward": 0.016154690645635128,
      "rewards/forensics_tiou_reward": 0.41275754570961,
      "step": 239
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.7373271889400922,
      "grad_norm": 4.574138164520264,
      "kl": 0.28515625,
      "learning_rate": 8.16153846153846e-07,
      "loss": 0.0864,
      "reward": 0.714754730463028,
      "reward_std": 0.08740680664777756,
      "reward_std_all": 0.08837679773569107,
      "rewards-max/forensics_sinkhorn_reward": 0.9807945191860199,
      "rewards-max/forensics_strict_edge_reward": 0.3653731048107147,
      "rewards-max/forensics_tiou_reward": 0.9556602537631989,
      "rewards/forensics_sinkhorn_reward": 0.6236929297447205,
      "rewards/forensics_strict_edge_reward": 0.024950464721769094,
      "rewards/forensics_tiou_reward": 0.3874892592430115,
      "step": 240
    },
    {
      "completion_length": 25.4375,
      "epoch": 0.7403993855606759,
      "grad_norm": 5.182683944702148,
      "kl": 0.322265625,
      "learning_rate": 8.153846153846154e-07,
      "loss": 0.0985,
      "reward": 0.773391842842102,
      "reward_std": 0.04517829138785601,
      "reward_std_all": 0.042932769283652306,
      "rewards-max/forensics_sinkhorn_reward": 0.8960436284542084,
      "rewards-max/forensics_strict_edge_reward": 0.13581721056834795,
      "rewards-max/forensics_tiou_reward": 0.7286640703678131,
      "rewards/forensics_sinkhorn_reward": 0.671199381351471,
      "rewards/forensics_strict_edge_reward": 0.006297944351899787,
      "rewards/forensics_tiou_reward": 0.39607082307338715,
      "step": 241
    },
    {
      "completion_length": 23.21875,
      "epoch": 0.7434715821812596,
      "grad_norm": 6.112059116363525,
      "kl": 0.2470703125,
      "learning_rate": 8.146153846153846e-07,
      "loss": 0.1615,
      "reward": 0.7415451109409332,
      "reward_std": 0.06765593029558659,
      "reward_std_all": 0.067054009065032,
      "rewards-max/forensics_sinkhorn_reward": 0.9578644335269928,
      "rewards-max/forensics_strict_edge_reward": 0.5064293146133423,
      "rewards-max/forensics_tiou_reward": 0.9082694053649902,
      "rewards/forensics_sinkhorn_reward": 0.6532285213470459,
      "rewards/forensics_strict_edge_reward": 0.026178543455898762,
      "rewards/forensics_tiou_reward": 0.3606088310480118,
      "step": 242
    },
    {
      "completion_length": 23.203125,
      "epoch": 0.7465437788018433,
      "grad_norm": 6.200862884521484,
      "kl": 0.3955078125,
      "learning_rate": 8.138461538461538e-07,
      "loss": 0.0898,
      "reward": 0.7564707100391388,
      "reward_std": 0.03936810093000531,
      "reward_std_all": 0.04722985252737999,
      "rewards-max/forensics_sinkhorn_reward": 0.9597988426685333,
      "rewards-max/forensics_strict_edge_reward": 0.052367767319083214,
      "rewards-max/forensics_tiou_reward": 0.907819390296936,
      "rewards/forensics_sinkhorn_reward": 0.6760427951812744,
      "rewards/forensics_strict_edge_reward": 0.006811261351685971,
      "rewards/forensics_tiou_reward": 0.4586354196071625,
      "step": 243
    },
    {
      "completion_length": 26.640625,
      "epoch": 0.7496159754224271,
      "grad_norm": 4.953660011291504,
      "kl": 0.265625,
      "learning_rate": 8.13076923076923e-07,
      "loss": 0.1742,
      "reward": 0.7610979080200195,
      "reward_std": 0.05989600718021393,
      "reward_std_all": 0.05604473128914833,
      "rewards-max/forensics_sinkhorn_reward": 0.9471487998962402,
      "rewards-max/forensics_strict_edge_reward": 0.12351470440626144,
      "rewards-max/forensics_tiou_reward": 0.8767999112606049,
      "rewards/forensics_sinkhorn_reward": 0.6829396784305573,
      "rewards/forensics_strict_edge_reward": 0.010095731122419238,
      "rewards/forensics_tiou_reward": 0.44668909907341003,
      "step": 244
    },
    {
      "completion_length": 22.265625,
      "epoch": 0.7526881720430108,
      "grad_norm": 4.3272528648376465,
      "kl": 0.3017578125,
      "learning_rate": 8.123076923076922e-07,
      "loss": 0.0591,
      "reward": 0.708684116601944,
      "reward_std": 0.0500837042927742,
      "reward_std_all": 0.07210523076355457,
      "rewards-max/forensics_sinkhorn_reward": 0.8968393504619598,
      "rewards-max/forensics_strict_edge_reward": 0.01100675156340003,
      "rewards-max/forensics_tiou_reward": 0.7327249348163605,
      "rewards/forensics_sinkhorn_reward": 0.6225424408912659,
      "rewards/forensics_strict_edge_reward": 0.0008806507685221732,
      "rewards/forensics_tiou_reward": 0.3296092599630356,
      "step": 245
    },
    {
      "completion_length": 23.859375,
      "epoch": 0.7557603686635944,
      "grad_norm": 5.18081521987915,
      "kl": 0.337890625,
      "learning_rate": 8.115384615384615e-07,
      "loss": 0.1782,
      "reward": 0.7126737833023071,
      "reward_std": 0.06954910233616829,
      "reward_std_all": 0.08030804246664047,
      "rewards-max/forensics_sinkhorn_reward": 0.9499187767505646,
      "rewards-max/forensics_strict_edge_reward": 0.026573223527520895,
      "rewards-max/forensics_tiou_reward": 0.8686620593070984,
      "rewards/forensics_sinkhorn_reward": 0.6250316798686981,
      "rewards/forensics_strict_edge_reward": 0.001994547463255003,
      "rewards/forensics_tiou_reward": 0.365469828248024,
      "step": 246
    },
    {
      "completion_length": 25.234375,
      "epoch": 0.7588325652841782,
      "grad_norm": 4.883122444152832,
      "kl": 0.3173828125,
      "learning_rate": 8.107692307692308e-07,
      "loss": 0.0703,
      "reward": 0.7114982604980469,
      "reward_std": 0.061088623479008675,
      "reward_std_all": 0.07467616721987724,
      "rewards-max/forensics_sinkhorn_reward": 0.9764444530010223,
      "rewards-max/forensics_strict_edge_reward": 0.4010360911488533,
      "rewards-max/forensics_tiou_reward": 0.9479314386844635,
      "rewards/forensics_sinkhorn_reward": 0.6511180102825165,
      "rewards/forensics_strict_edge_reward": 0.021486511453986168,
      "rewards/forensics_tiou_reward": 0.4016834944486618,
      "step": 247
    },
    {
      "completion_length": 24.984375,
      "epoch": 0.7619047619047619,
      "grad_norm": 5.041945457458496,
      "kl": 0.2763671875,
      "learning_rate": 8.1e-07,
      "loss": 0.0559,
      "reward": 0.7655148804187775,
      "reward_std": 0.052260374650359154,
      "reward_std_all": 0.049763066694140434,
      "rewards-max/forensics_sinkhorn_reward": 0.9186955392360687,
      "rewards-max/forensics_strict_edge_reward": 0.18909518793225288,
      "rewards-max/forensics_tiou_reward": 0.812380850315094,
      "rewards/forensics_sinkhorn_reward": 0.7036595940589905,
      "rewards/forensics_strict_edge_reward": 0.017070486210286617,
      "rewards/forensics_tiou_reward": 0.4043332040309906,
      "step": 248
    },
    {
      "completion_length": 25.234375,
      "epoch": 0.7649769585253456,
      "grad_norm": 5.213740825653076,
      "kl": 0.3427734375,
      "learning_rate": 8.092307692307692e-07,
      "loss": 0.0525,
      "reward": 0.7457149922847748,
      "reward_std": 0.05210607312619686,
      "reward_std_all": 0.0551172848790884,
      "rewards-max/forensics_sinkhorn_reward": 0.9173265099525452,
      "rewards-max/forensics_strict_edge_reward": 0.08345298952190205,
      "rewards-max/forensics_tiou_reward": 0.7987596690654755,
      "rewards/forensics_sinkhorn_reward": 0.6693086922168732,
      "rewards/forensics_strict_edge_reward": 0.004309992025810061,
      "rewards/forensics_tiou_reward": 0.41477328538894653,
      "step": 249
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.7680491551459293,
      "grad_norm": 4.354286193847656,
      "kl": 0.3125,
      "learning_rate": 8.084615384615384e-07,
      "loss": 0.0009,
      "reward": 0.7309122681617737,
      "reward_std": 0.04485316202044487,
      "reward_std_all": 0.05704641342163086,
      "rewards-max/forensics_sinkhorn_reward": 0.9431789517402649,
      "rewards-max/forensics_strict_edge_reward": 0.460111266002059,
      "rewards-max/forensics_tiou_reward": 0.8721075654029846,
      "rewards/forensics_sinkhorn_reward": 0.6765971779823303,
      "rewards/forensics_strict_edge_reward": 0.020083147566765547,
      "rewards/forensics_tiou_reward": 0.3638397753238678,
      "step": 250
    },
    {
      "completion_length": 24.484375,
      "epoch": 0.7711213517665131,
      "grad_norm": 5.151031970977783,
      "kl": 0.427734375,
      "learning_rate": 8.076923076923077e-07,
      "loss": 0.0048,
      "reward": 0.7027541995048523,
      "reward_std": 0.0681388508528471,
      "reward_std_all": 0.08600303903222084,
      "rewards-max/forensics_sinkhorn_reward": 0.943649023771286,
      "rewards-max/forensics_strict_edge_reward": 0.39635732662281953,
      "rewards-max/forensics_tiou_reward": 0.875056803226471,
      "rewards/forensics_sinkhorn_reward": 0.623783141374588,
      "rewards/forensics_strict_edge_reward": 0.04497911487487727,
      "rewards/forensics_tiou_reward": 0.3413987457752228,
      "step": 251
    },
    {
      "completion_length": 23.0,
      "epoch": 0.7741935483870968,
      "grad_norm": 6.8619818687438965,
      "kl": 0.27734375,
      "learning_rate": 8.069230769230769e-07,
      "loss": 0.1378,
      "reward": 0.6996592879295349,
      "reward_std": 0.10445890948176384,
      "reward_std_all": 0.10322392731904984,
      "rewards-max/forensics_sinkhorn_reward": 0.9627162516117096,
      "rewards-max/forensics_strict_edge_reward": 0.1309044286608696,
      "rewards-max/forensics_tiou_reward": 0.9009429216384888,
      "rewards/forensics_sinkhorn_reward": 0.5961721241474152,
      "rewards/forensics_strict_edge_reward": 0.00939383078366518,
      "rewards/forensics_tiou_reward": 0.3391399011015892,
      "step": 252
    },
    {
      "completion_length": 26.125,
      "epoch": 0.7772657450076805,
      "grad_norm": 4.365909576416016,
      "kl": 0.2744140625,
      "learning_rate": 8.06153846153846e-07,
      "loss": 0.1968,
      "reward": 0.7191071212291718,
      "reward_std": 0.08884645067155361,
      "reward_std_all": 0.08841427974402905,
      "rewards-max/forensics_sinkhorn_reward": 0.954598605632782,
      "rewards-max/forensics_strict_edge_reward": 0.118141770362854,
      "rewards-max/forensics_tiou_reward": 0.8929459154605865,
      "rewards/forensics_sinkhorn_reward": 0.6179481744766235,
      "rewards/forensics_strict_edge_reward": 0.009173786267638206,
      "rewards/forensics_tiou_reward": 0.3948448598384857,
      "step": 253
    },
    {
      "completion_length": 25.484375,
      "epoch": 0.7803379416282642,
      "grad_norm": 4.502674102783203,
      "kl": 0.4501953125,
      "learning_rate": 8.053846153846154e-07,
      "loss": 0.1348,
      "reward": 0.6987678706645966,
      "reward_std": 0.07899901643395424,
      "reward_std_all": 0.090611282736063,
      "rewards-max/forensics_sinkhorn_reward": 0.9162541925907135,
      "rewards-max/forensics_strict_edge_reward": 0.09668138157576323,
      "rewards-max/forensics_tiou_reward": 0.7893295884132385,
      "rewards/forensics_sinkhorn_reward": 0.5392085909843445,
      "rewards/forensics_strict_edge_reward": 0.006187671446241438,
      "rewards/forensics_tiou_reward": 0.3028506338596344,
      "step": 254
    },
    {
      "completion_length": 25.078125,
      "epoch": 0.783410138248848,
      "grad_norm": 4.007931232452393,
      "kl": 0.310546875,
      "learning_rate": 8.046153846153846e-07,
      "loss": 0.2625,
      "reward": 0.7344664037227631,
      "reward_std": 0.08187410980463028,
      "reward_std_all": 0.07808592915534973,
      "rewards-max/forensics_sinkhorn_reward": 0.8708572089672089,
      "rewards-max/forensics_strict_edge_reward": 0.02601626329123974,
      "rewards-max/forensics_tiou_reward": 0.6882697641849518,
      "rewards/forensics_sinkhorn_reward": 0.6507926285266876,
      "rewards/forensics_strict_edge_reward": 0.0017719421302899718,
      "rewards/forensics_tiou_reward": 0.3752298504114151,
      "step": 255
    },
    {
      "completion_length": 26.578125,
      "epoch": 0.7864823348694316,
      "grad_norm": 3.835949659347534,
      "kl": 0.2333984375,
      "learning_rate": 8.038461538461538e-07,
      "loss": 0.1897,
      "reward": 0.7188631296157837,
      "reward_std": 0.09994343295693398,
      "reward_std_all": 0.09703970700502396,
      "rewards-max/forensics_sinkhorn_reward": 0.9308961033821106,
      "rewards-max/forensics_strict_edge_reward": 0.35049770772457123,
      "rewards-max/forensics_tiou_reward": 0.834295392036438,
      "rewards/forensics_sinkhorn_reward": 0.6131581962108612,
      "rewards/forensics_strict_edge_reward": 0.018258281983435154,
      "rewards/forensics_tiou_reward": 0.3347260504961014,
      "step": 256
    },
    {
      "completion_length": 25.5,
      "epoch": 0.7895545314900153,
      "grad_norm": 5.829096794128418,
      "kl": 0.2861328125,
      "learning_rate": 8.03076923076923e-07,
      "loss": 0.0219,
      "reward": 0.7094749212265015,
      "reward_std": 0.07934527099132538,
      "reward_std_all": 0.08566857501864433,
      "rewards-max/forensics_sinkhorn_reward": 0.9849169552326202,
      "rewards-max/forensics_strict_edge_reward": 0.33644743263721466,
      "rewards-max/forensics_tiou_reward": 0.9662609994411469,
      "rewards/forensics_sinkhorn_reward": 0.5693491548299789,
      "rewards/forensics_strict_edge_reward": 0.023094769567251205,
      "rewards/forensics_tiou_reward": 0.41097623109817505,
      "step": 257
    },
    {
      "completion_length": 25.484375,
      "epoch": 0.7926267281105991,
      "grad_norm": 4.741184234619141,
      "kl": 0.236328125,
      "learning_rate": 8.023076923076922e-07,
      "loss": 0.1249,
      "reward": 0.7296726405620575,
      "reward_std": 0.0966964103281498,
      "reward_std_all": 0.08999524265527725,
      "rewards-max/forensics_sinkhorn_reward": 0.9275553822517395,
      "rewards-max/forensics_strict_edge_reward": 0.20831449516117573,
      "rewards-max/forensics_tiou_reward": 0.8362731635570526,
      "rewards/forensics_sinkhorn_reward": 0.6320928633213043,
      "rewards/forensics_strict_edge_reward": 0.013556137564592063,
      "rewards/forensics_tiou_reward": 0.3825366944074631,
      "step": 258
    },
    {
      "completion_length": 26.3125,
      "epoch": 0.7956989247311828,
      "grad_norm": 5.336219310760498,
      "kl": 0.328125,
      "learning_rate": 8.015384615384615e-07,
      "loss": 0.1284,
      "reward": 0.74684277176857,
      "reward_std": 0.04946681670844555,
      "reward_std_all": 0.05359136499464512,
      "rewards-max/forensics_sinkhorn_reward": 0.9650745391845703,
      "rewards-max/forensics_strict_edge_reward": 0.1803518794476986,
      "rewards-max/forensics_tiou_reward": 0.9199700057506561,
      "rewards/forensics_sinkhorn_reward": 0.6396794319152832,
      "rewards/forensics_strict_edge_reward": 0.01246474776417017,
      "rewards/forensics_tiou_reward": 0.36011040210723877,
      "step": 259
    },
    {
      "completion_length": 24.328125,
      "epoch": 0.7987711213517665,
      "grad_norm": 5.3160786628723145,
      "kl": 0.322265625,
      "learning_rate": 8.007692307692308e-07,
      "loss": 0.0761,
      "reward": 0.752243310213089,
      "reward_std": 0.042003890965133905,
      "reward_std_all": 0.04923712043091655,
      "rewards-max/forensics_sinkhorn_reward": 0.9707094728946686,
      "rewards-max/forensics_strict_edge_reward": 0.4235072433948517,
      "rewards-max/forensics_tiou_reward": 0.9312589764595032,
      "rewards/forensics_sinkhorn_reward": 0.6530576050281525,
      "rewards/forensics_strict_edge_reward": 0.019401563797146082,
      "rewards/forensics_tiou_reward": 0.3818538635969162,
      "step": 260
    },
    {
      "completion_length": 25.5,
      "epoch": 0.8018433179723502,
      "grad_norm": 6.425746917724609,
      "kl": 0.4814453125,
      "learning_rate": 8e-07,
      "loss": 0.1434,
      "reward": 0.7551786601543427,
      "reward_std": 0.04727778211236,
      "reward_std_all": 0.05146810971200466,
      "rewards-max/forensics_sinkhorn_reward": 0.9347720146179199,
      "rewards-max/forensics_strict_edge_reward": 0.17633773386478424,
      "rewards-max/forensics_tiou_reward": 0.8484395146369934,
      "rewards/forensics_sinkhorn_reward": 0.6572072207927704,
      "rewards/forensics_strict_edge_reward": 0.007649358827620745,
      "rewards/forensics_tiou_reward": 0.3765058368444443,
      "step": 261
    },
    {
      "completion_length": 22.4375,
      "epoch": 0.804915514592934,
      "grad_norm": 4.553653717041016,
      "kl": 0.3515625,
      "learning_rate": 7.992307692307692e-07,
      "loss": 0.1301,
      "reward": 0.7523095309734344,
      "reward_std": 0.06304671987891197,
      "reward_std_all": 0.06043292209506035,
      "rewards-max/forensics_sinkhorn_reward": 0.9698308706283569,
      "rewards-max/forensics_strict_edge_reward": 0.20645379647612572,
      "rewards-max/forensics_tiou_reward": 0.9299036264419556,
      "rewards/forensics_sinkhorn_reward": 0.683692455291748,
      "rewards/forensics_strict_edge_reward": 0.012455023068469018,
      "rewards/forensics_tiou_reward": 0.39825353026390076,
      "step": 262
    },
    {
      "completion_length": 25.578125,
      "epoch": 0.8079877112135176,
      "grad_norm": 4.580052375793457,
      "kl": 0.28125,
      "learning_rate": 7.984615384615384e-07,
      "loss": 0.0571,
      "reward": 0.7609669268131256,
      "reward_std": 0.0389129975810647,
      "reward_std_all": 0.04201574623584747,
      "rewards-max/forensics_sinkhorn_reward": 0.9296020269393921,
      "rewards-max/forensics_strict_edge_reward": 0.1846668004291132,
      "rewards-max/forensics_tiou_reward": 0.8464520275592804,
      "rewards/forensics_sinkhorn_reward": 0.7024769484996796,
      "rewards/forensics_strict_edge_reward": 0.012796092603821307,
      "rewards/forensics_tiou_reward": 0.40015581250190735,
      "step": 263
    },
    {
      "completion_length": 27.390625,
      "epoch": 0.8110599078341014,
      "grad_norm": 4.783110618591309,
      "kl": 0.296875,
      "learning_rate": 7.976923076923077e-07,
      "loss": 0.2537,
      "reward": 0.723940908908844,
      "reward_std": 0.07028837315738201,
      "reward_std_all": 0.07724399864673615,
      "rewards-max/forensics_sinkhorn_reward": 0.9579587876796722,
      "rewards-max/forensics_strict_edge_reward": 0.35922926664352417,
      "rewards-max/forensics_tiou_reward": 0.8961041569709778,
      "rewards/forensics_sinkhorn_reward": 0.6098937690258026,
      "rewards/forensics_strict_edge_reward": 0.018758569844067097,
      "rewards/forensics_tiou_reward": 0.4051879048347473,
      "step": 264
    },
    {
      "completion_length": 24.046875,
      "epoch": 0.8141321044546851,
      "grad_norm": 3.923250675201416,
      "kl": 0.23828125,
      "learning_rate": 7.969230769230769e-07,
      "loss": 0.1102,
      "reward": 0.7365060150623322,
      "reward_std": 0.04423149116337299,
      "reward_std_all": 0.0548733863979578,
      "rewards-max/forensics_sinkhorn_reward": 0.9275980293750763,
      "rewards-max/forensics_strict_edge_reward": 0.14295656001195312,
      "rewards-max/forensics_tiou_reward": 0.8213159441947937,
      "rewards/forensics_sinkhorn_reward": 0.6002789735794067,
      "rewards/forensics_strict_edge_reward": 0.006352142954710871,
      "rewards/forensics_tiou_reward": 0.2985354959964752,
      "step": 265
    },
    {
      "completion_length": 24.6875,
      "epoch": 0.8172043010752689,
      "grad_norm": 4.388193607330322,
      "kl": 0.25537109375,
      "learning_rate": 7.96153846153846e-07,
      "loss": 0.1147,
      "reward": 0.7103153765201569,
      "reward_std": 0.08917748183012009,
      "reward_std_all": 0.09111903235316277,
      "rewards-max/forensics_sinkhorn_reward": 0.9525988698005676,
      "rewards-max/forensics_strict_edge_reward": 0.24959099292755127,
      "rewards-max/forensics_tiou_reward": 0.8682361543178558,
      "rewards/forensics_sinkhorn_reward": 0.6439236402511597,
      "rewards/forensics_strict_edge_reward": 0.01674196543172002,
      "rewards/forensics_tiou_reward": 0.39695167541503906,
      "step": 266
    },
    {
      "completion_length": 33.296875,
      "epoch": 0.8202764976958525,
      "grad_norm": 4.30687141418457,
      "kl": 0.22265625,
      "learning_rate": 7.953846153846154e-07,
      "loss": 0.1233,
      "reward": 0.7118294537067413,
      "reward_std": 0.08429055009037256,
      "reward_std_all": 0.08820042572915554,
      "rewards-max/forensics_sinkhorn_reward": 0.9208289682865143,
      "rewards-max/forensics_strict_edge_reward": 0.19277227809652686,
      "rewards-max/forensics_tiou_reward": 0.8097359538078308,
      "rewards/forensics_sinkhorn_reward": 0.5674940943717957,
      "rewards/forensics_strict_edge_reward": 0.01263452009879984,
      "rewards/forensics_tiou_reward": 0.32732440531253815,
      "step": 267
    },
    {
      "completion_length": 24.46875,
      "epoch": 0.8233486943164362,
      "grad_norm": 5.0788798332214355,
      "kl": 0.2529296875,
      "learning_rate": 7.946153846153846e-07,
      "loss": 0.0681,
      "reward": 0.7042532563209534,
      "reward_std": 0.09305993467569351,
      "reward_std_all": 0.09457523003220558,
      "rewards-max/forensics_sinkhorn_reward": 0.9620408117771149,
      "rewards-max/forensics_strict_edge_reward": 0.18443760368973017,
      "rewards-max/forensics_tiou_reward": 0.9000942707061768,
      "rewards/forensics_sinkhorn_reward": 0.6194361448287964,
      "rewards/forensics_strict_edge_reward": 0.012640130880754441,
      "rewards/forensics_tiou_reward": 0.3535468876361847,
      "step": 268
    },
    {
      "completion_length": 25.328125,
      "epoch": 0.82642089093702,
      "grad_norm": 5.7245378494262695,
      "kl": 0.2353515625,
      "learning_rate": 7.938461538461538e-07,
      "loss": 0.0408,
      "reward": 0.7168485522270203,
      "reward_std": 0.09969986602663994,
      "reward_std_all": 0.09551751241087914,
      "rewards-max/forensics_sinkhorn_reward": 0.9676398038864136,
      "rewards-max/forensics_strict_edge_reward": 0.3706423342227936,
      "rewards-max/forensics_tiou_reward": 0.9209119975566864,
      "rewards/forensics_sinkhorn_reward": 0.6506679356098175,
      "rewards/forensics_strict_edge_reward": 0.020593533758074045,
      "rewards/forensics_tiou_reward": 0.4341921955347061,
      "step": 269
    },
    {
      "completion_length": 26.296875,
      "epoch": 0.8294930875576036,
      "grad_norm": 4.230697154998779,
      "kl": 0.23681640625,
      "learning_rate": 7.930769230769231e-07,
      "loss": 0.1945,
      "reward": 0.7008492350578308,
      "reward_std": 0.1080688014626503,
      "reward_std_all": 0.10505619458854198,
      "rewards-max/forensics_sinkhorn_reward": 0.9680174589157104,
      "rewards-max/forensics_strict_edge_reward": 0.2592229098081589,
      "rewards-max/forensics_tiou_reward": 0.9266889691352844,
      "rewards/forensics_sinkhorn_reward": 0.5839362442493439,
      "rewards/forensics_strict_edge_reward": 0.012122190557420254,
      "rewards/forensics_tiou_reward": 0.38726599514484406,
      "step": 270
    },
    {
      "completion_length": 25.6875,
      "epoch": 0.8325652841781874,
      "grad_norm": 4.555980205535889,
      "kl": 0.2900390625,
      "learning_rate": 7.923076923076922e-07,
      "loss": 0.1066,
      "reward": 0.6817401647567749,
      "reward_std": 0.0934850163757801,
      "reward_std_all": 0.10383686795830727,
      "rewards-max/forensics_sinkhorn_reward": 0.9801746606826782,
      "rewards-max/forensics_strict_edge_reward": 0.3230769634246826,
      "rewards-max/forensics_tiou_reward": 0.9474005699157715,
      "rewards/forensics_sinkhorn_reward": 0.5916937589645386,
      "rewards/forensics_strict_edge_reward": 0.03290106542408466,
      "rewards/forensics_tiou_reward": 0.33720920979976654,
      "step": 271
    },
    {
      "completion_length": 25.15625,
      "epoch": 0.8356374807987711,
      "grad_norm": 3.9161062240600586,
      "kl": 0.271484375,
      "learning_rate": 7.915384615384615e-07,
      "loss": 0.0299,
      "reward": 0.7940370440483093,
      "reward_std": 0.01770540326833725,
      "reward_std_all": 0.018140370026230812,
      "rewards-max/forensics_sinkhorn_reward": 0.9870138466358185,
      "rewards-max/forensics_strict_edge_reward": 0.6347885429859161,
      "rewards-max/forensics_tiou_reward": 0.9715569019317627,
      "rewards/forensics_sinkhorn_reward": 0.7256059944629669,
      "rewards/forensics_strict_edge_reward": 0.04224783182144165,
      "rewards/forensics_tiou_reward": 0.5153276920318604,
      "step": 272
    },
    {
      "completion_length": 25.40625,
      "epoch": 0.8387096774193549,
      "grad_norm": 12.118537902832031,
      "kl": 0.810546875,
      "learning_rate": 7.907692307692307e-07,
      "loss": 0.1342,
      "reward": 0.7209584414958954,
      "reward_std": 0.039786240085959435,
      "reward_std_all": 0.06282072886824608,
      "rewards-max/forensics_sinkhorn_reward": 0.9572558701038361,
      "rewards-max/forensics_strict_edge_reward": 0.30113139748573303,
      "rewards-max/forensics_tiou_reward": 0.9041696488857269,
      "rewards/forensics_sinkhorn_reward": 0.6301787495613098,
      "rewards/forensics_strict_edge_reward": 0.027869985438883305,
      "rewards/forensics_tiou_reward": 0.4380412846803665,
      "step": 273
    },
    {
      "completion_length": 28.25,
      "epoch": 0.8417818740399385,
      "grad_norm": 5.6038079261779785,
      "kl": 0.2392578125,
      "learning_rate": 7.9e-07,
      "loss": -0.0066,
      "reward": 0.6905644237995148,
      "reward_std": 0.10797831043601036,
      "reward_std_all": 0.1092749610543251,
      "rewards-max/forensics_sinkhorn_reward": 0.9523365795612335,
      "rewards-max/forensics_strict_edge_reward": 0.2337907776236534,
      "rewards-max/forensics_tiou_reward": 0.8953344225883484,
      "rewards/forensics_sinkhorn_reward": 0.5724833309650421,
      "rewards/forensics_strict_edge_reward": 0.016120712272822857,
      "rewards/forensics_tiou_reward": 0.3878564089536667,
      "step": 274
    },
    {
      "completion_length": 26.078125,
      "epoch": 0.8448540706605223,
      "grad_norm": 4.084033966064453,
      "kl": 0.2783203125,
      "learning_rate": 7.892307692307692e-07,
      "loss": 0.1103,
      "reward": 0.7351576685905457,
      "reward_std": 0.05418616347014904,
      "reward_std_all": 0.06140579842031002,
      "rewards-max/forensics_sinkhorn_reward": 0.9329347610473633,
      "rewards-max/forensics_strict_edge_reward": 0.1796595575287938,
      "rewards-max/forensics_tiou_reward": 0.8257084786891937,
      "rewards/forensics_sinkhorn_reward": 0.6743606626987457,
      "rewards/forensics_strict_edge_reward": 0.009573508403263986,
      "rewards/forensics_tiou_reward": 0.4093363434076309,
      "step": 275
    },
    {
      "completion_length": 24.671875,
      "epoch": 0.847926267281106,
      "grad_norm": 4.262040615081787,
      "kl": 0.291015625,
      "learning_rate": 7.884615384615384e-07,
      "loss": 0.0833,
      "reward": 0.7509112656116486,
      "reward_std": 0.04879472032189369,
      "reward_std_all": 0.05138329416513443,
      "rewards-max/forensics_sinkhorn_reward": 0.9850480556488037,
      "rewards-max/forensics_strict_edge_reward": 0.7311047911643982,
      "rewards-max/forensics_tiou_reward": 0.9704148471355438,
      "rewards/forensics_sinkhorn_reward": 0.656895786523819,
      "rewards/forensics_strict_edge_reward": 0.0426906943321228,
      "rewards/forensics_tiou_reward": 0.4048893451690674,
      "step": 276
    },
    {
      "completion_length": 27.296875,
      "epoch": 0.8509984639016898,
      "grad_norm": 4.049679756164551,
      "kl": 0.30859375,
      "learning_rate": 7.876923076923077e-07,
      "loss": 0.1253,
      "reward": 0.7218156456947327,
      "reward_std": 0.09187478199601173,
      "reward_std_all": 0.08984985947608948,
      "rewards-max/forensics_sinkhorn_reward": 0.988174319267273,
      "rewards-max/forensics_strict_edge_reward": 0.469220295548439,
      "rewards-max/forensics_tiou_reward": 0.9743853509426117,
      "rewards/forensics_sinkhorn_reward": 0.6641445755958557,
      "rewards/forensics_strict_edge_reward": 0.028622944839298725,
      "rewards/forensics_tiou_reward": 0.4040311574935913,
      "step": 277
    },
    {
      "completion_length": 26.359375,
      "epoch": 0.8540706605222734,
      "grad_norm": 4.151381492614746,
      "kl": 0.275390625,
      "learning_rate": 7.869230769230769e-07,
      "loss": 0.0957,
      "reward": 0.7211277186870575,
      "reward_std": 0.07591012865304947,
      "reward_std_all": 0.07856893911957741,
      "rewards-max/forensics_sinkhorn_reward": 0.9353167414665222,
      "rewards-max/forensics_strict_edge_reward": 0.06890122219920158,
      "rewards-max/forensics_tiou_reward": 0.8348047137260437,
      "rewards/forensics_sinkhorn_reward": 0.6179069876670837,
      "rewards/forensics_strict_edge_reward": 0.004075767181348056,
      "rewards/forensics_tiou_reward": 0.33222775161266327,
      "step": 278
    },
    {
      "completion_length": 24.25,
      "epoch": 0.8571428571428571,
      "grad_norm": 4.595129013061523,
      "kl": 0.3515625,
      "learning_rate": 7.861538461538461e-07,
      "loss": 0.0665,
      "reward": 0.7315496802330017,
      "reward_std": 0.06761333346366882,
      "reward_std_all": 0.0713946521282196,
      "rewards-max/forensics_sinkhorn_reward": 0.9235555529594421,
      "rewards-max/forensics_strict_edge_reward": 0.1733165681362152,
      "rewards-max/forensics_tiou_reward": 0.8288706243038177,
      "rewards/forensics_sinkhorn_reward": 0.6140186190605164,
      "rewards/forensics_strict_edge_reward": 0.009256073273718357,
      "rewards/forensics_tiou_reward": 0.3695206791162491,
      "step": 279
    },
    {
      "completion_length": 24.0625,
      "epoch": 0.8602150537634409,
      "grad_norm": 5.028102874755859,
      "kl": 0.541015625,
      "learning_rate": 7.853846153846153e-07,
      "loss": 0.1268,
      "reward": 0.7513160705566406,
      "reward_std": 0.07803966477513313,
      "reward_std_all": 0.07169101759791374,
      "rewards-max/forensics_sinkhorn_reward": 0.922698050737381,
      "rewards-max/forensics_strict_edge_reward": 0.051623858511447906,
      "rewards-max/forensics_tiou_reward": 0.819439709186554,
      "rewards/forensics_sinkhorn_reward": 0.6781514286994934,
      "rewards/forensics_strict_edge_reward": 0.0038017850602045655,
      "rewards/forensics_tiou_reward": 0.41111353039741516,
      "step": 280
    },
    {
      "completion_length": 24.0,
      "epoch": 0.8632872503840245,
      "grad_norm": 4.814053535461426,
      "kl": 0.2802734375,
      "learning_rate": 7.846153846153846e-07,
      "loss": 0.1567,
      "reward": 0.727038562297821,
      "reward_std": 0.07956260815262794,
      "reward_std_all": 0.08051066100597382,
      "rewards-max/forensics_sinkhorn_reward": 0.9842165410518646,
      "rewards-max/forensics_strict_edge_reward": 0.6284831166267395,
      "rewards-max/forensics_tiou_reward": 0.9662265181541443,
      "rewards/forensics_sinkhorn_reward": 0.646048367023468,
      "rewards/forensics_strict_edge_reward": 0.03351854532957077,
      "rewards/forensics_tiou_reward": 0.40569475293159485,
      "step": 281
    },
    {
      "completion_length": 24.9375,
      "epoch": 0.8663594470046083,
      "grad_norm": 4.751376152038574,
      "kl": 0.2939453125,
      "learning_rate": 7.838461538461539e-07,
      "loss": 0.1111,
      "reward": 0.7100696563720703,
      "reward_std": 0.05728747509419918,
      "reward_std_all": 0.07394921034574509,
      "rewards-max/forensics_sinkhorn_reward": 0.9270657598972321,
      "rewards-max/forensics_strict_edge_reward": 0.22008467931300402,
      "rewards-max/forensics_tiou_reward": 0.8205093145370483,
      "rewards/forensics_sinkhorn_reward": 0.6124365031719208,
      "rewards/forensics_strict_edge_reward": 0.017745740828104317,
      "rewards/forensics_tiou_reward": 0.3258894234895706,
      "step": 282
    },
    {
      "completion_length": 24.1875,
      "epoch": 0.869431643625192,
      "grad_norm": 4.95525598526001,
      "kl": 0.3544921875,
      "learning_rate": 7.830769230769231e-07,
      "loss": 0.0883,
      "reward": 0.7119972407817841,
      "reward_std": 0.04394926642999053,
      "reward_std_all": 0.06600666977465153,
      "rewards-max/forensics_sinkhorn_reward": 0.9568223059177399,
      "rewards-max/forensics_strict_edge_reward": 0.0753604955971241,
      "rewards-max/forensics_tiou_reward": 0.8859187662601471,
      "rewards/forensics_sinkhorn_reward": 0.6218676567077637,
      "rewards/forensics_strict_edge_reward": 0.005104830197524279,
      "rewards/forensics_tiou_reward": 0.3542194962501526,
      "step": 283
    },
    {
      "completion_length": 23.3125,
      "epoch": 0.8725038402457758,
      "grad_norm": 4.349712371826172,
      "kl": 0.3212890625,
      "learning_rate": 7.823076923076922e-07,
      "loss": 0.1629,
      "reward": 0.752837598323822,
      "reward_std": 0.05873511917889118,
      "reward_std_all": 0.05701073631644249,
      "rewards-max/forensics_sinkhorn_reward": 0.9746110439300537,
      "rewards-max/forensics_strict_edge_reward": 0.4032929688692093,
      "rewards-max/forensics_tiou_reward": 0.9409944117069244,
      "rewards/forensics_sinkhorn_reward": 0.6813020408153534,
      "rewards/forensics_strict_edge_reward": 0.025106164626777172,
      "rewards/forensics_tiou_reward": 0.43119674921035767,
      "step": 284
    },
    {
      "completion_length": 26.03125,
      "epoch": 0.8755760368663594,
      "grad_norm": 4.670973777770996,
      "kl": 0.3203125,
      "learning_rate": 7.815384615384615e-07,
      "loss": 0.1956,
      "reward": 0.7286402583122253,
      "reward_std": 0.08486510068178177,
      "reward_std_all": 0.08153839409351349,
      "rewards-max/forensics_sinkhorn_reward": 0.9530328214168549,
      "rewards-max/forensics_strict_edge_reward": 0.11978623643517494,
      "rewards-max/forensics_tiou_reward": 0.8858204782009125,
      "rewards/forensics_sinkhorn_reward": 0.6629641652107239,
      "rewards/forensics_strict_edge_reward": 0.011808286421000957,
      "rewards/forensics_tiou_reward": 0.39265820384025574,
      "step": 285
    },
    {
      "completion_length": 25.546875,
      "epoch": 0.8786482334869432,
      "grad_norm": 4.900998115539551,
      "kl": 0.259765625,
      "learning_rate": 7.807692307692307e-07,
      "loss": -0.0123,
      "reward": 0.716162383556366,
      "reward_std": 0.09074003994464874,
      "reward_std_all": 0.09014752507209778,
      "rewards-max/forensics_sinkhorn_reward": 0.9280797243118286,
      "rewards-max/forensics_strict_edge_reward": 0.08894331753253937,
      "rewards-max/forensics_tiou_reward": 0.833419680595398,
      "rewards/forensics_sinkhorn_reward": 0.6112610399723053,
      "rewards/forensics_strict_edge_reward": 0.004216995555907488,
      "rewards/forensics_tiou_reward": 0.3314660042524338,
      "step": 286
    },
    {
      "completion_length": 23.25,
      "epoch": 0.8817204301075269,
      "grad_norm": 4.992166042327881,
      "kl": 0.318359375,
      "learning_rate": 7.799999999999999e-07,
      "loss": -0.0172,
      "reward": 0.7110496163368225,
      "reward_std": 0.06130369380116463,
      "reward_std_all": 0.07458132412284613,
      "rewards-max/forensics_sinkhorn_reward": 0.9424140751361847,
      "rewards-max/forensics_strict_edge_reward": 0.1355073731392622,
      "rewards-max/forensics_tiou_reward": 0.8692644536495209,
      "rewards/forensics_sinkhorn_reward": 0.6344906091690063,
      "rewards/forensics_strict_edge_reward": 0.0063860692316666245,
      "rewards/forensics_tiou_reward": 0.3692677170038223,
      "step": 287
    },
    {
      "completion_length": 26.515625,
      "epoch": 0.8847926267281107,
      "grad_norm": 4.901494026184082,
      "kl": 0.3076171875,
      "learning_rate": 7.792307692307693e-07,
      "loss": 0.1259,
      "reward": 0.6792297959327698,
      "reward_std": 0.08585517853498459,
      "reward_std_all": 0.10186997801065445,
      "rewards-max/forensics_sinkhorn_reward": 0.9653803408145905,
      "rewards-max/forensics_strict_edge_reward": 0.3590621203184128,
      "rewards-max/forensics_tiou_reward": 0.9110493063926697,
      "rewards/forensics_sinkhorn_reward": 0.5846048295497894,
      "rewards/forensics_strict_edge_reward": 0.023666291497647762,
      "rewards/forensics_tiou_reward": 0.31023117899894714,
      "step": 288
    },
    {
      "completion_length": 23.65625,
      "epoch": 0.8878648233486943,
      "grad_norm": 5.0127458572387695,
      "kl": 0.501953125,
      "learning_rate": 7.784615384615384e-07,
      "loss": 0.152,
      "reward": 0.71257284283638,
      "reward_std": 0.0769997090101242,
      "reward_std_all": 0.08337969332933426,
      "rewards-max/forensics_sinkhorn_reward": 0.9553621113300323,
      "rewards-max/forensics_strict_edge_reward": 0.3429119437932968,
      "rewards-max/forensics_tiou_reward": 0.9007878005504608,
      "rewards/forensics_sinkhorn_reward": 0.619607001543045,
      "rewards/forensics_strict_edge_reward": 0.014703911903779954,
      "rewards/forensics_tiou_reward": 0.3327748477458954,
      "step": 289
    },
    {
      "completion_length": 30.640625,
      "epoch": 0.890937019969278,
      "grad_norm": 3.701030731201172,
      "kl": 0.2841796875,
      "learning_rate": 7.776923076923077e-07,
      "loss": 0.0735,
      "reward": 0.7145346701145172,
      "reward_std": 0.06560999155044556,
      "reward_std_all": 0.08086594566702843,
      "rewards-max/forensics_sinkhorn_reward": 0.9595110714435577,
      "rewards-max/forensics_strict_edge_reward": 0.18310516932979226,
      "rewards-max/forensics_tiou_reward": 0.9053316414356232,
      "rewards/forensics_sinkhorn_reward": 0.609708696603775,
      "rewards/forensics_strict_edge_reward": 0.010596531181363389,
      "rewards/forensics_tiou_reward": 0.3889962434768677,
      "step": 290
    },
    {
      "completion_length": 25.578125,
      "epoch": 0.8940092165898618,
      "grad_norm": 4.757227897644043,
      "kl": 0.2841796875,
      "learning_rate": 7.769230769230769e-07,
      "loss": 0.1362,
      "reward": 0.7458843886852264,
      "reward_std": 0.052496738731861115,
      "reward_std_all": 0.05437737703323364,
      "rewards-max/forensics_sinkhorn_reward": 0.9821520745754242,
      "rewards-max/forensics_strict_edge_reward": 0.2955048084259033,
      "rewards-max/forensics_tiou_reward": 0.9577169716358185,
      "rewards/forensics_sinkhorn_reward": 0.6615010499954224,
      "rewards/forensics_strict_edge_reward": 0.02052654977887869,
      "rewards/forensics_tiou_reward": 0.4219054877758026,
      "step": 291
    },
    {
      "completion_length": 24.296875,
      "epoch": 0.8970814132104454,
      "grad_norm": 5.3576178550720215,
      "kl": 0.25048828125,
      "learning_rate": 7.761538461538461e-07,
      "loss": 0.0872,
      "reward": 0.7080753147602081,
      "reward_std": 0.0923723429441452,
      "reward_std_all": 0.09481393918395042,
      "rewards-max/forensics_sinkhorn_reward": 0.9725326299667358,
      "rewards-max/forensics_strict_edge_reward": 0.2322264015674591,
      "rewards-max/forensics_tiou_reward": 0.9367823004722595,
      "rewards/forensics_sinkhorn_reward": 0.589514821767807,
      "rewards/forensics_strict_edge_reward": 0.01937736477702856,
      "rewards/forensics_tiou_reward": 0.346298485994339,
      "step": 292
    },
    {
      "completion_length": 26.984375,
      "epoch": 0.9001536098310292,
      "grad_norm": 3.3342716693878174,
      "kl": 0.3310546875,
      "learning_rate": 7.753846153846153e-07,
      "loss": 0.17,
      "reward": 0.748090535402298,
      "reward_std": 0.0644543468952179,
      "reward_std_all": 0.06263652071356773,
      "rewards-max/forensics_sinkhorn_reward": 0.962280809879303,
      "rewards-max/forensics_strict_edge_reward": 0.02310904348269105,
      "rewards-max/forensics_tiou_reward": 0.8871198296546936,
      "rewards/forensics_sinkhorn_reward": 0.6451409459114075,
      "rewards/forensics_strict_edge_reward": 0.002175599045585841,
      "rewards/forensics_tiou_reward": 0.42589494585990906,
      "step": 293
    },
    {
      "completion_length": 30.984375,
      "epoch": 0.9032258064516129,
      "grad_norm": 4.416475296020508,
      "kl": 0.326171875,
      "learning_rate": 7.746153846153845e-07,
      "loss": -0.0022,
      "reward": 0.6586887240409851,
      "reward_std": 0.09139754623174667,
      "reward_std_all": 0.11628103628754616,
      "rewards-max/forensics_sinkhorn_reward": 0.9427001774311066,
      "rewards-max/forensics_strict_edge_reward": 0.07847883179783821,
      "rewards-max/forensics_tiou_reward": 0.8603664040565491,
      "rewards/forensics_sinkhorn_reward": 0.5091454982757568,
      "rewards/forensics_strict_edge_reward": 0.006111258873715997,
      "rewards/forensics_tiou_reward": 0.34496453404426575,
      "step": 294
    },
    {
      "completion_length": 25.125,
      "epoch": 0.9062980030721967,
      "grad_norm": 4.072754383087158,
      "kl": 0.2421875,
      "learning_rate": 7.738461538461539e-07,
      "loss": 0.0768,
      "reward": 0.744041919708252,
      "reward_std": 0.06740165874361992,
      "reward_std_all": 0.06523178517818451,
      "rewards-max/forensics_sinkhorn_reward": 0.976751983165741,
      "rewards-max/forensics_strict_edge_reward": 0.355279803276062,
      "rewards-max/forensics_tiou_reward": 0.9425927400588989,
      "rewards/forensics_sinkhorn_reward": 0.6750190258026123,
      "rewards/forensics_strict_edge_reward": 0.026194365695118904,
      "rewards/forensics_tiou_reward": 0.4566899389028549,
      "step": 295
    },
    {
      "completion_length": 26.953125,
      "epoch": 0.9093701996927803,
      "grad_norm": 3.549306869506836,
      "kl": 0.2587890625,
      "learning_rate": 7.730769230769231e-07,
      "loss": 0.1658,
      "reward": 0.701636016368866,
      "reward_std": 0.049377137795090675,
      "reward_std_all": 0.07588587142527103,
      "rewards-max/forensics_sinkhorn_reward": 0.9773352146148682,
      "rewards-max/forensics_strict_edge_reward": 0.3897012807428837,
      "rewards-max/forensics_tiou_reward": 0.9375050365924835,
      "rewards/forensics_sinkhorn_reward": 0.5854174196720123,
      "rewards/forensics_strict_edge_reward": 0.03424562024883926,
      "rewards/forensics_tiou_reward": 0.36783935129642487,
      "step": 296
    },
    {
      "completion_length": 25.671875,
      "epoch": 0.9124423963133641,
      "grad_norm": 7.148793697357178,
      "kl": 0.3115234375,
      "learning_rate": 7.723076923076923e-07,
      "loss": 0.1693,
      "reward": 0.6948562562465668,
      "reward_std": 0.1271141767501831,
      "reward_std_all": 0.12158352136611938,
      "rewards-max/forensics_sinkhorn_reward": 0.9791455268859863,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9516685605049133,
      "rewards/forensics_sinkhorn_reward": 0.576661229133606,
      "rewards/forensics_strict_edge_reward": 0.03192781563848257,
      "rewards/forensics_tiou_reward": 0.36685122549533844,
      "step": 297
    },
    {
      "completion_length": 25.765625,
      "epoch": 0.9155145929339478,
      "grad_norm": 5.484372615814209,
      "kl": 0.25927734375,
      "learning_rate": 7.715384615384615e-07,
      "loss": 0.0692,
      "reward": 0.731858104467392,
      "reward_std": 0.07484684884548187,
      "reward_std_all": 0.0746002271771431,
      "rewards-max/forensics_sinkhorn_reward": 0.982432097196579,
      "rewards-max/forensics_strict_edge_reward": 0.36063383519649506,
      "rewards-max/forensics_tiou_reward": 0.9601791501045227,
      "rewards/forensics_sinkhorn_reward": 0.6448026299476624,
      "rewards/forensics_strict_edge_reward": 0.02118685469031334,
      "rewards/forensics_tiou_reward": 0.38265088200569153,
      "step": 298
    },
    {
      "completion_length": 26.375,
      "epoch": 0.9185867895545314,
      "grad_norm": 5.741014003753662,
      "kl": 0.26611328125,
      "learning_rate": 7.707692307692307e-07,
      "loss": 0.0011,
      "reward": 0.7061553001403809,
      "reward_std": 0.07059621810913086,
      "reward_std_all": 0.0851699598133564,
      "rewards-max/forensics_sinkhorn_reward": 0.9713658392429352,
      "rewards-max/forensics_strict_edge_reward": 0.1950355195440352,
      "rewards-max/forensics_tiou_reward": 0.9332939982414246,
      "rewards/forensics_sinkhorn_reward": 0.6338282525539398,
      "rewards/forensics_strict_edge_reward": 0.010037013184046373,
      "rewards/forensics_tiou_reward": 0.44635428488254547,
      "step": 299
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.9216589861751152,
      "grad_norm": 5.6254191398620605,
      "kl": 0.3232421875,
      "learning_rate": 7.699999999999999e-07,
      "loss": 0.1131,
      "reward": 0.7326200008392334,
      "reward_std": 0.06473579071462154,
      "reward_std_all": 0.06837606243789196,
      "rewards-max/forensics_sinkhorn_reward": 0.9622634649276733,
      "rewards-max/forensics_strict_edge_reward": 0.28044702112674713,
      "rewards-max/forensics_tiou_reward": 0.9105583131313324,
      "rewards/forensics_sinkhorn_reward": 0.6693142950534821,
      "rewards/forensics_strict_edge_reward": 0.017177733592689037,
      "rewards/forensics_tiou_reward": 0.38881339132785797,
      "step": 300
    },
    {
      "completion_length": 23.703125,
      "epoch": 0.9247311827956989,
      "grad_norm": 4.713756084442139,
      "kl": 0.3017578125,
      "learning_rate": 7.692307692307693e-07,
      "loss": 0.0471,
      "reward": 0.7830922901630402,
      "reward_std": 0.017580422572791576,
      "reward_std_all": 0.022136639803647995,
      "rewards-max/forensics_sinkhorn_reward": 0.9893299639225006,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.980619490146637,
      "rewards/forensics_sinkhorn_reward": 0.7334621250629425,
      "rewards/forensics_strict_edge_reward": 0.05750100500881672,
      "rewards/forensics_tiou_reward": 0.4734223335981369,
      "step": 301
    },
    {
      "completion_length": 22.390625,
      "epoch": 0.9278033794162827,
      "grad_norm": 5.340027332305908,
      "kl": 0.22509765625,
      "learning_rate": 7.684615384615384e-07,
      "loss": 0.1324,
      "reward": 0.722709059715271,
      "reward_std": 0.09471708908677101,
      "reward_std_all": 0.09210639260709286,
      "rewards-max/forensics_sinkhorn_reward": 0.9634408950805664,
      "rewards-max/forensics_strict_edge_reward": 0.43877362087368965,
      "rewards-max/forensics_tiou_reward": 0.9079581797122955,
      "rewards/forensics_sinkhorn_reward": 0.681608647108078,
      "rewards/forensics_strict_edge_reward": 0.02528873924165964,
      "rewards/forensics_tiou_reward": 0.4162342846393585,
      "step": 302
    },
    {
      "completion_length": 24.09375,
      "epoch": 0.9308755760368663,
      "grad_norm": 4.383544445037842,
      "kl": 0.271484375,
      "learning_rate": 7.676923076923077e-07,
      "loss": 0.06,
      "reward": 0.7634690701961517,
      "reward_std": 0.032656434923410416,
      "reward_std_all": 0.038855550810694695,
      "rewards-max/forensics_sinkhorn_reward": 0.9765589237213135,
      "rewards-max/forensics_strict_edge_reward": 0.08226747810840607,
      "rewards-max/forensics_tiou_reward": 0.9415601789951324,
      "rewards/forensics_sinkhorn_reward": 0.6868748366832733,
      "rewards/forensics_strict_edge_reward": 0.008258746471256018,
      "rewards/forensics_tiou_reward": 0.4455902725458145,
      "step": 303
    },
    {
      "completion_length": 21.5,
      "epoch": 0.9339477726574501,
      "grad_norm": 5.343082904815674,
      "kl": 0.3818359375,
      "learning_rate": 7.669230769230769e-07,
      "loss": -0.0026,
      "reward": 0.7510720789432526,
      "reward_std": 0.06866856664419174,
      "reward_std_all": 0.06516973488032818,
      "rewards-max/forensics_sinkhorn_reward": 0.9814799427986145,
      "rewards-max/forensics_strict_edge_reward": 0.46891139447689056,
      "rewards-max/forensics_tiou_reward": 0.953219473361969,
      "rewards/forensics_sinkhorn_reward": 0.724892795085907,
      "rewards/forensics_strict_edge_reward": 0.04622722044587135,
      "rewards/forensics_tiou_reward": 0.48915746808052063,
      "step": 304
    },
    {
      "completion_length": 21.8125,
      "epoch": 0.9370199692780338,
      "grad_norm": 5.70215368270874,
      "kl": 0.3486328125,
      "learning_rate": 7.661538461538461e-07,
      "loss": 0.065,
      "reward": 0.7119764089584351,
      "reward_std": 0.07741862628608942,
      "reward_std_all": 0.08235625550150871,
      "rewards-max/forensics_sinkhorn_reward": 0.977140486240387,
      "rewards-max/forensics_strict_edge_reward": 0.4011317938566208,
      "rewards-max/forensics_tiou_reward": 0.9457796514034271,
      "rewards/forensics_sinkhorn_reward": 0.6630585491657257,
      "rewards/forensics_strict_edge_reward": 0.02002935577183962,
      "rewards/forensics_tiou_reward": 0.3916662931442261,
      "step": 305
    },
    {
      "completion_length": 23.890625,
      "epoch": 0.9400921658986175,
      "grad_norm": 6.29269552230835,
      "kl": 0.296875,
      "learning_rate": 7.653846153846153e-07,
      "loss": 0.1278,
      "reward": 0.7250014245510101,
      "reward_std": 0.05795764736831188,
      "reward_std_all": 0.06858319230377674,
      "rewards-max/forensics_sinkhorn_reward": 0.9527399837970734,
      "rewards-max/forensics_strict_edge_reward": 0.11205768212676048,
      "rewards-max/forensics_tiou_reward": 0.8783151209354401,
      "rewards/forensics_sinkhorn_reward": 0.6232127249240875,
      "rewards/forensics_strict_edge_reward": 0.005796311190351844,
      "rewards/forensics_tiou_reward": 0.3610943555831909,
      "step": 306
    },
    {
      "completion_length": 23.203125,
      "epoch": 0.9431643625192012,
      "grad_norm": 6.634469509124756,
      "kl": 0.25,
      "learning_rate": 7.646153846153845e-07,
      "loss": 0.1094,
      "reward": 0.7507896423339844,
      "reward_std": 0.037088717333972454,
      "reward_std_all": 0.04767800681293011,
      "rewards-max/forensics_sinkhorn_reward": 0.9513516426086426,
      "rewards-max/forensics_strict_edge_reward": 0.26522907335311174,
      "rewards-max/forensics_tiou_reward": 0.853004664182663,
      "rewards/forensics_sinkhorn_reward": 0.6775113046169281,
      "rewards/forensics_strict_edge_reward": 0.02287769503891468,
      "rewards/forensics_tiou_reward": 0.39425618946552277,
      "step": 307
    },
    {
      "completion_length": 25.53125,
      "epoch": 0.946236559139785,
      "grad_norm": 4.776953220367432,
      "kl": 0.3740234375,
      "learning_rate": 7.638461538461539e-07,
      "loss": 0.1297,
      "reward": 0.6889554262161255,
      "reward_std": 0.03821714222431183,
      "reward_std_all": 0.07468636333942413,
      "rewards-max/forensics_sinkhorn_reward": 0.9682039022445679,
      "rewards-max/forensics_strict_edge_reward": 0.13335512240882963,
      "rewards-max/forensics_tiou_reward": 0.9041194319725037,
      "rewards/forensics_sinkhorn_reward": 0.5829698145389557,
      "rewards/forensics_strict_edge_reward": 0.010530665807891637,
      "rewards/forensics_tiou_reward": 0.3196098655462265,
      "step": 308
    },
    {
      "completion_length": 23.578125,
      "epoch": 0.9493087557603687,
      "grad_norm": 5.6136884689331055,
      "kl": 0.296875,
      "learning_rate": 7.630769230769231e-07,
      "loss": 0.049,
      "reward": 0.7472631335258484,
      "reward_std": 0.04971313662827015,
      "reward_std_all": 0.05475369282066822,
      "rewards-max/forensics_sinkhorn_reward": 0.9550102949142456,
      "rewards-max/forensics_strict_edge_reward": 0.34765754640102386,
      "rewards-max/forensics_tiou_reward": 0.8967837393283844,
      "rewards/forensics_sinkhorn_reward": 0.6757359206676483,
      "rewards/forensics_strict_edge_reward": 0.01580230053514242,
      "rewards/forensics_tiou_reward": 0.4100499451160431,
      "step": 309
    },
    {
      "completion_length": 24.59375,
      "epoch": 0.9523809523809523,
      "grad_norm": 5.188948631286621,
      "kl": 0.3359375,
      "learning_rate": 7.623076923076923e-07,
      "loss": 0.0719,
      "reward": 0.7624137997627258,
      "reward_std": 0.044197939336299896,
      "reward_std_all": 0.04463505744934082,
      "rewards-max/forensics_sinkhorn_reward": 0.9747107326984406,
      "rewards-max/forensics_strict_edge_reward": 0.4544278457760811,
      "rewards-max/forensics_tiou_reward": 0.9446885883808136,
      "rewards/forensics_sinkhorn_reward": 0.6945573091506958,
      "rewards/forensics_strict_edge_reward": 0.028446555137634277,
      "rewards/forensics_tiou_reward": 0.44090262055397034,
      "step": 310
    },
    {
      "completion_length": 23.484375,
      "epoch": 0.9554531490015361,
      "grad_norm": 4.841995716094971,
      "kl": 0.2900390625,
      "learning_rate": 7.615384615384615e-07,
      "loss": 0.0461,
      "reward": 0.6904614865779877,
      "reward_std": 0.08131346851587296,
      "reward_std_all": 0.09474927559494972,
      "rewards-max/forensics_sinkhorn_reward": 0.9124986529350281,
      "rewards-max/forensics_strict_edge_reward": 0.11163530126214027,
      "rewards-max/forensics_tiou_reward": 0.8003610372543335,
      "rewards/forensics_sinkhorn_reward": 0.5898636281490326,
      "rewards/forensics_strict_edge_reward": 0.005496127298101783,
      "rewards/forensics_tiou_reward": 0.30351030826568604,
      "step": 311
    },
    {
      "completion_length": 25.4375,
      "epoch": 0.9585253456221198,
      "grad_norm": 6.090603828430176,
      "kl": 0.369140625,
      "learning_rate": 7.607692307692307e-07,
      "loss": 0.201,
      "reward": 0.7550218999385834,
      "reward_std": 0.05228547379374504,
      "reward_std_all": 0.05198209919035435,
      "rewards-max/forensics_sinkhorn_reward": 0.9173561334609985,
      "rewards-max/forensics_strict_edge_reward": 0.2646277844905853,
      "rewards-max/forensics_tiou_reward": 0.808405876159668,
      "rewards/forensics_sinkhorn_reward": 0.6561045348644257,
      "rewards/forensics_strict_edge_reward": 0.015317996498197317,
      "rewards/forensics_tiou_reward": 0.3959325850009918,
      "step": 312
    },
    {
      "completion_length": 23.015625,
      "epoch": 0.9615975422427036,
      "grad_norm": 6.367893695831299,
      "kl": 0.2919921875,
      "learning_rate": 7.599999999999999e-07,
      "loss": -0.0375,
      "reward": 0.696325957775116,
      "reward_std": 0.08453569374978542,
      "reward_std_all": 0.09193956106901169,
      "rewards-max/forensics_sinkhorn_reward": 0.9471707046031952,
      "rewards-max/forensics_strict_edge_reward": 0.1824184108991176,
      "rewards-max/forensics_tiou_reward": 0.8481729328632355,
      "rewards/forensics_sinkhorn_reward": 0.6069150567054749,
      "rewards/forensics_strict_edge_reward": 0.02186583523871377,
      "rewards/forensics_tiou_reward": 0.3598766475915909,
      "step": 313
    },
    {
      "completion_length": 21.921875,
      "epoch": 0.9646697388632872,
      "grad_norm": 5.522693634033203,
      "kl": 0.375,
      "learning_rate": 7.592307692307692e-07,
      "loss": 0.1444,
      "reward": 0.7382170259952545,
      "reward_std": 0.06688813865184784,
      "reward_std_all": 0.0695832371711731,
      "rewards-max/forensics_sinkhorn_reward": 0.9624743461608887,
      "rewards-max/forensics_strict_edge_reward": 0.26738379895687103,
      "rewards-max/forensics_tiou_reward": 0.911062628030777,
      "rewards/forensics_sinkhorn_reward": 0.6681926846504211,
      "rewards/forensics_strict_edge_reward": 0.015447293873876333,
      "rewards/forensics_tiou_reward": 0.3874537795782089,
      "step": 314
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.967741935483871,
      "grad_norm": 5.00614595413208,
      "kl": 0.3447265625,
      "learning_rate": 7.584615384615384e-07,
      "loss": 0.2535,
      "reward": 0.6834148466587067,
      "reward_std": 0.06999539211392403,
      "reward_std_all": 0.09606542438268661,
      "rewards-max/forensics_sinkhorn_reward": 0.9472007751464844,
      "rewards-max/forensics_strict_edge_reward": 0.02390596689656377,
      "rewards-max/forensics_tiou_reward": 0.87624192237854,
      "rewards/forensics_sinkhorn_reward": 0.6045136451721191,
      "rewards/forensics_strict_edge_reward": 0.0017766065138857812,
      "rewards/forensics_tiou_reward": 0.3473287522792816,
      "step": 315
    },
    {
      "completion_length": 27.015625,
      "epoch": 0.9708141321044547,
      "grad_norm": 5.350013732910156,
      "kl": 0.36328125,
      "learning_rate": 7.576923076923077e-07,
      "loss": 0.1507,
      "reward": 0.7024968266487122,
      "reward_std": 0.10240915417671204,
      "reward_std_all": 0.10107357054948807,
      "rewards-max/forensics_sinkhorn_reward": 0.9392214715480804,
      "rewards-max/forensics_strict_edge_reward": 0.24046190467197448,
      "rewards-max/forensics_tiou_reward": 0.8496206402778625,
      "rewards/forensics_sinkhorn_reward": 0.5515646934509277,
      "rewards/forensics_strict_edge_reward": 0.023079138292814605,
      "rewards/forensics_tiou_reward": 0.4024845063686371,
      "step": 316
    },
    {
      "completion_length": 23.140625,
      "epoch": 0.9738863287250384,
      "grad_norm": 6.43074369430542,
      "kl": 0.314453125,
      "learning_rate": 7.569230769230769e-07,
      "loss": 0.0608,
      "reward": 0.7451949715614319,
      "reward_std": 0.07907018810510635,
      "reward_std_all": 0.07370072230696678,
      "rewards-max/forensics_sinkhorn_reward": 0.9730345010757446,
      "rewards-max/forensics_strict_edge_reward": 0.5274569988250732,
      "rewards-max/forensics_tiou_reward": 0.9415369033813477,
      "rewards/forensics_sinkhorn_reward": 0.6872986257076263,
      "rewards/forensics_strict_edge_reward": 0.03235382214188576,
      "rewards/forensics_tiou_reward": 0.4476949870586395,
      "step": 317
    },
    {
      "completion_length": 26.140625,
      "epoch": 0.9769585253456221,
      "grad_norm": 5.047083377838135,
      "kl": 0.396484375,
      "learning_rate": 7.561538461538461e-07,
      "loss": -0.0307,
      "reward": 0.747819572687149,
      "reward_std": 0.05675739608705044,
      "reward_std_all": 0.057459814473986626,
      "rewards-max/forensics_sinkhorn_reward": 0.9181763529777527,
      "rewards-max/forensics_strict_edge_reward": 0.08544253627769649,
      "rewards-max/forensics_tiou_reward": 0.79874587059021,
      "rewards/forensics_sinkhorn_reward": 0.6536891460418701,
      "rewards/forensics_strict_edge_reward": 0.004264141360181384,
      "rewards/forensics_tiou_reward": 0.3826508969068527,
      "step": 318
    },
    {
      "completion_length": 23.015625,
      "epoch": 0.9800307219662059,
      "grad_norm": 5.370683670043945,
      "kl": 0.2841796875,
      "learning_rate": 7.553846153846154e-07,
      "loss": 0.1553,
      "reward": 0.7042341232299805,
      "reward_std": 0.06031334213912487,
      "reward_std_all": 0.0815784465521574,
      "rewards-max/forensics_sinkhorn_reward": 0.9632158577442169,
      "rewards-max/forensics_strict_edge_reward": 0.4052793085575104,
      "rewards-max/forensics_tiou_reward": 0.923926830291748,
      "rewards/forensics_sinkhorn_reward": 0.6202033162117004,
      "rewards/forensics_strict_edge_reward": 0.02364042680710554,
      "rewards/forensics_tiou_reward": 0.3660489618778229,
      "step": 319
    },
    {
      "completion_length": 23.15625,
      "epoch": 0.9831029185867896,
      "grad_norm": 4.8148088455200195,
      "kl": 0.3798828125,
      "learning_rate": 7.546153846153845e-07,
      "loss": 0.1036,
      "reward": 0.7581039071083069,
      "reward_std": 0.060286276042461395,
      "reward_std_all": 0.05783028341829777,
      "rewards-max/forensics_sinkhorn_reward": 0.9555373787879944,
      "rewards-max/forensics_strict_edge_reward": 0.22393766790628433,
      "rewards-max/forensics_tiou_reward": 0.870750367641449,
      "rewards/forensics_sinkhorn_reward": 0.7137945592403412,
      "rewards/forensics_strict_edge_reward": 0.021005351096391678,
      "rewards/forensics_tiou_reward": 0.4718737453222275,
      "step": 320
    },
    {
      "completion_length": 25.140625,
      "epoch": 0.9861751152073732,
      "grad_norm": 4.561981678009033,
      "kl": 0.275390625,
      "learning_rate": 7.538461538461538e-07,
      "loss": 0.1698,
      "reward": 0.7347049713134766,
      "reward_std": 0.049342477694153786,
      "reward_std_all": 0.060416026040911674,
      "rewards-max/forensics_sinkhorn_reward": 0.9099955856800079,
      "rewards-max/forensics_strict_edge_reward": 0.25610521296039224,
      "rewards-max/forensics_tiou_reward": 0.7970456779003143,
      "rewards/forensics_sinkhorn_reward": 0.6193738877773285,
      "rewards/forensics_strict_edge_reward": 0.029981961473822594,
      "rewards/forensics_tiou_reward": 0.36443543434143066,
      "step": 321
    },
    {
      "completion_length": 25.53125,
      "epoch": 0.989247311827957,
      "grad_norm": 4.95803689956665,
      "kl": 0.4306640625,
      "learning_rate": 7.530769230769231e-07,
      "loss": 0.0079,
      "reward": 0.7314677834510803,
      "reward_std": 0.05704781785607338,
      "reward_std_all": 0.06507156789302826,
      "rewards-max/forensics_sinkhorn_reward": 0.9487815499305725,
      "rewards-max/forensics_strict_edge_reward": 0.28422586619853973,
      "rewards-max/forensics_tiou_reward": 0.8774893283843994,
      "rewards/forensics_sinkhorn_reward": 0.6143768429756165,
      "rewards/forensics_strict_edge_reward": 0.024993329774588346,
      "rewards/forensics_tiou_reward": 0.3579961806535721,
      "step": 322
    },
    {
      "completion_length": 26.421875,
      "epoch": 0.9923195084485407,
      "grad_norm": 4.968814373016357,
      "kl": 0.337890625,
      "learning_rate": 7.523076923076923e-07,
      "loss": 0.1113,
      "reward": 0.6965408623218536,
      "reward_std": 0.08358418196439743,
      "reward_std_all": 0.09099721536040306,
      "rewards-max/forensics_sinkhorn_reward": 0.9382732510566711,
      "rewards-max/forensics_strict_edge_reward": 0.22613869735505432,
      "rewards-max/forensics_tiou_reward": 0.8309124708175659,
      "rewards/forensics_sinkhorn_reward": 0.574590802192688,
      "rewards/forensics_strict_edge_reward": 0.012936044680827763,
      "rewards/forensics_tiou_reward": 0.32927605509757996,
      "step": 323
    },
    {
      "completion_length": 25.46875,
      "epoch": 0.9953917050691244,
      "grad_norm": 4.9866790771484375,
      "kl": 0.287109375,
      "learning_rate": 7.515384615384615e-07,
      "loss": 0.102,
      "reward": 0.7553023397922516,
      "reward_std": 0.0373892355710268,
      "reward_std_all": 0.04542150627821684,
      "rewards-max/forensics_sinkhorn_reward": 0.9698121845722198,
      "rewards-max/forensics_strict_edge_reward": 0.16485833376646042,
      "rewards-max/forensics_tiou_reward": 0.9260726869106293,
      "rewards/forensics_sinkhorn_reward": 0.6042277812957764,
      "rewards/forensics_strict_edge_reward": 0.008106150664389133,
      "rewards/forensics_tiou_reward": 0.46237511932849884,
      "step": 324
    },
    {
      "completion_length": 28.796875,
      "epoch": 0.9984639016897081,
      "grad_norm": 6.032171249389648,
      "kl": 0.4150390625,
      "learning_rate": 7.507692307692307e-07,
      "loss": 0.1237,
      "reward": 0.746825248003006,
      "reward_std": 0.07903770357370377,
      "reward_std_all": 0.07331644371151924,
      "rewards-max/forensics_sinkhorn_reward": 0.9719253480434418,
      "rewards-max/forensics_strict_edge_reward": 0.31592947617173195,
      "rewards-max/forensics_tiou_reward": 0.9331642091274261,
      "rewards/forensics_sinkhorn_reward": 0.6593940556049347,
      "rewards/forensics_strict_edge_reward": 0.02645696420222521,
      "rewards/forensics_tiou_reward": 0.42927588522434235,
      "step": 325
    },
    {
      "completion_length": 19.0,
      "epoch": 1.0,
      "grad_norm": 3.6982741355895996,
      "kl": 0.2421875,
      "learning_rate": 7.5e-07,
      "loss": 0.0464,
      "reward": 0.7716385722160339,
      "reward_std": 0.03111307881772518,
      "reward_std_all": 0.03417545184493065,
      "rewards-max/forensics_sinkhorn_reward": 0.9903830885887146,
      "rewards-max/forensics_strict_edge_reward": 0.16529889404773712,
      "rewards-max/forensics_tiou_reward": 0.978690505027771,
      "rewards/forensics_sinkhorn_reward": 0.734673798084259,
      "rewards/forensics_strict_edge_reward": 0.033059779554605484,
      "rewards/forensics_tiou_reward": 0.3819165527820587,
      "step": 326
    },
    {
      "completion_length": 27.921875,
      "epoch": 1.0030721966205838,
      "grad_norm": 6.053960800170898,
      "kl": 0.328125,
      "learning_rate": 7.492307692307692e-07,
      "loss": -0.0504,
      "reward": 0.7322807312011719,
      "reward_std": 0.0630243793129921,
      "reward_std_all": 0.06938368082046509,
      "rewards-max/forensics_sinkhorn_reward": 0.9712465405464172,
      "rewards-max/forensics_strict_edge_reward": 0.5067306756973267,
      "rewards-max/forensics_tiou_reward": 0.9311782121658325,
      "rewards/forensics_sinkhorn_reward": 0.6755574643611908,
      "rewards/forensics_strict_edge_reward": 0.050701506435871124,
      "rewards/forensics_tiou_reward": 0.4516947567462921,
      "step": 327
    },
    {
      "completion_length": 24.40625,
      "epoch": 1.0061443932411673,
      "grad_norm": 6.523731708526611,
      "kl": 0.3525390625,
      "learning_rate": 7.484615384615384e-07,
      "loss": 0.0702,
      "reward": 0.7566879391670227,
      "reward_std": 0.05633295141160488,
      "reward_std_all": 0.05445053614675999,
      "rewards-max/forensics_sinkhorn_reward": 0.9873247444629669,
      "rewards-max/forensics_strict_edge_reward": 0.2879621237516403,
      "rewards-max/forensics_tiou_reward": 0.970937967300415,
      "rewards/forensics_sinkhorn_reward": 0.6742902398109436,
      "rewards/forensics_strict_edge_reward": 0.028880962170660496,
      "rewards/forensics_tiou_reward": 0.46690621972084045,
      "step": 328
    },
    {
      "completion_length": 28.609375,
      "epoch": 1.0092165898617511,
      "grad_norm": 5.657609939575195,
      "kl": 0.5107421875,
      "learning_rate": 7.476923076923077e-07,
      "loss": 0.2063,
      "reward": 0.7197582125663757,
      "reward_std": 0.056040581315755844,
      "reward_std_all": 0.0690949335694313,
      "rewards-max/forensics_sinkhorn_reward": 0.978311151266098,
      "rewards-max/forensics_strict_edge_reward": 0.22341300919651985,
      "rewards-max/forensics_tiou_reward": 0.9449766874313354,
      "rewards/forensics_sinkhorn_reward": 0.6263788938522339,
      "rewards/forensics_strict_edge_reward": 0.021973980590701103,
      "rewards/forensics_tiou_reward": 0.4360915422439575,
      "step": 329
    },
    {
      "completion_length": 26.203125,
      "epoch": 1.012288786482335,
      "grad_norm": 4.906993865966797,
      "kl": 0.26416015625,
      "learning_rate": 7.469230769230769e-07,
      "loss": -0.0375,
      "reward": 0.769832044839859,
      "reward_std": 0.030513545963913202,
      "reward_std_all": 0.03522611781954765,
      "rewards-max/forensics_sinkhorn_reward": 0.9833421409130096,
      "rewards-max/forensics_strict_edge_reward": 0.5244297683238983,
      "rewards-max/forensics_tiou_reward": 0.9614847302436829,
      "rewards/forensics_sinkhorn_reward": 0.6779633462429047,
      "rewards/forensics_strict_edge_reward": 0.032521747052669525,
      "rewards/forensics_tiou_reward": 0.5150730311870575,
      "step": 330
    },
    {
      "completion_length": 26.3125,
      "epoch": 1.0153609831029187,
      "grad_norm": 5.4476141929626465,
      "kl": 0.3193359375,
      "learning_rate": 7.461538461538461e-07,
      "loss": 0.0438,
      "reward": 0.736178457736969,
      "reward_std": 0.05654860474169254,
      "reward_std_all": 0.062176480889320374,
      "rewards-max/forensics_sinkhorn_reward": 0.97391077876091,
      "rewards-max/forensics_strict_edge_reward": 0.2371476050466299,
      "rewards-max/forensics_tiou_reward": 0.9307741224765778,
      "rewards/forensics_sinkhorn_reward": 0.667151004076004,
      "rewards/forensics_strict_edge_reward": 0.023456374532543123,
      "rewards/forensics_tiou_reward": 0.4407843053340912,
      "step": 331
    },
    {
      "completion_length": 28.65625,
      "epoch": 1.0184331797235022,
      "grad_norm": 4.880171298980713,
      "kl": 0.2685546875,
      "learning_rate": 7.453846153846154e-07,
      "loss": 0.1419,
      "reward": 0.6785339415073395,
      "reward_std": 0.09455898404121399,
      "reward_std_all": 0.11175209283828735,
      "rewards-max/forensics_sinkhorn_reward": 0.9543092250823975,
      "rewards-max/forensics_strict_edge_reward": 0.09461709490278736,
      "rewards-max/forensics_tiou_reward": 0.8957318663597107,
      "rewards/forensics_sinkhorn_reward": 0.5454255938529968,
      "rewards/forensics_strict_edge_reward": 0.003993974278273527,
      "rewards/forensics_tiou_reward": 0.32422003149986267,
      "step": 332
    },
    {
      "completion_length": 23.65625,
      "epoch": 1.021505376344086,
      "grad_norm": 6.187910079956055,
      "kl": 0.6005859375,
      "learning_rate": 7.446153846153845e-07,
      "loss": 0.1221,
      "reward": 0.7549003064632416,
      "reward_std": 0.06322949193418026,
      "reward_std_all": 0.060041225515305996,
      "rewards-max/forensics_sinkhorn_reward": 0.9744700491428375,
      "rewards-max/forensics_strict_edge_reward": 0.6575055122375488,
      "rewards-max/forensics_tiou_reward": 0.9459608197212219,
      "rewards/forensics_sinkhorn_reward": 0.6768729090690613,
      "rewards/forensics_strict_edge_reward": 0.0509930532425642,
      "rewards/forensics_tiou_reward": 0.4355808347463608,
      "step": 333
    },
    {
      "completion_length": 27.421875,
      "epoch": 1.0245775729646698,
      "grad_norm": 4.486147880554199,
      "kl": 0.31640625,
      "learning_rate": 7.438461538461538e-07,
      "loss": 0.1824,
      "reward": 0.718163788318634,
      "reward_std": 0.07065749727189541,
      "reward_std_all": 0.07545038871467113,
      "rewards-max/forensics_sinkhorn_reward": 0.9716711044311523,
      "rewards-max/forensics_strict_edge_reward": 0.6071805953979492,
      "rewards-max/forensics_tiou_reward": 0.9397375285625458,
      "rewards/forensics_sinkhorn_reward": 0.6109645962715149,
      "rewards/forensics_strict_edge_reward": 0.04028963577002287,
      "rewards/forensics_tiou_reward": 0.4069874584674835,
      "step": 334
    },
    {
      "completion_length": 26.328125,
      "epoch": 1.0276497695852536,
      "grad_norm": 7.008718490600586,
      "kl": 0.6865234375,
      "learning_rate": 7.43076923076923e-07,
      "loss": 0.0574,
      "reward": 0.6745060384273529,
      "reward_std": 0.09076014906167984,
      "reward_std_all": 0.11007561907172203,
      "rewards-max/forensics_sinkhorn_reward": 0.9715362787246704,
      "rewards-max/forensics_strict_edge_reward": 0.28076549246907234,
      "rewards-max/forensics_tiou_reward": 0.9410808980464935,
      "rewards/forensics_sinkhorn_reward": 0.5876816809177399,
      "rewards/forensics_strict_edge_reward": 0.017909941729158163,
      "rewards/forensics_tiou_reward": 0.36089302599430084,
      "step": 335
    },
    {
      "completion_length": 27.1875,
      "epoch": 1.0307219662058371,
      "grad_norm": 5.92833948135376,
      "kl": 0.3916015625,
      "learning_rate": 7.423076923076923e-07,
      "loss": 0.1617,
      "reward": 0.7365908920764923,
      "reward_std": 0.07049261778593063,
      "reward_std_all": 0.06921960785984993,
      "rewards-max/forensics_sinkhorn_reward": 0.9900942444801331,
      "rewards-max/forensics_strict_edge_reward": 0.5776711404323578,
      "rewards-max/forensics_tiou_reward": 0.9785096049308777,
      "rewards/forensics_sinkhorn_reward": 0.6686213910579681,
      "rewards/forensics_strict_edge_reward": 0.0365391131490469,
      "rewards/forensics_tiou_reward": 0.4698401093482971,
      "step": 336
    },
    {
      "completion_length": 21.609375,
      "epoch": 1.033794162826421,
      "grad_norm": 5.672828197479248,
      "kl": 0.3212890625,
      "learning_rate": 7.415384615384616e-07,
      "loss": -0.0029,
      "reward": 0.7817280292510986,
      "reward_std": 0.02616690658032894,
      "reward_std_all": 0.02690250799059868,
      "rewards-max/forensics_sinkhorn_reward": 0.9667794704437256,
      "rewards-max/forensics_strict_edge_reward": 0.4661003053188324,
      "rewards-max/forensics_tiou_reward": 0.9238562881946564,
      "rewards/forensics_sinkhorn_reward": 0.7243885397911072,
      "rewards/forensics_strict_edge_reward": 0.042424144223332405,
      "rewards/forensics_tiou_reward": 0.4368223398923874,
      "step": 337
    },
    {
      "completion_length": 24.15625,
      "epoch": 1.0368663594470047,
      "grad_norm": 5.499756336212158,
      "kl": 0.3876953125,
      "learning_rate": 7.407692307692307e-07,
      "loss": 0.1193,
      "reward": 0.7361589670181274,
      "reward_std": 0.0939432680606842,
      "reward_std_all": 0.08670523762702942,
      "rewards-max/forensics_sinkhorn_reward": 0.9285554587841034,
      "rewards-max/forensics_strict_edge_reward": 0.012379245832562447,
      "rewards-max/forensics_tiou_reward": 0.7856303453445435,
      "rewards/forensics_sinkhorn_reward": 0.648642361164093,
      "rewards/forensics_strict_edge_reward": 0.0009659362258389592,
      "rewards/forensics_tiou_reward": 0.33490581810474396,
      "step": 338
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.0399385560675882,
      "grad_norm": 5.3579182624816895,
      "kl": 0.3173828125,
      "learning_rate": 7.4e-07,
      "loss": 0.0724,
      "reward": 0.7460521459579468,
      "reward_std": 0.06412536837160587,
      "reward_std_all": 0.06381513923406601,
      "rewards-max/forensics_sinkhorn_reward": 0.9832559525966644,
      "rewards-max/forensics_strict_edge_reward": 0.1731217112392187,
      "rewards-max/forensics_tiou_reward": 0.9474225640296936,
      "rewards/forensics_sinkhorn_reward": 0.6808442175388336,
      "rewards/forensics_strict_edge_reward": 0.019941541366279125,
      "rewards/forensics_tiou_reward": 0.4670376777648926,
      "step": 339
    },
    {
      "completion_length": 25.421875,
      "epoch": 1.043010752688172,
      "grad_norm": 3.6914358139038086,
      "kl": 0.3037109375,
      "learning_rate": 7.392307692307692e-07,
      "loss": 0.0658,
      "reward": 0.7336945235729218,
      "reward_std": 0.0698823630809784,
      "reward_std_all": 0.06987759470939636,
      "rewards-max/forensics_sinkhorn_reward": 0.977751761674881,
      "rewards-max/forensics_strict_edge_reward": 0.54204261302948,
      "rewards-max/forensics_tiou_reward": 0.9447542130947113,
      "rewards/forensics_sinkhorn_reward": 0.6722068190574646,
      "rewards/forensics_strict_edge_reward": 0.042304662987589836,
      "rewards/forensics_tiou_reward": 0.4379262179136276,
      "step": 340
    },
    {
      "completion_length": 23.15625,
      "epoch": 1.0460829493087558,
      "grad_norm": 4.525136470794678,
      "kl": 0.318359375,
      "learning_rate": 7.384615384615384e-07,
      "loss": 0.1409,
      "reward": 0.7395889461040497,
      "reward_std": 0.032728361897170544,
      "reward_std_all": 0.05002903938293457,
      "rewards-max/forensics_sinkhorn_reward": 0.9561438262462616,
      "rewards-max/forensics_strict_edge_reward": 0.14766891719773412,
      "rewards-max/forensics_tiou_reward": 0.9034851491451263,
      "rewards/forensics_sinkhorn_reward": 0.6296536028385162,
      "rewards/forensics_strict_edge_reward": 0.009929052408551797,
      "rewards/forensics_tiou_reward": 0.4093469977378845,
      "step": 341
    },
    {
      "completion_length": 24.609375,
      "epoch": 1.0491551459293396,
      "grad_norm": 4.483613967895508,
      "kl": 0.29296875,
      "learning_rate": 7.376923076923076e-07,
      "loss": 0.1216,
      "reward": 0.7373875379562378,
      "reward_std": 0.07834037020802498,
      "reward_std_all": 0.07535667717456818,
      "rewards-max/forensics_sinkhorn_reward": 0.9816523492336273,
      "rewards-max/forensics_strict_edge_reward": 0.48830981831997633,
      "rewards-max/forensics_tiou_reward": 0.952809751033783,
      "rewards/forensics_sinkhorn_reward": 0.6473222970962524,
      "rewards/forensics_strict_edge_reward": 0.020981411973480135,
      "rewards/forensics_tiou_reward": 0.42682453989982605,
      "step": 342
    },
    {
      "completion_length": 25.625,
      "epoch": 1.0522273425499231,
      "grad_norm": 5.196666240692139,
      "kl": 0.3515625,
      "learning_rate": 7.369230769230769e-07,
      "loss": 0.0541,
      "reward": 0.7500228583812714,
      "reward_std": 0.07268390618264675,
      "reward_std_all": 0.06818460673093796,
      "rewards-max/forensics_sinkhorn_reward": 0.9926057755947113,
      "rewards-max/forensics_strict_edge_reward": 0.7709200084209442,
      "rewards-max/forensics_tiou_reward": 0.9874565005302429,
      "rewards/forensics_sinkhorn_reward": 0.6462994515895844,
      "rewards/forensics_strict_edge_reward": 0.06844725087285042,
      "rewards/forensics_tiou_reward": 0.499808669090271,
      "step": 343
    },
    {
      "completion_length": 26.90625,
      "epoch": 1.055299539170507,
      "grad_norm": 4.201338291168213,
      "kl": 0.359375,
      "learning_rate": 7.361538461538461e-07,
      "loss": 0.2105,
      "reward": 0.7299250960350037,
      "reward_std": 0.05856454372406006,
      "reward_std_all": 0.06687572970986366,
      "rewards-max/forensics_sinkhorn_reward": 0.9538283050060272,
      "rewards-max/forensics_strict_edge_reward": 0.2616788297891617,
      "rewards-max/forensics_tiou_reward": 0.887940913438797,
      "rewards/forensics_sinkhorn_reward": 0.5907409489154816,
      "rewards/forensics_strict_edge_reward": 0.01193592231720686,
      "rewards/forensics_tiou_reward": 0.41165396571159363,
      "step": 344
    },
    {
      "completion_length": 24.328125,
      "epoch": 1.0583717357910907,
      "grad_norm": 5.369482040405273,
      "kl": 0.3134765625,
      "learning_rate": 7.353846153846154e-07,
      "loss": 0.0223,
      "reward": 0.7645653188228607,
      "reward_std": 0.033743432722985744,
      "reward_std_all": 0.03984854929149151,
      "rewards-max/forensics_sinkhorn_reward": 0.9845390319824219,
      "rewards-max/forensics_strict_edge_reward": 0.4908616244792938,
      "rewards-max/forensics_tiou_reward": 0.9642451405525208,
      "rewards/forensics_sinkhorn_reward": 0.7109436094760895,
      "rewards/forensics_strict_edge_reward": 0.027177225332707167,
      "rewards/forensics_tiou_reward": 0.48998142778873444,
      "step": 345
    },
    {
      "completion_length": 27.09375,
      "epoch": 1.0614439324116745,
      "grad_norm": 4.69118595123291,
      "kl": 0.3623046875,
      "learning_rate": 7.346153846153846e-07,
      "loss": 0.183,
      "reward": 0.7366916239261627,
      "reward_std": 0.04024070315063,
      "reward_std_all": 0.05656473897397518,
      "rewards-max/forensics_sinkhorn_reward": 0.972521185874939,
      "rewards-max/forensics_strict_edge_reward": 0.19274798035621643,
      "rewards-max/forensics_tiou_reward": 0.9337239861488342,
      "rewards/forensics_sinkhorn_reward": 0.6343013346195221,
      "rewards/forensics_strict_edge_reward": 0.016479130368679762,
      "rewards/forensics_tiou_reward": 0.44255736470222473,
      "step": 346
    },
    {
      "completion_length": 23.21875,
      "epoch": 1.064516129032258,
      "grad_norm": 4.7144551277160645,
      "kl": 0.267578125,
      "learning_rate": 7.338461538461538e-07,
      "loss": 0.1034,
      "reward": 0.7757697105407715,
      "reward_std": 0.020126574207097292,
      "reward_std_all": 0.025498709641397,
      "rewards-max/forensics_sinkhorn_reward": 0.9534589052200317,
      "rewards-max/forensics_strict_edge_reward": 0.3751392289996147,
      "rewards-max/forensics_tiou_reward": 0.8806508779525757,
      "rewards/forensics_sinkhorn_reward": 0.6875450313091278,
      "rewards/forensics_strict_edge_reward": 0.018726933281868696,
      "rewards/forensics_tiou_reward": 0.4299336224794388,
      "step": 347
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.0675883256528418,
      "grad_norm": 5.908445835113525,
      "kl": 0.3955078125,
      "learning_rate": 7.33076923076923e-07,
      "loss": 0.0368,
      "reward": 0.732391893863678,
      "reward_std": 0.03238851670175791,
      "reward_std_all": 0.054823772981762886,
      "rewards-max/forensics_sinkhorn_reward": 0.9802570939064026,
      "rewards-max/forensics_strict_edge_reward": 0.561550498008728,
      "rewards-max/forensics_tiou_reward": 0.9568520486354828,
      "rewards/forensics_sinkhorn_reward": 0.7057739496231079,
      "rewards/forensics_strict_edge_reward": 0.03728227689862251,
      "rewards/forensics_tiou_reward": 0.43121057748794556,
      "step": 348
    },
    {
      "completion_length": 24.546875,
      "epoch": 1.0706605222734256,
      "grad_norm": 4.812714099884033,
      "kl": 0.31640625,
      "learning_rate": 7.323076923076923e-07,
      "loss": 0.1306,
      "reward": 0.7438877820968628,
      "reward_std": 0.051853885874152184,
      "reward_std_all": 0.057019591331481934,
      "rewards-max/forensics_sinkhorn_reward": 0.978744775056839,
      "rewards-max/forensics_strict_edge_reward": 0.5055623948574066,
      "rewards-max/forensics_tiou_reward": 0.9525539875030518,
      "rewards/forensics_sinkhorn_reward": 0.6743994057178497,
      "rewards/forensics_strict_edge_reward": 0.03455452527850866,
      "rewards/forensics_tiou_reward": 0.4541558772325516,
      "step": 349
    },
    {
      "completion_length": 25.8125,
      "epoch": 1.0737327188940091,
      "grad_norm": 5.065425872802734,
      "kl": 0.279296875,
      "learning_rate": 7.315384615384616e-07,
      "loss": 0.0246,
      "reward": 0.7700782418251038,
      "reward_std": 0.041251176968216896,
      "reward_std_all": 0.040503544732928276,
      "rewards-max/forensics_sinkhorn_reward": 0.9769330322742462,
      "rewards-max/forensics_strict_edge_reward": 0.480892151594162,
      "rewards-max/forensics_tiou_reward": 0.9430415332317352,
      "rewards/forensics_sinkhorn_reward": 0.6761998236179352,
      "rewards/forensics_strict_edge_reward": 0.05638452619314194,
      "rewards/forensics_tiou_reward": 0.49566879868507385,
      "step": 350
    },
    {
      "completion_length": 26.359375,
      "epoch": 1.076804915514593,
      "grad_norm": 4.91592264175415,
      "kl": 0.3515625,
      "learning_rate": 7.307692307692307e-07,
      "loss": 0.1905,
      "reward": 0.6958819925785065,
      "reward_std": 0.08233917877078056,
      "reward_std_all": 0.09427381306886673,
      "rewards-max/forensics_sinkhorn_reward": 0.9642932415008545,
      "rewards-max/forensics_strict_edge_reward": 0.22761881723999977,
      "rewards-max/forensics_tiou_reward": 0.910120040178299,
      "rewards/forensics_sinkhorn_reward": 0.6149665117263794,
      "rewards/forensics_strict_edge_reward": 0.013045589439570904,
      "rewards/forensics_tiou_reward": 0.39176730811595917,
      "step": 351
    },
    {
      "completion_length": 25.6875,
      "epoch": 1.0798771121351767,
      "grad_norm": 5.41831111907959,
      "kl": 0.2939453125,
      "learning_rate": 7.3e-07,
      "loss": -0.0207,
      "reward": 0.7154322564601898,
      "reward_std": 0.07697588205337524,
      "reward_std_all": 0.0813494622707367,
      "rewards-max/forensics_sinkhorn_reward": 0.9614632427692413,
      "rewards-max/forensics_strict_edge_reward": 0.16686122864484787,
      "rewards-max/forensics_tiou_reward": 0.8967937231063843,
      "rewards/forensics_sinkhorn_reward": 0.6216698586940765,
      "rewards/forensics_strict_edge_reward": 0.013129981700330973,
      "rewards/forensics_tiou_reward": 0.4218800812959671,
      "step": 352
    },
    {
      "completion_length": 25.359375,
      "epoch": 1.0829493087557605,
      "grad_norm": 4.567417144775391,
      "kl": 0.3232421875,
      "learning_rate": 7.292307692307692e-07,
      "loss": 0.1643,
      "reward": 0.7374228537082672,
      "reward_std": 0.06492344662547112,
      "reward_std_all": 0.06861294060945511,
      "rewards-max/forensics_sinkhorn_reward": 0.9259472191333771,
      "rewards-max/forensics_strict_edge_reward": 0.12220471608452499,
      "rewards-max/forensics_tiou_reward": 0.8205493688583374,
      "rewards/forensics_sinkhorn_reward": 0.6570886373519897,
      "rewards/forensics_strict_edge_reward": 0.008410518465097994,
      "rewards/forensics_tiou_reward": 0.41685381531715393,
      "step": 353
    },
    {
      "completion_length": 24.0,
      "epoch": 1.086021505376344,
      "grad_norm": 6.109921455383301,
      "kl": 0.2890625,
      "learning_rate": 7.284615384615384e-07,
      "loss": 0.0109,
      "reward": 0.7406730353832245,
      "reward_std": 0.06713435240089893,
      "reward_std_all": 0.06681536138057709,
      "rewards-max/forensics_sinkhorn_reward": 0.8233279287815094,
      "rewards-max/forensics_strict_edge_reward": 0.1432810164988041,
      "rewards-max/forensics_tiou_reward": 0.5875130295753479,
      "rewards/forensics_sinkhorn_reward": 0.5956279635429382,
      "rewards/forensics_strict_edge_reward": 0.007144587114453316,
      "rewards/forensics_tiou_reward": 0.335319459438324,
      "step": 354
    },
    {
      "completion_length": 23.859375,
      "epoch": 1.0890937019969278,
      "grad_norm": 4.719030380249023,
      "kl": 0.361328125,
      "learning_rate": 7.276923076923076e-07,
      "loss": 0.0018,
      "reward": 0.7406118512153625,
      "reward_std": 0.0477613378316164,
      "reward_std_all": 0.054659707471728325,
      "rewards-max/forensics_sinkhorn_reward": 0.9896613955497742,
      "rewards-max/forensics_strict_edge_reward": 0.2972189001739025,
      "rewards-max/forensics_tiou_reward": 0.9763225317001343,
      "rewards/forensics_sinkhorn_reward": 0.7023066878318787,
      "rewards/forensics_strict_edge_reward": 0.02159753767773509,
      "rewards/forensics_tiou_reward": 0.4299383610486984,
      "step": 355
    },
    {
      "completion_length": 26.28125,
      "epoch": 1.0921658986175116,
      "grad_norm": 4.928651332855225,
      "kl": 0.30859375,
      "learning_rate": 7.269230769230769e-07,
      "loss": 0.2511,
      "reward": 0.7095547318458557,
      "reward_std": 0.06180699169635773,
      "reward_std_all": 0.07765555195510387,
      "rewards-max/forensics_sinkhorn_reward": 0.9179870784282684,
      "rewards-max/forensics_strict_edge_reward": 0.08807932212948799,
      "rewards-max/forensics_tiou_reward": 0.7985296249389648,
      "rewards/forensics_sinkhorn_reward": 0.5843871831893921,
      "rewards/forensics_strict_edge_reward": 0.005056128138676286,
      "rewards/forensics_tiou_reward": 0.35601507127285004,
      "step": 356
    },
    {
      "completion_length": 25.75,
      "epoch": 1.0952380952380953,
      "grad_norm": 4.874016284942627,
      "kl": 0.361328125,
      "learning_rate": 7.261538461538461e-07,
      "loss": 0.0984,
      "reward": 0.7816912829875946,
      "reward_std": 0.034680393524467945,
      "reward_std_all": 0.03294038772583008,
      "rewards-max/forensics_sinkhorn_reward": 0.9785676300525665,
      "rewards-max/forensics_strict_edge_reward": 0.28275663033127785,
      "rewards-max/forensics_tiou_reward": 0.9489217698574066,
      "rewards/forensics_sinkhorn_reward": 0.7301738262176514,
      "rewards/forensics_strict_edge_reward": 0.022221714025363326,
      "rewards/forensics_tiou_reward": 0.5198104679584503,
      "step": 357
    },
    {
      "completion_length": 25.640625,
      "epoch": 1.098310291858679,
      "grad_norm": 5.10036563873291,
      "kl": 0.2587890625,
      "learning_rate": 7.253846153846154e-07,
      "loss": 0.1063,
      "reward": 0.7262226939201355,
      "reward_std": 0.05605196766555309,
      "reward_std_all": 0.06700367480516434,
      "rewards-max/forensics_sinkhorn_reward": 0.9732484519481659,
      "rewards-max/forensics_strict_edge_reward": 0.5303425639867783,
      "rewards-max/forensics_tiou_reward": 0.9344777464866638,
      "rewards/forensics_sinkhorn_reward": 0.6316530108451843,
      "rewards/forensics_strict_edge_reward": 0.04080430045723915,
      "rewards/forensics_tiou_reward": 0.43947505950927734,
      "step": 358
    },
    {
      "completion_length": 24.90625,
      "epoch": 1.1013824884792627,
      "grad_norm": 5.539621829986572,
      "kl": 0.33740234375,
      "learning_rate": 7.246153846153846e-07,
      "loss": 0.0584,
      "reward": 0.7549615800380707,
      "reward_std": 0.019663209095597267,
      "reward_std_all": 0.03508997615426779,
      "rewards-max/forensics_sinkhorn_reward": 0.9217298924922943,
      "rewards-max/forensics_strict_edge_reward": 0.13229437702102587,
      "rewards-max/forensics_tiou_reward": 0.808546781539917,
      "rewards/forensics_sinkhorn_reward": 0.6615115404129028,
      "rewards/forensics_strict_edge_reward": 0.007971467432071222,
      "rewards/forensics_tiou_reward": 0.4017200469970703,
      "step": 359
    },
    {
      "completion_length": 23.796875,
      "epoch": 1.1044546850998465,
      "grad_norm": 5.659606456756592,
      "kl": 0.361328125,
      "learning_rate": 7.238461538461538e-07,
      "loss": 0.1215,
      "reward": 0.7772691547870636,
      "reward_std": 0.03252823418006301,
      "reward_std_all": 0.03323388798162341,
      "rewards-max/forensics_sinkhorn_reward": 0.99406698346138,
      "rewards-max/forensics_strict_edge_reward": 0.10633997619152069,
      "rewards-max/forensics_tiou_reward": 0.9881842732429504,
      "rewards/forensics_sinkhorn_reward": 0.6877748966217041,
      "rewards/forensics_strict_edge_reward": 0.006788789294660091,
      "rewards/forensics_tiou_reward": 0.5307824909687042,
      "step": 360
    },
    {
      "completion_length": 24.328125,
      "epoch": 1.10752688172043,
      "grad_norm": 4.866223335266113,
      "kl": 0.2939453125,
      "learning_rate": 7.23076923076923e-07,
      "loss": 0.0969,
      "reward": 0.7519306540489197,
      "reward_std": 0.054466817528009415,
      "reward_std_all": 0.054357290267944336,
      "rewards-max/forensics_sinkhorn_reward": 0.9518052935600281,
      "rewards-max/forensics_strict_edge_reward": 0.18931212509050965,
      "rewards-max/forensics_tiou_reward": 0.8837524056434631,
      "rewards/forensics_sinkhorn_reward": 0.6603226065635681,
      "rewards/forensics_strict_edge_reward": 0.012932087003719062,
      "rewards/forensics_tiou_reward": 0.4509400427341461,
      "step": 361
    },
    {
      "completion_length": 25.953125,
      "epoch": 1.1105990783410138,
      "grad_norm": 4.763815879821777,
      "kl": 0.3173828125,
      "learning_rate": 7.223076923076922e-07,
      "loss": 0.046,
      "reward": 0.7485194504261017,
      "reward_std": 0.04758797772228718,
      "reward_std_all": 0.05101892352104187,
      "rewards-max/forensics_sinkhorn_reward": 0.9517664909362793,
      "rewards-max/forensics_strict_edge_reward": 0.3661011606454849,
      "rewards-max/forensics_tiou_reward": 0.8625157475471497,
      "rewards/forensics_sinkhorn_reward": 0.626522421836853,
      "rewards/forensics_strict_edge_reward": 0.027027607895433903,
      "rewards/forensics_tiou_reward": 0.4457990825176239,
      "step": 362
    },
    {
      "completion_length": 26.90625,
      "epoch": 1.1136712749615976,
      "grad_norm": 5.507283687591553,
      "kl": 0.30859375,
      "learning_rate": 7.215384615384616e-07,
      "loss": 0.0126,
      "reward": 0.754037469625473,
      "reward_std": 0.06628512032330036,
      "reward_std_all": 0.062332672998309135,
      "rewards-max/forensics_sinkhorn_reward": 0.9778943061828613,
      "rewards-max/forensics_strict_edge_reward": 0.46924614906311035,
      "rewards-max/forensics_tiou_reward": 0.9500379264354706,
      "rewards/forensics_sinkhorn_reward": 0.7095658779144287,
      "rewards/forensics_strict_edge_reward": 0.03804783709347248,
      "rewards/forensics_tiou_reward": 0.4893764853477478,
      "step": 363
    },
    {
      "completion_length": 26.46875,
      "epoch": 1.1167434715821813,
      "grad_norm": 4.920435428619385,
      "kl": 0.361328125,
      "learning_rate": 7.207692307692307e-07,
      "loss": 0.0699,
      "reward": 0.7357412278652191,
      "reward_std": 0.04175851307809353,
      "reward_std_all": 0.05771678872406483,
      "rewards-max/forensics_sinkhorn_reward": 0.9469936490058899,
      "rewards-max/forensics_strict_edge_reward": 0.3522630734369159,
      "rewards-max/forensics_tiou_reward": 0.8718544542789459,
      "rewards/forensics_sinkhorn_reward": 0.658154308795929,
      "rewards/forensics_strict_edge_reward": 0.021014528232626617,
      "rewards/forensics_tiou_reward": 0.42568767070770264,
      "step": 364
    },
    {
      "completion_length": 22.234375,
      "epoch": 1.119815668202765,
      "grad_norm": 6.189783096313477,
      "kl": 0.349609375,
      "learning_rate": 7.2e-07,
      "loss": 0.0231,
      "reward": 0.7509569227695465,
      "reward_std": 0.026288275606930256,
      "reward_std_all": 0.04040690325200558,
      "rewards-max/forensics_sinkhorn_reward": 0.9882825911045074,
      "rewards-max/forensics_strict_edge_reward": 0.6680246889591217,
      "rewards-max/forensics_tiou_reward": 0.9772022366523743,
      "rewards/forensics_sinkhorn_reward": 0.6966638565063477,
      "rewards/forensics_strict_edge_reward": 0.04304313659667969,
      "rewards/forensics_tiou_reward": 0.4014739841222763,
      "step": 365
    },
    {
      "completion_length": 23.421875,
      "epoch": 1.1228878648233487,
      "grad_norm": 5.707892417907715,
      "kl": 0.353515625,
      "learning_rate": 7.192307692307692e-07,
      "loss": 0.076,
      "reward": 0.7483622431755066,
      "reward_std": 0.056164147797971964,
      "reward_std_all": 0.05660988762974739,
      "rewards-max/forensics_sinkhorn_reward": 0.979590892791748,
      "rewards-max/forensics_strict_edge_reward": 0.288161039352417,
      "rewards-max/forensics_tiou_reward": 0.9515066146850586,
      "rewards/forensics_sinkhorn_reward": 0.7079764902591705,
      "rewards/forensics_strict_edge_reward": 0.01415075920522213,
      "rewards/forensics_tiou_reward": 0.45719389617443085,
      "step": 366
    },
    {
      "completion_length": 27.78125,
      "epoch": 1.1259600614439325,
      "grad_norm": 4.722048282623291,
      "kl": 0.3583984375,
      "learning_rate": 7.184615384615384e-07,
      "loss": 0.1305,
      "reward": 0.7188207805156708,
      "reward_std": 0.09141794964671135,
      "reward_std_all": 0.0893256813287735,
      "rewards-max/forensics_sinkhorn_reward": 0.9700005948543549,
      "rewards-max/forensics_strict_edge_reward": 0.4307430237531662,
      "rewards-max/forensics_tiou_reward": 0.9174594283103943,
      "rewards/forensics_sinkhorn_reward": 0.6324282884597778,
      "rewards/forensics_strict_edge_reward": 0.026602616533637047,
      "rewards/forensics_tiou_reward": 0.47748930752277374,
      "step": 367
    },
    {
      "completion_length": 26.921875,
      "epoch": 1.129032258064516,
      "grad_norm": 4.666380405426025,
      "kl": 0.43359375,
      "learning_rate": 7.176923076923077e-07,
      "loss": 0.1164,
      "reward": 0.764978289604187,
      "reward_std": 0.05241192318499088,
      "reward_std_all": 0.05046766437590122,
      "rewards-max/forensics_sinkhorn_reward": 0.9706814587116241,
      "rewards-max/forensics_strict_edge_reward": 0.11363368853926659,
      "rewards-max/forensics_tiou_reward": 0.9274121522903442,
      "rewards/forensics_sinkhorn_reward": 0.6771484613418579,
      "rewards/forensics_strict_edge_reward": 0.006560050882399082,
      "rewards/forensics_tiou_reward": 0.3851304054260254,
      "step": 368
    },
    {
      "completion_length": 27.1875,
      "epoch": 1.1321044546850998,
      "grad_norm": 4.837963104248047,
      "kl": 0.3544921875,
      "learning_rate": 7.169230769230768e-07,
      "loss": 0.1692,
      "reward": 0.7423298060894012,
      "reward_std": 0.06861775740981102,
      "reward_std_all": 0.06811423785984516,
      "rewards-max/forensics_sinkhorn_reward": 0.9882912039756775,
      "rewards-max/forensics_strict_edge_reward": 0.11443570349365473,
      "rewards-max/forensics_tiou_reward": 0.9666639268398285,
      "rewards/forensics_sinkhorn_reward": 0.6467721164226532,
      "rewards/forensics_strict_edge_reward": 0.005978668807074428,
      "rewards/forensics_tiou_reward": 0.4575549513101578,
      "step": 369
    },
    {
      "completion_length": 26.5625,
      "epoch": 1.1351766513056836,
      "grad_norm": 5.81068229675293,
      "kl": 0.37109375,
      "learning_rate": 7.161538461538461e-07,
      "loss": 0.1148,
      "reward": 0.6989960074424744,
      "reward_std": 0.11431531608104706,
      "reward_std_all": 0.11056223884224892,
      "rewards-max/forensics_sinkhorn_reward": 0.9561473429203033,
      "rewards-max/forensics_strict_edge_reward": 0.4194317013025284,
      "rewards-max/forensics_tiou_reward": 0.8901887238025665,
      "rewards/forensics_sinkhorn_reward": 0.5917913615703583,
      "rewards/forensics_strict_edge_reward": 0.02461397834122181,
      "rewards/forensics_tiou_reward": 0.4135069102048874,
      "step": 370
    },
    {
      "completion_length": 22.59375,
      "epoch": 1.1382488479262673,
      "grad_norm": 5.293907642364502,
      "kl": 0.365234375,
      "learning_rate": 7.153846153846154e-07,
      "loss": -0.0488,
      "reward": 0.7569887042045593,
      "reward_std": 0.034401215612888336,
      "reward_std_all": 0.045553444884717464,
      "rewards-max/forensics_sinkhorn_reward": 0.9688210189342499,
      "rewards-max/forensics_strict_edge_reward": 0.1447548232972622,
      "rewards-max/forensics_tiou_reward": 0.9197334051132202,
      "rewards/forensics_sinkhorn_reward": 0.7003194391727448,
      "rewards/forensics_strict_edge_reward": 0.010885520372539759,
      "rewards/forensics_tiou_reward": 0.43786993622779846,
      "step": 371
    },
    {
      "completion_length": 26.546875,
      "epoch": 1.141321044546851,
      "grad_norm": 4.998113632202148,
      "kl": 0.35546875,
      "learning_rate": 7.146153846153846e-07,
      "loss": 0.0308,
      "reward": 0.7491040825843811,
      "reward_std": 0.0405058404430747,
      "reward_std_all": 0.05011710152029991,
      "rewards-max/forensics_sinkhorn_reward": 0.965063065290451,
      "rewards-max/forensics_strict_edge_reward": 0.14025845751166344,
      "rewards-max/forensics_tiou_reward": 0.9152015447616577,
      "rewards/forensics_sinkhorn_reward": 0.6804773509502411,
      "rewards/forensics_strict_edge_reward": 0.010141447419300675,
      "rewards/forensics_tiou_reward": 0.46110183000564575,
      "step": 372
    },
    {
      "completion_length": 25.03125,
      "epoch": 1.1443932411674347,
      "grad_norm": 5.41843843460083,
      "kl": 0.3125,
      "learning_rate": 7.138461538461538e-07,
      "loss": 0.0646,
      "reward": 0.7669853568077087,
      "reward_std": 0.0580275971442461,
      "reward_std_all": 0.0534922294318676,
      "rewards-max/forensics_sinkhorn_reward": 0.9547394216060638,
      "rewards-max/forensics_strict_edge_reward": 0.2193003073334694,
      "rewards-max/forensics_tiou_reward": 0.8916864693164825,
      "rewards/forensics_sinkhorn_reward": 0.6971619129180908,
      "rewards/forensics_strict_edge_reward": 0.018689066171646118,
      "rewards/forensics_tiou_reward": 0.4899093508720398,
      "step": 373
    },
    {
      "completion_length": 24.40625,
      "epoch": 1.1474654377880185,
      "grad_norm": 5.452053546905518,
      "kl": 0.400390625,
      "learning_rate": 7.13076923076923e-07,
      "loss": 0.1991,
      "reward": 0.7215258181095123,
      "reward_std": 0.0844029039144516,
      "reward_std_all": 0.08671204745769501,
      "rewards-max/forensics_sinkhorn_reward": 0.9726901352405548,
      "rewards-max/forensics_strict_edge_reward": 0.27964235097169876,
      "rewards-max/forensics_tiou_reward": 0.9247352480888367,
      "rewards/forensics_sinkhorn_reward": 0.6638968586921692,
      "rewards/forensics_strict_edge_reward": 0.02609150856733322,
      "rewards/forensics_tiou_reward": 0.42996135354042053,
      "step": 374
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.1505376344086022,
      "grad_norm": 5.583658695220947,
      "kl": 0.22412109375,
      "learning_rate": 7.123076923076922e-07,
      "loss": 0.1241,
      "reward": 0.7249955236911774,
      "reward_std": 0.04320471966639161,
      "reward_std_all": 0.05947355180978775,
      "rewards-max/forensics_sinkhorn_reward": 0.9442494213581085,
      "rewards-max/forensics_strict_edge_reward": 0.27156948670744896,
      "rewards-max/forensics_tiou_reward": 0.8761278092861176,
      "rewards/forensics_sinkhorn_reward": 0.5974512994289398,
      "rewards/forensics_strict_edge_reward": 0.02761226613074541,
      "rewards/forensics_tiou_reward": 0.3885409086942673,
      "step": 375
    },
    {
      "completion_length": 26.15625,
      "epoch": 1.1536098310291858,
      "grad_norm": 5.17355489730835,
      "kl": 0.3369140625,
      "learning_rate": 7.115384615384616e-07,
      "loss": 0.1196,
      "reward": 0.6911277770996094,
      "reward_std": 0.07186531648039818,
      "reward_std_all": 0.08861636370420456,
      "rewards-max/forensics_sinkhorn_reward": 0.9834083616733551,
      "rewards-max/forensics_strict_edge_reward": 0.49686121940612793,
      "rewards-max/forensics_tiou_reward": 0.962589830160141,
      "rewards/forensics_sinkhorn_reward": 0.6299414336681366,
      "rewards/forensics_strict_edge_reward": 0.04004084877669811,
      "rewards/forensics_tiou_reward": 0.4554920941591263,
      "step": 376
    },
    {
      "completion_length": 24.859375,
      "epoch": 1.1566820276497696,
      "grad_norm": 5.392961502075195,
      "kl": 0.28515625,
      "learning_rate": 7.107692307692308e-07,
      "loss": 0.0505,
      "reward": 0.7669446170330048,
      "reward_std": 0.03584728669375181,
      "reward_std_all": 0.03756070043891668,
      "rewards-max/forensics_sinkhorn_reward": 0.9735502004623413,
      "rewards-max/forensics_strict_edge_reward": 0.21773628145456314,
      "rewards-max/forensics_tiou_reward": 0.9243435859680176,
      "rewards/forensics_sinkhorn_reward": 0.67772376537323,
      "rewards/forensics_strict_edge_reward": 0.012228496838361025,
      "rewards/forensics_tiou_reward": 0.3921969532966614,
      "step": 377
    },
    {
      "completion_length": 22.90625,
      "epoch": 1.1597542242703534,
      "grad_norm": 4.93278169631958,
      "kl": 0.375,
      "learning_rate": 7.1e-07,
      "loss": -0.0041,
      "reward": 0.7699393630027771,
      "reward_std": 0.05077749490737915,
      "reward_std_all": 0.047361183911561966,
      "rewards-max/forensics_sinkhorn_reward": 0.9601918160915375,
      "rewards-max/forensics_strict_edge_reward": 0.34727809578180313,
      "rewards-max/forensics_tiou_reward": 0.9052928388118744,
      "rewards/forensics_sinkhorn_reward": 0.7182230353355408,
      "rewards/forensics_strict_edge_reward": 0.022628742037341,
      "rewards/forensics_tiou_reward": 0.46036049723625183,
      "step": 378
    },
    {
      "completion_length": 22.921875,
      "epoch": 1.1628264208909371,
      "grad_norm": 4.641942501068115,
      "kl": 0.412109375,
      "learning_rate": 7.092307692307692e-07,
      "loss": 0.0063,
      "reward": 0.7563817799091339,
      "reward_std": 0.013365383725613356,
      "reward_std_all": 0.03206893429160118,
      "rewards-max/forensics_sinkhorn_reward": 0.9490380883216858,
      "rewards-max/forensics_strict_edge_reward": 0.37350234389305115,
      "rewards-max/forensics_tiou_reward": 0.8763048946857452,
      "rewards/forensics_sinkhorn_reward": 0.6511643528938293,
      "rewards/forensics_strict_edge_reward": 0.021002952940762043,
      "rewards/forensics_tiou_reward": 0.4444734901189804,
      "step": 379
    },
    {
      "completion_length": 21.96875,
      "epoch": 1.1658986175115207,
      "grad_norm": 5.789433002471924,
      "kl": 0.4453125,
      "learning_rate": 7.084615384615384e-07,
      "loss": 0.0662,
      "reward": 0.7386005222797394,
      "reward_std": 0.06075270939618349,
      "reward_std_all": 0.06867910362780094,
      "rewards-max/forensics_sinkhorn_reward": 0.967946857213974,
      "rewards-max/forensics_strict_edge_reward": 0.15277573466300964,
      "rewards-max/forensics_tiou_reward": 0.9137626886367798,
      "rewards/forensics_sinkhorn_reward": 0.7132026553153992,
      "rewards/forensics_strict_edge_reward": 0.011254480807110667,
      "rewards/forensics_tiou_reward": 0.48322694003582,
      "step": 380
    },
    {
      "completion_length": 23.859375,
      "epoch": 1.1689708141321045,
      "grad_norm": 7.187836170196533,
      "kl": 0.5693359375,
      "learning_rate": 7.076923076923077e-07,
      "loss": -0.0116,
      "reward": 0.7429559826850891,
      "reward_std": 0.03420859947800636,
      "reward_std_all": 0.04703456535935402,
      "rewards-max/forensics_sinkhorn_reward": 0.9707276225090027,
      "rewards-max/forensics_strict_edge_reward": 0.3543456792831421,
      "rewards-max/forensics_tiou_reward": 0.9291967153549194,
      "rewards/forensics_sinkhorn_reward": 0.6851527392864227,
      "rewards/forensics_strict_edge_reward": 0.03791977418586612,
      "rewards/forensics_tiou_reward": 0.42744219303131104,
      "step": 381
    },
    {
      "completion_length": 21.6875,
      "epoch": 1.1720430107526882,
      "grad_norm": 5.2382049560546875,
      "kl": 0.341796875,
      "learning_rate": 7.069230769230768e-07,
      "loss": 0.0405,
      "reward": 0.7269852757453918,
      "reward_std": 0.041824253275990486,
      "reward_std_all": 0.0582379586994648,
      "rewards-max/forensics_sinkhorn_reward": 0.9463434219360352,
      "rewards-max/forensics_strict_edge_reward": 0.32564708963036537,
      "rewards-max/forensics_tiou_reward": 0.858163982629776,
      "rewards/forensics_sinkhorn_reward": 0.6493801772594452,
      "rewards/forensics_strict_edge_reward": 0.018560192547738552,
      "rewards/forensics_tiou_reward": 0.35820239782333374,
      "step": 382
    },
    {
      "completion_length": 25.34375,
      "epoch": 1.1751152073732718,
      "grad_norm": 4.591296195983887,
      "kl": 0.4462890625,
      "learning_rate": 7.061538461538462e-07,
      "loss": -0.0005,
      "reward": 0.7801892161369324,
      "reward_std": 0.02333345264196396,
      "reward_std_all": 0.025259695015847683,
      "rewards-max/forensics_sinkhorn_reward": 0.971091091632843,
      "rewards-max/forensics_strict_edge_reward": 0.39240114390850067,
      "rewards-max/forensics_tiou_reward": 0.9269677698612213,
      "rewards/forensics_sinkhorn_reward": 0.6914655566215515,
      "rewards/forensics_strict_edge_reward": 0.02673044055700302,
      "rewards/forensics_tiou_reward": 0.481902539730072,
      "step": 383
    },
    {
      "completion_length": 25.96875,
      "epoch": 1.1781874039938556,
      "grad_norm": 4.868441104888916,
      "kl": 0.2548828125,
      "learning_rate": 7.053846153846154e-07,
      "loss": 0.1643,
      "reward": 0.7582250237464905,
      "reward_std": 0.05395969469100237,
      "reward_std_all": 0.054555175825953484,
      "rewards-max/forensics_sinkhorn_reward": 0.9936092793941498,
      "rewards-max/forensics_strict_edge_reward": 0.6600648462772369,
      "rewards-max/forensics_tiou_reward": 0.9896641075611115,
      "rewards/forensics_sinkhorn_reward": 0.692709356546402,
      "rewards/forensics_strict_edge_reward": 0.043305269442498684,
      "rewards/forensics_tiou_reward": 0.4694010615348816,
      "step": 384
    },
    {
      "completion_length": 21.625,
      "epoch": 1.1812596006144394,
      "grad_norm": 6.582262992858887,
      "kl": 0.416015625,
      "learning_rate": 7.046153846153846e-07,
      "loss": 0.0498,
      "reward": 0.7361218631267548,
      "reward_std": 0.08209607750177383,
      "reward_std_all": 0.07803916744887829,
      "rewards-max/forensics_sinkhorn_reward": 0.9664205610752106,
      "rewards-max/forensics_strict_edge_reward": 0.287113593891263,
      "rewards-max/forensics_tiou_reward": 0.915979653596878,
      "rewards/forensics_sinkhorn_reward": 0.6828438639640808,
      "rewards/forensics_strict_edge_reward": 0.02531897625885904,
      "rewards/forensics_tiou_reward": 0.42640742659568787,
      "step": 385
    },
    {
      "completion_length": 21.359375,
      "epoch": 1.1843317972350231,
      "grad_norm": 5.869760990142822,
      "kl": 0.34521484375,
      "learning_rate": 7.038461538461539e-07,
      "loss": -0.0053,
      "reward": 0.7554660439491272,
      "reward_std": 0.042806767858564854,
      "reward_std_all": 0.04556088708341122,
      "rewards-max/forensics_sinkhorn_reward": 0.9823467135429382,
      "rewards-max/forensics_strict_edge_reward": 0.6048385128378868,
      "rewards-max/forensics_tiou_reward": 0.9573124647140503,
      "rewards/forensics_sinkhorn_reward": 0.7135446071624756,
      "rewards/forensics_strict_edge_reward": 0.03492504917085171,
      "rewards/forensics_tiou_reward": 0.44670841097831726,
      "step": 386
    },
    {
      "completion_length": 23.859375,
      "epoch": 1.1874039938556067,
      "grad_norm": 5.4833807945251465,
      "kl": 0.28076171875,
      "learning_rate": 7.03076923076923e-07,
      "loss": 0.1794,
      "reward": 0.7528852224349976,
      "reward_std": 0.05801033787429333,
      "reward_std_all": 0.05668332800269127,
      "rewards-max/forensics_sinkhorn_reward": 0.9706824123859406,
      "rewards-max/forensics_strict_edge_reward": 0.11725732311606407,
      "rewards-max/forensics_tiou_reward": 0.9264115989208221,
      "rewards/forensics_sinkhorn_reward": 0.6849562525749207,
      "rewards/forensics_strict_edge_reward": 0.01365137193351984,
      "rewards/forensics_tiou_reward": 0.44305771589279175,
      "step": 387
    },
    {
      "completion_length": 22.3125,
      "epoch": 1.1904761904761905,
      "grad_norm": 4.624153137207031,
      "kl": 0.32177734375,
      "learning_rate": 7.023076923076922e-07,
      "loss": 0.0586,
      "reward": 0.7841036915779114,
      "reward_std": 0.03365197032690048,
      "reward_std_all": 0.03163154702633619,
      "rewards-max/forensics_sinkhorn_reward": 0.9734630584716797,
      "rewards-max/forensics_strict_edge_reward": 0.5433230996131897,
      "rewards-max/forensics_tiou_reward": 0.9300622045993805,
      "rewards/forensics_sinkhorn_reward": 0.7534496188163757,
      "rewards/forensics_strict_edge_reward": 0.039387209340929985,
      "rewards/forensics_tiou_reward": 0.4960188567638397,
      "step": 388
    },
    {
      "completion_length": 25.390625,
      "epoch": 1.1935483870967742,
      "grad_norm": 5.267916202545166,
      "kl": 0.26953125,
      "learning_rate": 7.015384615384615e-07,
      "loss": 0.1287,
      "reward": 0.7233773171901703,
      "reward_std": 0.06707997620105743,
      "reward_std_all": 0.07231282070279121,
      "rewards-max/forensics_sinkhorn_reward": 0.986571341753006,
      "rewards-max/forensics_strict_edge_reward": 0.4592139273881912,
      "rewards-max/forensics_tiou_reward": 0.9718522727489471,
      "rewards/forensics_sinkhorn_reward": 0.6353369653224945,
      "rewards/forensics_strict_edge_reward": 0.024750807788223028,
      "rewards/forensics_tiou_reward": 0.40279240906238556,
      "step": 389
    },
    {
      "completion_length": 23.703125,
      "epoch": 1.1966205837173578,
      "grad_norm": 4.38957405090332,
      "kl": 0.2763671875,
      "learning_rate": 7.007692307692308e-07,
      "loss": 0.1132,
      "reward": 0.7693385183811188,
      "reward_std": 0.0489594005048275,
      "reward_std_all": 0.046319400891661644,
      "rewards-max/forensics_sinkhorn_reward": 0.9575026631355286,
      "rewards-max/forensics_strict_edge_reward": 0.5484330058097839,
      "rewards-max/forensics_tiou_reward": 0.9013011753559113,
      "rewards/forensics_sinkhorn_reward": 0.6889999210834503,
      "rewards/forensics_strict_edge_reward": 0.028293427079916,
      "rewards/forensics_tiou_reward": 0.44010908901691437,
      "step": 390
    },
    {
      "completion_length": 23.984375,
      "epoch": 1.1996927803379416,
      "grad_norm": 5.2644734382629395,
      "kl": 0.283203125,
      "learning_rate": 7e-07,
      "loss": 0.0544,
      "reward": 0.7639277875423431,
      "reward_std": 0.045645587146282196,
      "reward_std_all": 0.04519299976527691,
      "rewards-max/forensics_sinkhorn_reward": 0.9224826693534851,
      "rewards-max/forensics_strict_edge_reward": 0.17691032588481903,
      "rewards-max/forensics_tiou_reward": 0.8224073052406311,
      "rewards/forensics_sinkhorn_reward": 0.69784876704216,
      "rewards/forensics_strict_edge_reward": 0.01678573968820274,
      "rewards/forensics_tiou_reward": 0.4675176292657852,
      "step": 391
    },
    {
      "completion_length": 27.28125,
      "epoch": 1.2027649769585254,
      "grad_norm": 5.1530303955078125,
      "kl": 0.390625,
      "learning_rate": 6.992307692307692e-07,
      "loss": 0.0432,
      "reward": 0.757203221321106,
      "reward_std": 0.046124090906232595,
      "reward_std_all": 0.04736920353025198,
      "rewards-max/forensics_sinkhorn_reward": 0.9417102038860321,
      "rewards-max/forensics_strict_edge_reward": 0.1432527406141162,
      "rewards-max/forensics_tiou_reward": 0.822542667388916,
      "rewards/forensics_sinkhorn_reward": 0.6619702279567719,
      "rewards/forensics_strict_edge_reward": 0.006346856476739049,
      "rewards/forensics_tiou_reward": 0.40904465317726135,
      "step": 392
    },
    {
      "completion_length": 27.375,
      "epoch": 1.2058371735791091,
      "grad_norm": 4.3076090812683105,
      "kl": 0.3857421875,
      "learning_rate": 6.984615384615384e-07,
      "loss": 0.096,
      "reward": 0.7205041944980621,
      "reward_std": 0.059458451345562935,
      "reward_std_all": 0.07047024741768837,
      "rewards-max/forensics_sinkhorn_reward": 0.9726658463478088,
      "rewards-max/forensics_strict_edge_reward": 0.48739658296108246,
      "rewards-max/forensics_tiou_reward": 0.9215121865272522,
      "rewards/forensics_sinkhorn_reward": 0.6167637407779694,
      "rewards/forensics_strict_edge_reward": 0.022067095153033733,
      "rewards/forensics_tiou_reward": 0.36460213363170624,
      "step": 393
    },
    {
      "completion_length": 21.96875,
      "epoch": 1.2089093701996927,
      "grad_norm": 4.113668441772461,
      "kl": 0.3388671875,
      "learning_rate": 6.976923076923077e-07,
      "loss": 0.0177,
      "reward": 0.7635107338428497,
      "reward_std": 0.03572086896747351,
      "reward_std_all": 0.038787256460636854,
      "rewards-max/forensics_sinkhorn_reward": 0.9818221926689148,
      "rewards-max/forensics_strict_edge_reward": 0.6839397251605988,
      "rewards-max/forensics_tiou_reward": 0.9625101983547211,
      "rewards/forensics_sinkhorn_reward": 0.7036314904689789,
      "rewards/forensics_strict_edge_reward": 0.05531918257474899,
      "rewards/forensics_tiou_reward": 0.5032571852207184,
      "step": 394
    },
    {
      "completion_length": 23.671875,
      "epoch": 1.2119815668202765,
      "grad_norm": 6.669389247894287,
      "kl": 0.3134765625,
      "learning_rate": 6.969230769230769e-07,
      "loss": 0.0834,
      "reward": 0.7537776827812195,
      "reward_std": 0.047250498086214066,
      "reward_std_all": 0.04956083185970783,
      "rewards-max/forensics_sinkhorn_reward": 0.9866058826446533,
      "rewards-max/forensics_strict_edge_reward": 0.4194317013025284,
      "rewards-max/forensics_tiou_reward": 0.966254323720932,
      "rewards/forensics_sinkhorn_reward": 0.703373372554779,
      "rewards/forensics_strict_edge_reward": 0.046631019562482834,
      "rewards/forensics_tiou_reward": 0.4582591652870178,
      "step": 395
    },
    {
      "completion_length": 25.234375,
      "epoch": 1.2150537634408602,
      "grad_norm": 5.790189743041992,
      "kl": 0.45703125,
      "learning_rate": 6.961538461538461e-07,
      "loss": 0.0364,
      "reward": 0.7521827220916748,
      "reward_std": 0.054236773401498795,
      "reward_std_all": 0.05496343597769737,
      "rewards-max/forensics_sinkhorn_reward": 0.9601864516735077,
      "rewards-max/forensics_strict_edge_reward": 0.16005106642842293,
      "rewards-max/forensics_tiou_reward": 0.9093449115753174,
      "rewards/forensics_sinkhorn_reward": 0.6843173205852509,
      "rewards/forensics_strict_edge_reward": 0.013354179449379444,
      "rewards/forensics_tiou_reward": 0.42023609578609467,
      "step": 396
    },
    {
      "completion_length": 26.1875,
      "epoch": 1.218125960061444,
      "grad_norm": 5.369973659515381,
      "kl": 0.4169921875,
      "learning_rate": 6.953846153846154e-07,
      "loss": 0.12,
      "reward": 0.7280365824699402,
      "reward_std": 0.06972828134894371,
      "reward_std_all": 0.07269333675503731,
      "rewards-max/forensics_sinkhorn_reward": 0.9720582365989685,
      "rewards-max/forensics_strict_edge_reward": 0.2938617302570492,
      "rewards-max/forensics_tiou_reward": 0.9277516603469849,
      "rewards/forensics_sinkhorn_reward": 0.6199251711368561,
      "rewards/forensics_strict_edge_reward": 0.020838339605688816,
      "rewards/forensics_tiou_reward": 0.4082692414522171,
      "step": 397
    },
    {
      "completion_length": 22.140625,
      "epoch": 1.2211981566820276,
      "grad_norm": 7.131803035736084,
      "kl": 0.412109375,
      "learning_rate": 6.946153846153846e-07,
      "loss": 0.0336,
      "reward": 0.7618114352226257,
      "reward_std": 0.05415790528059006,
      "reward_std_all": 0.05156530253589153,
      "rewards-max/forensics_sinkhorn_reward": 0.9869585335254669,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9707041680812836,
      "rewards/forensics_sinkhorn_reward": 0.7510764896869659,
      "rewards/forensics_strict_edge_reward": 0.06362840346992016,
      "rewards/forensics_tiou_reward": 0.5306039452552795,
      "step": 398
    },
    {
      "completion_length": 25.90625,
      "epoch": 1.2242703533026114,
      "grad_norm": 4.823288917541504,
      "kl": 0.3349609375,
      "learning_rate": 6.938461538461539e-07,
      "loss": 0.0768,
      "reward": 0.7473236322402954,
      "reward_std": 0.06705239322036505,
      "reward_std_all": 0.06536923348903656,
      "rewards-max/forensics_sinkhorn_reward": 0.9717973172664642,
      "rewards-max/forensics_strict_edge_reward": 0.6317985653877258,
      "rewards-max/forensics_tiou_reward": 0.9395293891429901,
      "rewards/forensics_sinkhorn_reward": 0.658687174320221,
      "rewards/forensics_strict_edge_reward": 0.056618633680045605,
      "rewards/forensics_tiou_reward": 0.4994862526655197,
      "step": 399
    },
    {
      "completion_length": 28.21875,
      "epoch": 1.2273425499231951,
      "grad_norm": 4.1308369636535645,
      "kl": 0.3486328125,
      "learning_rate": 6.93076923076923e-07,
      "loss": 0.1115,
      "reward": 0.7527110576629639,
      "reward_std": 0.049120549112558365,
      "reward_std_all": 0.05519590154290199,
      "rewards-max/forensics_sinkhorn_reward": 0.9879539310932159,
      "rewards-max/forensics_strict_edge_reward": 0.42733459174633026,
      "rewards-max/forensics_tiou_reward": 0.9693936705589294,
      "rewards/forensics_sinkhorn_reward": 0.7387517392635345,
      "rewards/forensics_strict_edge_reward": 0.04617260582745075,
      "rewards/forensics_tiou_reward": 0.5508589446544647,
      "step": 400
    },
    {
      "completion_length": 26.359375,
      "epoch": 1.230414746543779,
      "grad_norm": 3.9666967391967773,
      "kl": 0.3173828125,
      "learning_rate": 6.923076923076922e-07,
      "loss": 0.0722,
      "reward": 0.752916544675827,
      "reward_std": 0.043252624571323395,
      "reward_std_all": 0.048598913475871086,
      "rewards-max/forensics_sinkhorn_reward": 0.9924980998039246,
      "rewards-max/forensics_strict_edge_reward": 0.36393168568611145,
      "rewards-max/forensics_tiou_reward": 0.984396755695343,
      "rewards/forensics_sinkhorn_reward": 0.6662412583827972,
      "rewards/forensics_strict_edge_reward": 0.04128481075167656,
      "rewards/forensics_tiou_reward": 0.5175532400608063,
      "step": 401
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.2334869431643625,
      "grad_norm": 4.840784072875977,
      "kl": 0.3720703125,
      "learning_rate": 6.915384615384615e-07,
      "loss": 0.1396,
      "reward": 0.6963960528373718,
      "reward_std": 0.09124956466257572,
      "reward_std_all": 0.09869369491934776,
      "rewards-max/forensics_sinkhorn_reward": 0.976394534111023,
      "rewards-max/forensics_strict_edge_reward": 0.20177672244608402,
      "rewards-max/forensics_tiou_reward": 0.9390417635440826,
      "rewards/forensics_sinkhorn_reward": 0.6188611090183258,
      "rewards/forensics_strict_edge_reward": 0.01716280821710825,
      "rewards/forensics_tiou_reward": 0.3673231825232506,
      "step": 402
    },
    {
      "completion_length": 24.71875,
      "epoch": 1.2365591397849462,
      "grad_norm": 5.233423709869385,
      "kl": 0.4169921875,
      "learning_rate": 6.907692307692307e-07,
      "loss": 0.0863,
      "reward": 0.7605732679367065,
      "reward_std": 0.0473434878513217,
      "reward_std_all": 0.04949565231800079,
      "rewards-max/forensics_sinkhorn_reward": 0.9688110947608948,
      "rewards-max/forensics_strict_edge_reward": 0.3826456218957901,
      "rewards-max/forensics_tiou_reward": 0.9311678409576416,
      "rewards/forensics_sinkhorn_reward": 0.6878921985626221,
      "rewards/forensics_strict_edge_reward": 0.03358905948698521,
      "rewards/forensics_tiou_reward": 0.4486214816570282,
      "step": 403
    },
    {
      "completion_length": 24.96875,
      "epoch": 1.23963133640553,
      "grad_norm": 5.5680623054504395,
      "kl": 0.3115234375,
      "learning_rate": 6.9e-07,
      "loss": 0.0702,
      "reward": 0.7510505616664886,
      "reward_std": 0.06687517743557692,
      "reward_std_all": 0.06383536569774151,
      "rewards-max/forensics_sinkhorn_reward": 0.9650123119354248,
      "rewards-max/forensics_strict_edge_reward": 0.38718849420547485,
      "rewards-max/forensics_tiou_reward": 0.9084735214710236,
      "rewards/forensics_sinkhorn_reward": 0.6822973191738129,
      "rewards/forensics_strict_edge_reward": 0.02477780170738697,
      "rewards/forensics_tiou_reward": 0.452099934220314,
      "step": 404
    },
    {
      "completion_length": 25.671875,
      "epoch": 1.2427035330261136,
      "grad_norm": 5.868531227111816,
      "kl": 0.310546875,
      "learning_rate": 6.892307692307692e-07,
      "loss": 0.1289,
      "reward": 0.7070065140724182,
      "reward_std": 0.08287698589265347,
      "reward_std_all": 0.09470966272056103,
      "rewards-max/forensics_sinkhorn_reward": 0.9889845252037048,
      "rewards-max/forensics_strict_edge_reward": 0.41483041644096375,
      "rewards-max/forensics_tiou_reward": 0.9773056507110596,
      "rewards/forensics_sinkhorn_reward": 0.6115294098854065,
      "rewards/forensics_strict_edge_reward": 0.028707236982882023,
      "rewards/forensics_tiou_reward": 0.4381200522184372,
      "step": 405
    },
    {
      "completion_length": 23.015625,
      "epoch": 1.2457757296466974,
      "grad_norm": 4.194766044616699,
      "kl": 0.345703125,
      "learning_rate": 6.884615384615384e-07,
      "loss": -0.0095,
      "reward": 0.780114084482193,
      "reward_std": 0.020400687586516142,
      "reward_std_all": 0.024496266152709723,
      "rewards-max/forensics_sinkhorn_reward": 0.9823126196861267,
      "rewards-max/forensics_strict_edge_reward": 0.5069752782583237,
      "rewards-max/forensics_tiou_reward": 0.9586417078971863,
      "rewards/forensics_sinkhorn_reward": 0.7466187477111816,
      "rewards/forensics_strict_edge_reward": 0.05034544877707958,
      "rewards/forensics_tiou_reward": 0.511938601732254,
      "step": 406
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.2488479262672811,
      "grad_norm": 5.159938812255859,
      "kl": 0.2705078125,
      "learning_rate": 6.876923076923077e-07,
      "loss": -0.0879,
      "reward": 0.7356584072113037,
      "reward_std": 0.02895463816821575,
      "reward_std_all": 0.049526846036314964,
      "rewards-max/forensics_sinkhorn_reward": 0.9793164134025574,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9537118077278137,
      "rewards/forensics_sinkhorn_reward": 0.6876153647899628,
      "rewards/forensics_strict_edge_reward": 0.03419883921742439,
      "rewards/forensics_tiou_reward": 0.44230589270591736,
      "step": 407
    },
    {
      "completion_length": 24.390625,
      "epoch": 1.251920122887865,
      "grad_norm": 4.79763650894165,
      "kl": 0.2939453125,
      "learning_rate": 6.869230769230769e-07,
      "loss": 0.1105,
      "reward": 0.7565384805202484,
      "reward_std": 0.04670144431293011,
      "reward_std_all": 0.0514631774276495,
      "rewards-max/forensics_sinkhorn_reward": 0.9729525744915009,
      "rewards-max/forensics_strict_edge_reward": 0.37596484273672104,
      "rewards-max/forensics_tiou_reward": 0.9391630589962006,
      "rewards/forensics_sinkhorn_reward": 0.7038261294364929,
      "rewards/forensics_strict_edge_reward": 0.025723968632519245,
      "rewards/forensics_tiou_reward": 0.4795930087566376,
      "step": 408
    },
    {
      "completion_length": 24.9375,
      "epoch": 1.2549923195084485,
      "grad_norm": 6.290646076202393,
      "kl": 0.330078125,
      "learning_rate": 6.861538461538461e-07,
      "loss": 0.142,
      "reward": 0.723054438829422,
      "reward_std": 0.06053803488612175,
      "reward_std_all": 0.07129521295428276,
      "rewards-max/forensics_sinkhorn_reward": 0.9743848443031311,
      "rewards-max/forensics_strict_edge_reward": 0.45590444654226303,
      "rewards-max/forensics_tiou_reward": 0.9364423155784607,
      "rewards/forensics_sinkhorn_reward": 0.6544384360313416,
      "rewards/forensics_strict_edge_reward": 0.030347459018230438,
      "rewards/forensics_tiou_reward": 0.46050965785980225,
      "step": 409
    },
    {
      "completion_length": 25.109375,
      "epoch": 1.2580645161290323,
      "grad_norm": 5.04243803024292,
      "kl": 0.3203125,
      "learning_rate": 6.853846153846153e-07,
      "loss": 0.0294,
      "reward": 0.7353610098361969,
      "reward_std": 0.06343322061002254,
      "reward_std_all": 0.07024967297911644,
      "rewards-max/forensics_sinkhorn_reward": 0.9493546783924103,
      "rewards-max/forensics_strict_edge_reward": 0.0680053336545825,
      "rewards-max/forensics_tiou_reward": 0.8589741289615631,
      "rewards/forensics_sinkhorn_reward": 0.6391811370849609,
      "rewards/forensics_strict_edge_reward": 0.008216053596697748,
      "rewards/forensics_tiou_reward": 0.44134649634361267,
      "step": 410
    },
    {
      "completion_length": 26.546875,
      "epoch": 1.261136712749616,
      "grad_norm": 5.043822765350342,
      "kl": 0.3095703125,
      "learning_rate": 6.846153846153846e-07,
      "loss": 0.0071,
      "reward": 0.7763749659061432,
      "reward_std": 0.032201020047068596,
      "reward_std_all": 0.03280991595238447,
      "rewards-max/forensics_sinkhorn_reward": 0.9785609841346741,
      "rewards-max/forensics_strict_edge_reward": 0.39014607667922974,
      "rewards-max/forensics_tiou_reward": 0.9494666755199432,
      "rewards/forensics_sinkhorn_reward": 0.7424322366714478,
      "rewards/forensics_strict_edge_reward": 0.030162204056978226,
      "rewards/forensics_tiou_reward": 0.5031673163175583,
      "step": 411
    },
    {
      "completion_length": 26.671875,
      "epoch": 1.2642089093701996,
      "grad_norm": 4.767712593078613,
      "kl": 0.326171875,
      "learning_rate": 6.838461538461539e-07,
      "loss": 0.1591,
      "reward": 0.7396848797798157,
      "reward_std": 0.033478979021310806,
      "reward_std_all": 0.05236761551350355,
      "rewards-max/forensics_sinkhorn_reward": 0.9586946964263916,
      "rewards-max/forensics_strict_edge_reward": 0.27454555965960026,
      "rewards-max/forensics_tiou_reward": 0.8927743136882782,
      "rewards/forensics_sinkhorn_reward": 0.685847133398056,
      "rewards/forensics_strict_edge_reward": 0.02965627983212471,
      "rewards/forensics_tiou_reward": 0.5146412551403046,
      "step": 412
    },
    {
      "completion_length": 24.4375,
      "epoch": 1.2672811059907834,
      "grad_norm": 5.021236419677734,
      "kl": 0.3330078125,
      "learning_rate": 6.83076923076923e-07,
      "loss": 0.0651,
      "reward": 0.739305168390274,
      "reward_std": 0.03190580592490733,
      "reward_std_all": 0.04901126027107239,
      "rewards-max/forensics_sinkhorn_reward": 0.9732182919979095,
      "rewards-max/forensics_strict_edge_reward": 0.4211113452911377,
      "rewards-max/forensics_tiou_reward": 0.9372412264347076,
      "rewards/forensics_sinkhorn_reward": 0.675116777420044,
      "rewards/forensics_strict_edge_reward": 0.02784483041614294,
      "rewards/forensics_tiou_reward": 0.4998699277639389,
      "step": 413
    },
    {
      "completion_length": 24.0625,
      "epoch": 1.2703533026113671,
      "grad_norm": 5.199314117431641,
      "kl": 0.26123046875,
      "learning_rate": 6.823076923076922e-07,
      "loss": 0.039,
      "reward": 0.7212823927402496,
      "reward_std": 0.07377473451197147,
      "reward_std_all": 0.07846421748399734,
      "rewards-max/forensics_sinkhorn_reward": 0.9857768714427948,
      "rewards-max/forensics_strict_edge_reward": 0.732279509305954,
      "rewards-max/forensics_tiou_reward": 0.9710111618041992,
      "rewards/forensics_sinkhorn_reward": 0.6676640510559082,
      "rewards/forensics_strict_edge_reward": 0.04696938954293728,
      "rewards/forensics_tiou_reward": 0.4494040459394455,
      "step": 414
    },
    {
      "completion_length": 27.265625,
      "epoch": 1.273425499231951,
      "grad_norm": 4.570586204528809,
      "kl": 0.296875,
      "learning_rate": 6.815384615384615e-07,
      "loss": 0.0642,
      "reward": 0.7666043043136597,
      "reward_std": 0.0444315318018198,
      "reward_std_all": 0.0452618021517992,
      "rewards-max/forensics_sinkhorn_reward": 0.9892932772636414,
      "rewards-max/forensics_strict_edge_reward": 0.5711847841739655,
      "rewards-max/forensics_tiou_reward": 0.977714329957962,
      "rewards/forensics_sinkhorn_reward": 0.6762014031410217,
      "rewards/forensics_strict_edge_reward": 0.039366573095321655,
      "rewards/forensics_tiou_reward": 0.4921724200248718,
      "step": 415
    },
    {
      "completion_length": 25.96875,
      "epoch": 1.2764976958525347,
      "grad_norm": 4.819485664367676,
      "kl": 0.3857421875,
      "learning_rate": 6.807692307692307e-07,
      "loss": 0.1981,
      "reward": 0.7697456777095795,
      "reward_std": 0.04685019701719284,
      "reward_std_all": 0.04410008154809475,
      "rewards-max/forensics_sinkhorn_reward": 0.9262037873268127,
      "rewards-max/forensics_strict_edge_reward": 0.24983903020620346,
      "rewards-max/forensics_tiou_reward": 0.8330709338188171,
      "rewards/forensics_sinkhorn_reward": 0.6763121783733368,
      "rewards/forensics_strict_edge_reward": 0.025992006761953235,
      "rewards/forensics_tiou_reward": 0.45793554186820984,
      "step": 416
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.2795698924731183,
      "grad_norm": 5.286667346954346,
      "kl": 0.3583984375,
      "learning_rate": 6.800000000000001e-07,
      "loss": 0.06,
      "reward": 0.7222400605678558,
      "reward_std": 0.0641727689653635,
      "reward_std_all": 0.07110879570245743,
      "rewards-max/forensics_sinkhorn_reward": 0.9554994404315948,
      "rewards-max/forensics_strict_edge_reward": 0.17573142051696777,
      "rewards-max/forensics_tiou_reward": 0.8885584473609924,
      "rewards/forensics_sinkhorn_reward": 0.6472739577293396,
      "rewards/forensics_strict_edge_reward": 0.009667448233813047,
      "rewards/forensics_tiou_reward": 0.39360755681991577,
      "step": 417
    },
    {
      "completion_length": 25.171875,
      "epoch": 1.282642089093702,
      "grad_norm": 5.242687702178955,
      "kl": 0.3876953125,
      "learning_rate": 6.792307692307692e-07,
      "loss": 0.0913,
      "reward": 0.7482171356678009,
      "reward_std": 0.04730677045881748,
      "reward_std_all": 0.05091714859008789,
      "rewards-max/forensics_sinkhorn_reward": 0.9734885990619659,
      "rewards-max/forensics_strict_edge_reward": 0.26503096520900726,
      "rewards-max/forensics_tiou_reward": 0.9324507415294647,
      "rewards/forensics_sinkhorn_reward": 0.637352466583252,
      "rewards/forensics_strict_edge_reward": 0.015590431168675423,
      "rewards/forensics_tiou_reward": 0.44288912415504456,
      "step": 418
    },
    {
      "completion_length": 21.734375,
      "epoch": 1.2857142857142856,
      "grad_norm": 4.749456405639648,
      "kl": 0.275390625,
      "learning_rate": 6.784615384615384e-07,
      "loss": 0.0155,
      "reward": 0.7685718834400177,
      "reward_std": 0.03152161091566086,
      "reward_std_all": 0.03773858956992626,
      "rewards-max/forensics_sinkhorn_reward": 0.981592983007431,
      "rewards-max/forensics_strict_edge_reward": 0.4899137169122696,
      "rewards-max/forensics_tiou_reward": 0.9593485891819,
      "rewards/forensics_sinkhorn_reward": 0.7373380661010742,
      "rewards/forensics_strict_edge_reward": 0.05095701292157173,
      "rewards/forensics_tiou_reward": 0.5109794437885284,
      "step": 419
    },
    {
      "completion_length": 26.515625,
      "epoch": 1.2887864823348694,
      "grad_norm": 7.612191200256348,
      "kl": 0.947265625,
      "learning_rate": 6.776923076923077e-07,
      "loss": 0.2058,
      "reward": 0.7223058342933655,
      "reward_std": 0.043755751103162766,
      "reward_std_all": 0.06158086284995079,
      "rewards-max/forensics_sinkhorn_reward": 0.9434673190116882,
      "rewards-max/forensics_strict_edge_reward": 0.19650311768054962,
      "rewards-max/forensics_tiou_reward": 0.87498739361763,
      "rewards/forensics_sinkhorn_reward": 0.6491176784038544,
      "rewards/forensics_strict_edge_reward": 0.010678443126380444,
      "rewards/forensics_tiou_reward": 0.4030814468860626,
      "step": 420
    },
    {
      "completion_length": 27.390625,
      "epoch": 1.2918586789554531,
      "grad_norm": 7.451809883117676,
      "kl": 0.451171875,
      "learning_rate": 6.769230769230769e-07,
      "loss": 0.2084,
      "reward": 0.7460572123527527,
      "reward_std": 0.04496142640709877,
      "reward_std_all": 0.05325719527900219,
      "rewards-max/forensics_sinkhorn_reward": 0.9754391014575958,
      "rewards-max/forensics_strict_edge_reward": 0.26900146901607513,
      "rewards-max/forensics_tiou_reward": 0.9382804930210114,
      "rewards/forensics_sinkhorn_reward": 0.6817847490310669,
      "rewards/forensics_strict_edge_reward": 0.018584758043289185,
      "rewards/forensics_tiou_reward": 0.46799837052822113,
      "step": 421
    },
    {
      "completion_length": 24.203125,
      "epoch": 1.294930875576037,
      "grad_norm": 5.379922389984131,
      "kl": 0.533203125,
      "learning_rate": 6.761538461538461e-07,
      "loss": 0.1573,
      "reward": 0.7489473819732666,
      "reward_std": 0.06264505162835121,
      "reward_std_all": 0.060849277302622795,
      "rewards-max/forensics_sinkhorn_reward": 0.9691171944141388,
      "rewards-max/forensics_strict_edge_reward": 0.40512125939130783,
      "rewards-max/forensics_tiou_reward": 0.925758421421051,
      "rewards/forensics_sinkhorn_reward": 0.69586181640625,
      "rewards/forensics_strict_edge_reward": 0.030966007616370916,
      "rewards/forensics_tiou_reward": 0.4687935560941696,
      "step": 422
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.2980030721966207,
      "grad_norm": 5.972044944763184,
      "kl": 0.337890625,
      "learning_rate": 6.753846153846153e-07,
      "loss": 0.0558,
      "reward": 0.7892248928546906,
      "reward_std": 0.02662648633122444,
      "reward_std_all": 0.025244392454624176,
      "rewards-max/forensics_sinkhorn_reward": 0.9770308136940002,
      "rewards-max/forensics_strict_edge_reward": 0.5296280682086945,
      "rewards-max/forensics_tiou_reward": 0.9445714950561523,
      "rewards/forensics_sinkhorn_reward": 0.7502128183841705,
      "rewards/forensics_strict_edge_reward": 0.06072027236223221,
      "rewards/forensics_tiou_reward": 0.5365877747535706,
      "step": 423
    },
    {
      "completion_length": 23.90625,
      "epoch": 1.3010752688172043,
      "grad_norm": 5.129899501800537,
      "kl": 0.27294921875,
      "learning_rate": 6.746153846153845e-07,
      "loss": 0.051,
      "reward": 0.7320708930492401,
      "reward_std": 0.04722261428833008,
      "reward_std_all": 0.059814102947711945,
      "rewards-max/forensics_sinkhorn_reward": 0.9675523936748505,
      "rewards-max/forensics_strict_edge_reward": 0.3917350545525551,
      "rewards-max/forensics_tiou_reward": 0.9255998730659485,
      "rewards/forensics_sinkhorn_reward": 0.658074826002121,
      "rewards/forensics_strict_edge_reward": 0.023593347519636154,
      "rewards/forensics_tiou_reward": 0.46200497448444366,
      "step": 424
    },
    {
      "completion_length": 23.203125,
      "epoch": 1.304147465437788,
      "grad_norm": 6.005635738372803,
      "kl": 0.3388671875,
      "learning_rate": 6.738461538461539e-07,
      "loss": 0.0179,
      "reward": 0.7725616991519928,
      "reward_std": 0.018443956971168518,
      "reward_std_all": 0.02907814458012581,
      "rewards-max/forensics_sinkhorn_reward": 0.9951135218143463,
      "rewards-max/forensics_strict_edge_reward": 0.7657686173915863,
      "rewards-max/forensics_tiou_reward": 0.9916885793209076,
      "rewards/forensics_sinkhorn_reward": 0.7444738149642944,
      "rewards/forensics_strict_edge_reward": 0.06420903652906418,
      "rewards/forensics_tiou_reward": 0.5114055275917053,
      "step": 425
    },
    {
      "completion_length": 23.609375,
      "epoch": 1.3072196620583718,
      "grad_norm": 7.528182029724121,
      "kl": 0.3212890625,
      "learning_rate": 6.730769230769231e-07,
      "loss": 0.0527,
      "reward": 0.735118180513382,
      "reward_std": 0.08597632497549057,
      "reward_std_all": 0.08136529847979546,
      "rewards-max/forensics_sinkhorn_reward": 0.9471087753772736,
      "rewards-max/forensics_strict_edge_reward": 0.18939974904060364,
      "rewards-max/forensics_tiou_reward": 0.85979825258255,
      "rewards/forensics_sinkhorn_reward": 0.6687658429145813,
      "rewards/forensics_strict_edge_reward": 0.01097539288457483,
      "rewards/forensics_tiou_reward": 0.41907384991645813,
      "step": 426
    },
    {
      "completion_length": 25.5,
      "epoch": 1.3102918586789554,
      "grad_norm": 4.878290176391602,
      "kl": 0.3525390625,
      "learning_rate": 6.723076923076922e-07,
      "loss": 0.0857,
      "reward": 0.7931793928146362,
      "reward_std": 0.020074713975191116,
      "reward_std_all": 0.019724945072084665,
      "rewards-max/forensics_sinkhorn_reward": 0.9834484159946442,
      "rewards-max/forensics_strict_edge_reward": 0.23274682462215424,
      "rewards-max/forensics_tiou_reward": 0.9614087343215942,
      "rewards/forensics_sinkhorn_reward": 0.752771645784378,
      "rewards/forensics_strict_edge_reward": 0.02810784289613366,
      "rewards/forensics_tiou_reward": 0.5745340585708618,
      "step": 427
    },
    {
      "completion_length": 22.484375,
      "epoch": 1.3133640552995391,
      "grad_norm": 8.57641315460205,
      "kl": 0.3779296875,
      "learning_rate": 6.715384615384615e-07,
      "loss": -0.0107,
      "reward": 0.7602266073226929,
      "reward_std": 0.03654865734279156,
      "reward_std_all": 0.042194509878754616,
      "rewards-max/forensics_sinkhorn_reward": 0.9750988483428955,
      "rewards-max/forensics_strict_edge_reward": 0.4496074318885803,
      "rewards-max/forensics_tiou_reward": 0.9408679008483887,
      "rewards/forensics_sinkhorn_reward": 0.7350961267948151,
      "rewards/forensics_strict_edge_reward": 0.030012805480509996,
      "rewards/forensics_tiou_reward": 0.4953364431858063,
      "step": 428
    },
    {
      "completion_length": 24.765625,
      "epoch": 1.316436251920123,
      "grad_norm": 5.859391689300537,
      "kl": 0.6240234375,
      "learning_rate": 6.707692307692307e-07,
      "loss": -0.0148,
      "reward": 0.767844945192337,
      "reward_std": 0.036461275070905685,
      "reward_std_all": 0.038228985853493214,
      "rewards-max/forensics_sinkhorn_reward": 0.973044753074646,
      "rewards-max/forensics_strict_edge_reward": 0.3375049978494644,
      "rewards-max/forensics_tiou_reward": 0.93653804063797,
      "rewards/forensics_sinkhorn_reward": 0.7208165526390076,
      "rewards/forensics_strict_edge_reward": 0.02725549042224884,
      "rewards/forensics_tiou_reward": 0.4839472025632858,
      "step": 429
    },
    {
      "completion_length": 29.09375,
      "epoch": 1.3195084485407067,
      "grad_norm": 3.791060209274292,
      "kl": 0.365234375,
      "learning_rate": 6.7e-07,
      "loss": 0.0223,
      "reward": 0.7654302716255188,
      "reward_std": 0.039498030208051205,
      "reward_std_all": 0.04341813921928406,
      "rewards-max/forensics_sinkhorn_reward": 0.9804562330245972,
      "rewards-max/forensics_strict_edge_reward": 0.7388437688350677,
      "rewards-max/forensics_tiou_reward": 0.9553344547748566,
      "rewards/forensics_sinkhorn_reward": 0.6817099153995514,
      "rewards/forensics_strict_edge_reward": 0.05371679738163948,
      "rewards/forensics_tiou_reward": 0.4988051950931549,
      "step": 430
    },
    {
      "completion_length": 26.578125,
      "epoch": 1.3225806451612903,
      "grad_norm": 5.8831868171691895,
      "kl": 0.53125,
      "learning_rate": 6.692307692307692e-07,
      "loss": 0.0779,
      "reward": 0.7719815671443939,
      "reward_std": 0.049942292273044586,
      "reward_std_all": 0.04645910859107971,
      "rewards-max/forensics_sinkhorn_reward": 0.9689469635486603,
      "rewards-max/forensics_strict_edge_reward": 0.5048972964286804,
      "rewards-max/forensics_tiou_reward": 0.9263157546520233,
      "rewards/forensics_sinkhorn_reward": 0.6766954958438873,
      "rewards/forensics_strict_edge_reward": 0.027785975486040115,
      "rewards/forensics_tiou_reward": 0.4303443282842636,
      "step": 431
    },
    {
      "completion_length": 23.75,
      "epoch": 1.325652841781874,
      "grad_norm": 6.936856746673584,
      "kl": 0.3037109375,
      "learning_rate": 6.684615384615384e-07,
      "loss": 0.0193,
      "reward": 0.7738098800182343,
      "reward_std": 0.041944957338273525,
      "reward_std_all": 0.04040749557316303,
      "rewards-max/forensics_sinkhorn_reward": 0.9678277373313904,
      "rewards-max/forensics_strict_edge_reward": 0.5383955463767052,
      "rewards-max/forensics_tiou_reward": 0.9236985445022583,
      "rewards/forensics_sinkhorn_reward": 0.7107379734516144,
      "rewards/forensics_strict_edge_reward": 0.0416178407613188,
      "rewards/forensics_tiou_reward": 0.4382704347372055,
      "step": 432
    },
    {
      "completion_length": 23.578125,
      "epoch": 1.3287250384024578,
      "grad_norm": 5.754964351654053,
      "kl": 0.33251953125,
      "learning_rate": 6.676923076923077e-07,
      "loss": -0.0165,
      "reward": 0.7697289288043976,
      "reward_std": 0.0414979662746191,
      "reward_std_all": 0.040996018797159195,
      "rewards-max/forensics_sinkhorn_reward": 0.9728754758834839,
      "rewards-max/forensics_strict_edge_reward": 0.3819463551044464,
      "rewards-max/forensics_tiou_reward": 0.9280888438224792,
      "rewards/forensics_sinkhorn_reward": 0.7273005545139313,
      "rewards/forensics_strict_edge_reward": 0.019958394579589367,
      "rewards/forensics_tiou_reward": 0.5277333110570908,
      "step": 433
    },
    {
      "completion_length": 30.015625,
      "epoch": 1.3317972350230414,
      "grad_norm": 5.6763505935668945,
      "kl": 0.4384765625,
      "learning_rate": 6.669230769230769e-07,
      "loss": 0.0461,
      "reward": 0.7537785470485687,
      "reward_std": 0.06713858246803284,
      "reward_std_all": 0.06304395198822021,
      "rewards-max/forensics_sinkhorn_reward": 0.9729738235473633,
      "rewards-max/forensics_strict_edge_reward": 0.25695058472047094,
      "rewards-max/forensics_tiou_reward": 0.941587895154953,
      "rewards/forensics_sinkhorn_reward": 0.6743507981300354,
      "rewards/forensics_strict_edge_reward": 0.01088157872982265,
      "rewards/forensics_tiou_reward": 0.49528828263282776,
      "step": 434
    },
    {
      "completion_length": 25.796875,
      "epoch": 1.3348694316436251,
      "grad_norm": 5.345578193664551,
      "kl": 0.32421875,
      "learning_rate": 6.661538461538462e-07,
      "loss": 0.0815,
      "reward": 0.7846550047397614,
      "reward_std": 0.022639577742666006,
      "reward_std_all": 0.02352837473154068,
      "rewards-max/forensics_sinkhorn_reward": 0.9808811247348785,
      "rewards-max/forensics_strict_edge_reward": 0.49555233120918274,
      "rewards-max/forensics_tiou_reward": 0.9553032219409943,
      "rewards/forensics_sinkhorn_reward": 0.7151209712028503,
      "rewards/forensics_strict_edge_reward": 0.041420361027121544,
      "rewards/forensics_tiou_reward": 0.48221391439437866,
      "step": 435
    },
    {
      "completion_length": 24.125,
      "epoch": 1.337941628264209,
      "grad_norm": 7.775149822235107,
      "kl": 0.3232421875,
      "learning_rate": 6.653846153846153e-07,
      "loss": 0.0852,
      "reward": 0.7452496886253357,
      "reward_std": 0.03276566229760647,
      "reward_std_all": 0.04856918938457966,
      "rewards-max/forensics_sinkhorn_reward": 0.9493491053581238,
      "rewards-max/forensics_strict_edge_reward": 0.27180543541908264,
      "rewards-max/forensics_tiou_reward": 0.8720245659351349,
      "rewards/forensics_sinkhorn_reward": 0.6969619393348694,
      "rewards/forensics_strict_edge_reward": 0.02151755103841424,
      "rewards/forensics_tiou_reward": 0.40476179122924805,
      "step": 436
    },
    {
      "completion_length": 26.578125,
      "epoch": 1.3410138248847927,
      "grad_norm": 6.7810211181640625,
      "kl": 0.54296875,
      "learning_rate": 6.646153846153845e-07,
      "loss": -0.028,
      "reward": 0.7384874820709229,
      "reward_std": 0.04084782116115093,
      "reward_std_all": 0.05314313992857933,
      "rewards-max/forensics_sinkhorn_reward": 0.9835934340953827,
      "rewards-max/forensics_strict_edge_reward": 0.3677225112915039,
      "rewards-max/forensics_tiou_reward": 0.9610633552074432,
      "rewards/forensics_sinkhorn_reward": 0.6961931586265564,
      "rewards/forensics_strict_edge_reward": 0.028526434674859047,
      "rewards/forensics_tiou_reward": 0.47144511342048645,
      "step": 437
    },
    {
      "completion_length": 22.59375,
      "epoch": 1.3440860215053765,
      "grad_norm": 6.412522792816162,
      "kl": 0.4287109375,
      "learning_rate": 6.638461538461539e-07,
      "loss": 0.064,
      "reward": 0.7053207159042358,
      "reward_std": 0.05824531242251396,
      "reward_std_all": 0.07743701711297035,
      "rewards-max/forensics_sinkhorn_reward": 0.9381330907344818,
      "rewards-max/forensics_strict_edge_reward": 0.07524841651320457,
      "rewards-max/forensics_tiou_reward": 0.8580000996589661,
      "rewards/forensics_sinkhorn_reward": 0.6047974526882172,
      "rewards/forensics_strict_edge_reward": 0.006552547652972862,
      "rewards/forensics_tiou_reward": 0.28337274491786957,
      "step": 438
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.34715821812596,
      "grad_norm": 6.0023088455200195,
      "kl": 0.3056640625,
      "learning_rate": 6.630769230769231e-07,
      "loss": 0.0264,
      "reward": 0.6917373538017273,
      "reward_std": 0.0648723766207695,
      "reward_std_all": 0.0860745832324028,
      "rewards-max/forensics_sinkhorn_reward": 0.9709559381008148,
      "rewards-max/forensics_strict_edge_reward": 0.26912766695022583,
      "rewards-max/forensics_tiou_reward": 0.9297679960727692,
      "rewards/forensics_sinkhorn_reward": 0.6064906716346741,
      "rewards/forensics_strict_edge_reward": 0.03173753013834357,
      "rewards/forensics_tiou_reward": 0.3613751083612442,
      "step": 439
    },
    {
      "completion_length": 27.375,
      "epoch": 1.3502304147465438,
      "grad_norm": 5.210806369781494,
      "kl": 0.3076171875,
      "learning_rate": 6.623076923076923e-07,
      "loss": 0.07,
      "reward": 0.7190967500209808,
      "reward_std": 0.0698830746114254,
      "reward_std_all": 0.08072448521852493,
      "rewards-max/forensics_sinkhorn_reward": 0.9731327295303345,
      "rewards-max/forensics_strict_edge_reward": 0.26451872289180756,
      "rewards-max/forensics_tiou_reward": 0.9259267747402191,
      "rewards/forensics_sinkhorn_reward": 0.6455923616886139,
      "rewards/forensics_strict_edge_reward": 0.02381822280585766,
      "rewards/forensics_tiou_reward": 0.45636437833309174,
      "step": 440
    },
    {
      "completion_length": 29.015625,
      "epoch": 1.3533026113671274,
      "grad_norm": 5.862890720367432,
      "kl": 0.3447265625,
      "learning_rate": 6.615384615384615e-07,
      "loss": 0.0861,
      "reward": 0.66880202293396,
      "reward_std": 0.09468507766723633,
      "reward_std_all": 0.11563056334853172,
      "rewards-max/forensics_sinkhorn_reward": 0.9751864671707153,
      "rewards-max/forensics_strict_edge_reward": 0.2057781796902418,
      "rewards-max/forensics_tiou_reward": 0.9305475354194641,
      "rewards/forensics_sinkhorn_reward": 0.5640108585357666,
      "rewards/forensics_strict_edge_reward": 0.02589963946957141,
      "rewards/forensics_tiou_reward": 0.41043388843536377,
      "step": 441
    },
    {
      "completion_length": 22.6875,
      "epoch": 1.3563748079877112,
      "grad_norm": 6.539273262023926,
      "kl": 0.41796875,
      "learning_rate": 6.607692307692307e-07,
      "loss": 0.1647,
      "reward": 0.7341549396514893,
      "reward_std": 0.07455840706825256,
      "reward_std_all": 0.07367152534425259,
      "rewards-max/forensics_sinkhorn_reward": 0.9776392877101898,
      "rewards-max/forensics_strict_edge_reward": 0.4789932817220688,
      "rewards-max/forensics_tiou_reward": 0.945690929889679,
      "rewards/forensics_sinkhorn_reward": 0.6761864721775055,
      "rewards/forensics_strict_edge_reward": 0.03404463641345501,
      "rewards/forensics_tiou_reward": 0.436258465051651,
      "step": 442
    },
    {
      "completion_length": 26.125,
      "epoch": 1.359447004608295,
      "grad_norm": 4.580578327178955,
      "kl": 0.2939453125,
      "learning_rate": 6.6e-07,
      "loss": 0.1503,
      "reward": 0.6929620206356049,
      "reward_std": 0.09978897124528885,
      "reward_std_all": 0.10878659039735794,
      "rewards-max/forensics_sinkhorn_reward": 0.9754625856876373,
      "rewards-max/forensics_strict_edge_reward": 0.24913404136896133,
      "rewards-max/forensics_tiou_reward": 0.937529444694519,
      "rewards/forensics_sinkhorn_reward": 0.6496526896953583,
      "rewards/forensics_strict_edge_reward": 0.016296171117573977,
      "rewards/forensics_tiou_reward": 0.4436102956533432,
      "step": 443
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.3625192012288787,
      "grad_norm": 4.480441093444824,
      "kl": 0.3994140625,
      "learning_rate": 6.592307692307691e-07,
      "loss": 0.0443,
      "reward": 0.7474785447120667,
      "reward_std": 0.03872872143983841,
      "reward_std_all": 0.05274871364235878,
      "rewards-max/forensics_sinkhorn_reward": 0.9791190326213837,
      "rewards-max/forensics_strict_edge_reward": 0.6720768958330154,
      "rewards-max/forensics_tiou_reward": 0.9557563364505768,
      "rewards/forensics_sinkhorn_reward": 0.7114701867103577,
      "rewards/forensics_strict_edge_reward": 0.05943806283175945,
      "rewards/forensics_tiou_reward": 0.47804972529411316,
      "step": 444
    },
    {
      "completion_length": 23.28125,
      "epoch": 1.3655913978494625,
      "grad_norm": 5.395405292510986,
      "kl": 0.3408203125,
      "learning_rate": 6.584615384615384e-07,
      "loss": -0.0143,
      "reward": 0.7575426697731018,
      "reward_std": 0.026901341043412685,
      "reward_std_all": 0.03937657084316015,
      "rewards-max/forensics_sinkhorn_reward": 0.9754660129547119,
      "rewards-max/forensics_strict_edge_reward": 0.46190081536769867,
      "rewards-max/forensics_tiou_reward": 0.9455539584159851,
      "rewards/forensics_sinkhorn_reward": 0.71762615442276,
      "rewards/forensics_strict_edge_reward": 0.03465091437101364,
      "rewards/forensics_tiou_reward": 0.5199063420295715,
      "step": 445
    },
    {
      "completion_length": 23.765625,
      "epoch": 1.368663594470046,
      "grad_norm": 5.493749618530273,
      "kl": 0.3681640625,
      "learning_rate": 6.576923076923077e-07,
      "loss": -0.0024,
      "reward": 0.7787907719612122,
      "reward_std": 0.04156233975663781,
      "reward_std_all": 0.03848874568939209,
      "rewards-max/forensics_sinkhorn_reward": 0.9935645759105682,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9896591007709503,
      "rewards/forensics_sinkhorn_reward": 0.7277045845985413,
      "rewards/forensics_strict_edge_reward": 0.07263955473899841,
      "rewards/forensics_tiou_reward": 0.5872860252857208,
      "step": 446
    },
    {
      "completion_length": 24.515625,
      "epoch": 1.3717357910906298,
      "grad_norm": 5.699827194213867,
      "kl": 0.3076171875,
      "learning_rate": 6.569230769230769e-07,
      "loss": 0.1371,
      "reward": 0.7568706274032593,
      "reward_std": 0.05992312356829643,
      "reward_std_all": 0.05671736225485802,
      "rewards-max/forensics_sinkhorn_reward": 0.9961248338222504,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.995674341917038,
      "rewards/forensics_sinkhorn_reward": 0.6864757239818573,
      "rewards/forensics_strict_edge_reward": 0.056559547781944275,
      "rewards/forensics_tiou_reward": 0.488031268119812,
      "step": 447
    },
    {
      "completion_length": 22.25,
      "epoch": 1.3748079877112136,
      "grad_norm": 5.364807605743408,
      "kl": 0.3203125,
      "learning_rate": 6.561538461538462e-07,
      "loss": -0.0095,
      "reward": 0.7610397934913635,
      "reward_std": 0.054812220856547356,
      "reward_std_all": 0.05234001949429512,
      "rewards-max/forensics_sinkhorn_reward": 0.972507894039154,
      "rewards-max/forensics_strict_edge_reward": 0.5066853165626526,
      "rewards-max/forensics_tiou_reward": 0.9419445395469666,
      "rewards/forensics_sinkhorn_reward": 0.6909572780132294,
      "rewards/forensics_strict_edge_reward": 0.03387968987226486,
      "rewards/forensics_tiou_reward": 0.39565509557724,
      "step": 448
    },
    {
      "completion_length": 25.109375,
      "epoch": 1.3778801843317972,
      "grad_norm": 6.347201824188232,
      "kl": 0.29345703125,
      "learning_rate": 6.553846153846153e-07,
      "loss": 0.0786,
      "reward": 0.6948161721229553,
      "reward_std": 0.04946986958384514,
      "reward_std_all": 0.08392883464694023,
      "rewards-max/forensics_sinkhorn_reward": 0.9967571496963501,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.9976251721382141,
      "rewards/forensics_sinkhorn_reward": 0.6413154304027557,
      "rewards/forensics_strict_edge_reward": 0.03380558267235756,
      "rewards/forensics_tiou_reward": 0.5350203514099121,
      "step": 449
    },
    {
      "completion_length": 30.71875,
      "epoch": 1.380952380952381,
      "grad_norm": 5.837617874145508,
      "kl": 0.3486328125,
      "learning_rate": 6.546153846153845e-07,
      "loss": 0.1289,
      "reward": 0.6998925805091858,
      "reward_std": 0.07631746400147676,
      "reward_std_all": 0.08947013132274151,
      "rewards-max/forensics_sinkhorn_reward": 0.9580668807029724,
      "rewards-max/forensics_strict_edge_reward": 0.24919188767671585,
      "rewards-max/forensics_tiou_reward": 0.905278742313385,
      "rewards/forensics_sinkhorn_reward": 0.6102206707000732,
      "rewards/forensics_strict_edge_reward": 0.020118217449635267,
      "rewards/forensics_tiou_reward": 0.4137441962957382,
      "step": 450
    },
    {
      "completion_length": 27.0625,
      "epoch": 1.3840245775729647,
      "grad_norm": 5.963652610778809,
      "kl": 0.3037109375,
      "learning_rate": 6.538461538461538e-07,
      "loss": 0.0434,
      "reward": 0.7073022127151489,
      "reward_std": 0.07734328880906105,
      "reward_std_all": 0.08706700056791306,
      "rewards-max/forensics_sinkhorn_reward": 0.9442713856697083,
      "rewards-max/forensics_strict_edge_reward": 0.035393324913457036,
      "rewards-max/forensics_tiou_reward": 0.8601963818073273,
      "rewards/forensics_sinkhorn_reward": 0.6002360582351685,
      "rewards/forensics_strict_edge_reward": 0.0027553939289646223,
      "rewards/forensics_tiou_reward": 0.35709747672080994,
      "step": 451
    },
    {
      "completion_length": 24.75,
      "epoch": 1.3870967741935485,
      "grad_norm": 7.254609107971191,
      "kl": 0.2880859375,
      "learning_rate": 6.530769230769231e-07,
      "loss": 0.0011,
      "reward": 0.7289600968360901,
      "reward_std": 0.0874587893486023,
      "reward_std_all": 0.0843588076531887,
      "rewards-max/forensics_sinkhorn_reward": 0.9761658906936646,
      "rewards-max/forensics_strict_edge_reward": 0.34197375178337097,
      "rewards-max/forensics_tiou_reward": 0.9357706010341644,
      "rewards/forensics_sinkhorn_reward": 0.7070660293102264,
      "rewards/forensics_strict_edge_reward": 0.02783184591680765,
      "rewards/forensics_tiou_reward": 0.4901932179927826,
      "step": 452
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.390168970814132,
      "grad_norm": 6.895351409912109,
      "kl": 0.357421875,
      "learning_rate": 6.523076923076923e-07,
      "loss": -0.0042,
      "reward": 0.767707884311676,
      "reward_std": 0.0420451033860445,
      "reward_std_all": 0.041814353317022324,
      "rewards-max/forensics_sinkhorn_reward": 0.9851518869400024,
      "rewards-max/forensics_strict_edge_reward": 0.6976466774940491,
      "rewards-max/forensics_tiou_reward": 0.9690403640270233,
      "rewards/forensics_sinkhorn_reward": 0.740924060344696,
      "rewards/forensics_strict_edge_reward": 0.06594271026551723,
      "rewards/forensics_tiou_reward": 0.5278467535972595,
      "step": 453
    },
    {
      "completion_length": 22.359375,
      "epoch": 1.3932411674347158,
      "grad_norm": 4.727535724639893,
      "kl": 0.359375,
      "learning_rate": 6.515384615384615e-07,
      "loss": 0.026,
      "reward": 0.7911863923072815,
      "reward_std": 0.024332770553883165,
      "reward_std_all": 0.023371543968096375,
      "rewards-max/forensics_sinkhorn_reward": 0.9700840413570404,
      "rewards-max/forensics_strict_edge_reward": 0.44792895019054413,
      "rewards-max/forensics_tiou_reward": 0.9373635649681091,
      "rewards/forensics_sinkhorn_reward": 0.753113329410553,
      "rewards/forensics_strict_edge_reward": 0.026398727670311928,
      "rewards/forensics_tiou_reward": 0.5389980524778366,
      "step": 454
    },
    {
      "completion_length": 23.078125,
      "epoch": 1.3963133640552996,
      "grad_norm": 7.8520307540893555,
      "kl": 0.3818359375,
      "learning_rate": 6.507692307692307e-07,
      "loss": 0.1218,
      "reward": 0.7379292845726013,
      "reward_std": 0.07027684897184372,
      "reward_std_all": 0.07082763314247131,
      "rewards-max/forensics_sinkhorn_reward": 0.9819803535938263,
      "rewards-max/forensics_strict_edge_reward": 0.5512756779789925,
      "rewards-max/forensics_tiou_reward": 0.9583109021186829,
      "rewards/forensics_sinkhorn_reward": 0.6579373478889465,
      "rewards/forensics_strict_edge_reward": 0.049568927846848965,
      "rewards/forensics_tiou_reward": 0.46531379222869873,
      "step": 455
    },
    {
      "completion_length": 23.625,
      "epoch": 1.3993855606758832,
      "grad_norm": 5.177594184875488,
      "kl": 0.4716796875,
      "learning_rate": 6.5e-07,
      "loss": 0.0422,
      "reward": 0.7531994581222534,
      "reward_std": 0.04625563230365515,
      "reward_std_all": 0.053613954689353704,
      "rewards-max/forensics_sinkhorn_reward": 0.988226056098938,
      "rewards-max/forensics_strict_edge_reward": 0.34087733179330826,
      "rewards-max/forensics_tiou_reward": 0.9733248949050903,
      "rewards/forensics_sinkhorn_reward": 0.7320694029331207,
      "rewards/forensics_strict_edge_reward": 0.02638943213969469,
      "rewards/forensics_tiou_reward": 0.5339876413345337,
      "step": 456
    },
    {
      "completion_length": 22.34375,
      "epoch": 1.402457757296467,
      "grad_norm": 6.473525047302246,
      "kl": 0.384765625,
      "learning_rate": 6.492307692307692e-07,
      "loss": 0.0684,
      "reward": 0.7213196456432343,
      "reward_std": 0.07231850363314152,
      "reward_std_all": 0.07905011437833309,
      "rewards-max/forensics_sinkhorn_reward": 0.9886152148246765,
      "rewards-max/forensics_strict_edge_reward": 0.7727466821670532,
      "rewards-max/forensics_tiou_reward": 0.9782621264457703,
      "rewards/forensics_sinkhorn_reward": 0.6368878185749054,
      "rewards/forensics_strict_edge_reward": 0.05427061393857002,
      "rewards/forensics_tiou_reward": 0.39390093088150024,
      "step": 457
    },
    {
      "completion_length": 24.203125,
      "epoch": 1.4055299539170507,
      "grad_norm": 6.832784175872803,
      "kl": 0.7060546875,
      "learning_rate": 6.484615384615383e-07,
      "loss": 0.1063,
      "reward": 0.7526852786540985,
      "reward_std": 0.07279185950756073,
      "reward_std_all": 0.0674385167658329,
      "rewards-max/forensics_sinkhorn_reward": 0.9481285214424133,
      "rewards-max/forensics_strict_edge_reward": 0.07601466123014688,
      "rewards-max/forensics_tiou_reward": 0.8730416595935822,
      "rewards/forensics_sinkhorn_reward": 0.6938562095165253,
      "rewards/forensics_strict_edge_reward": 0.005129215482156724,
      "rewards/forensics_tiou_reward": 0.4422088861465454,
      "step": 458
    },
    {
      "completion_length": 22.78125,
      "epoch": 1.4086021505376345,
      "grad_norm": 5.9237470626831055,
      "kl": 0.3017578125,
      "learning_rate": 6.476923076923077e-07,
      "loss": -0.0726,
      "reward": 0.7863687574863434,
      "reward_std": 0.02333119115792215,
      "reward_std_all": 0.02371957851573825,
      "rewards-max/forensics_sinkhorn_reward": 0.990992397069931,
      "rewards-max/forensics_strict_edge_reward": 0.709127277135849,
      "rewards-max/forensics_tiou_reward": 0.9829406142234802,
      "rewards/forensics_sinkhorn_reward": 0.74073925614357,
      "rewards/forensics_strict_edge_reward": 0.07385635748505592,
      "rewards/forensics_tiou_reward": 0.5656376481056213,
      "step": 459
    },
    {
      "completion_length": 23.015625,
      "epoch": 1.411674347158218,
      "grad_norm": 5.652225971221924,
      "kl": 0.3310546875,
      "learning_rate": 6.469230769230769e-07,
      "loss": -0.103,
      "reward": 0.7156454622745514,
      "reward_std": 0.04497197922319174,
      "reward_std_all": 0.0663740374147892,
      "rewards-max/forensics_sinkhorn_reward": 0.9856489896774292,
      "rewards-max/forensics_strict_edge_reward": 0.49053970724344254,
      "rewards-max/forensics_tiou_reward": 0.9661656022071838,
      "rewards/forensics_sinkhorn_reward": 0.6662809252738953,
      "rewards/forensics_strict_edge_reward": 0.044019794557243586,
      "rewards/forensics_tiou_reward": 0.4566083997488022,
      "step": 460
    },
    {
      "completion_length": 21.90625,
      "epoch": 1.4147465437788018,
      "grad_norm": 7.621007919311523,
      "kl": 0.3623046875,
      "learning_rate": 6.461538461538462e-07,
      "loss": 0.1097,
      "reward": 0.7626620531082153,
      "reward_std": 0.02044275077059865,
      "reward_std_all": 0.03402521042153239,
      "rewards-max/forensics_sinkhorn_reward": 0.9607881605625153,
      "rewards-max/forensics_strict_edge_reward": 0.203890360891819,
      "rewards-max/forensics_tiou_reward": 0.8920003175735474,
      "rewards/forensics_sinkhorn_reward": 0.6834744513034821,
      "rewards/forensics_strict_edge_reward": 0.012752112001180649,
      "rewards/forensics_tiou_reward": 0.39674825966358185,
      "step": 461
    },
    {
      "completion_length": 23.421875,
      "epoch": 1.4178187403993856,
      "grad_norm": 5.5477519035339355,
      "kl": 0.3271484375,
      "learning_rate": 6.453846153846153e-07,
      "loss": 0.0543,
      "reward": 0.7749505639076233,
      "reward_std": 0.03289838694036007,
      "reward_std_all": 0.03308613412082195,
      "rewards-max/forensics_sinkhorn_reward": 0.9828037619590759,
      "rewards-max/forensics_strict_edge_reward": 0.6071805953979492,
      "rewards-max/forensics_tiou_reward": 0.9631331861019135,
      "rewards/forensics_sinkhorn_reward": 0.7644824087619781,
      "rewards/forensics_strict_edge_reward": 0.058795446529984474,
      "rewards/forensics_tiou_reward": 0.5541540384292603,
      "step": 462
    },
    {
      "completion_length": 24.328125,
      "epoch": 1.4208909370199692,
      "grad_norm": 4.968278408050537,
      "kl": 0.3603515625,
      "learning_rate": 6.446153846153845e-07,
      "loss": 0.0237,
      "reward": 0.7366734147071838,
      "reward_std": 0.05587916262447834,
      "reward_std_all": 0.06529590487480164,
      "rewards-max/forensics_sinkhorn_reward": 0.9716747105121613,
      "rewards-max/forensics_strict_edge_reward": 0.06048707815352827,
      "rewards-max/forensics_tiou_reward": 0.9264061152935028,
      "rewards/forensics_sinkhorn_reward": 0.7110584378242493,
      "rewards/forensics_strict_edge_reward": 0.003067952664423501,
      "rewards/forensics_tiou_reward": 0.425017774105072,
      "step": 463
    },
    {
      "completion_length": 23.640625,
      "epoch": 1.423963133640553,
      "grad_norm": 5.383334159851074,
      "kl": 0.3984375,
      "learning_rate": 6.438461538461538e-07,
      "loss": 0.0125,
      "reward": 0.7407340705394745,
      "reward_std": 0.03906828071922064,
      "reward_std_all": 0.05124819651246071,
      "rewards-max/forensics_sinkhorn_reward": 0.9084722101688385,
      "rewards-max/forensics_strict_edge_reward": 0.19880328490398824,
      "rewards-max/forensics_tiou_reward": 0.7956196367740631,
      "rewards/forensics_sinkhorn_reward": 0.6939685344696045,
      "rewards/forensics_strict_edge_reward": 0.02493449884786969,
      "rewards/forensics_tiou_reward": 0.421219140291214,
      "step": 464
    },
    {
      "completion_length": 24.15625,
      "epoch": 1.4270353302611367,
      "grad_norm": 6.055786609649658,
      "kl": 0.4267578125,
      "learning_rate": 6.43076923076923e-07,
      "loss": 0.0153,
      "reward": 0.7678110003471375,
      "reward_std": 0.046713512390851974,
      "reward_std_all": 0.04491518810391426,
      "rewards-max/forensics_sinkhorn_reward": 0.9787508249282837,
      "rewards-max/forensics_strict_edge_reward": 0.19105863198637962,
      "rewards-max/forensics_tiou_reward": 0.9495946168899536,
      "rewards/forensics_sinkhorn_reward": 0.7722712755203247,
      "rewards/forensics_strict_edge_reward": 0.02157820761203766,
      "rewards/forensics_tiou_reward": 0.5604220032691956,
      "step": 465
    },
    {
      "completion_length": 24.625,
      "epoch": 1.4301075268817205,
      "grad_norm": 4.410170078277588,
      "kl": 0.349609375,
      "learning_rate": 6.423076923076924e-07,
      "loss": 0.0906,
      "reward": 0.7466566562652588,
      "reward_std": 0.06549638137221336,
      "reward_std_all": 0.06357143633067608,
      "rewards-max/forensics_sinkhorn_reward": 0.9578869342803955,
      "rewards-max/forensics_strict_edge_reward": 0.27228695154190063,
      "rewards-max/forensics_tiou_reward": 0.908174067735672,
      "rewards/forensics_sinkhorn_reward": 0.6301026046276093,
      "rewards/forensics_strict_edge_reward": 0.024121065624058247,
      "rewards/forensics_tiou_reward": 0.4681481570005417,
      "step": 466
    },
    {
      "completion_length": 24.328125,
      "epoch": 1.4331797235023043,
      "grad_norm": 8.673785209655762,
      "kl": 0.6474609375,
      "learning_rate": 6.415384615384615e-07,
      "loss": 0.0591,
      "reward": 0.7657935619354248,
      "reward_std": 0.05080403760075569,
      "reward_std_all": 0.04844868369400501,
      "rewards-max/forensics_sinkhorn_reward": 0.9768467843532562,
      "rewards-max/forensics_strict_edge_reward": 0.2693897099234164,
      "rewards-max/forensics_tiou_reward": 0.9225429594516754,
      "rewards/forensics_sinkhorn_reward": 0.7305530309677124,
      "rewards/forensics_strict_edge_reward": 0.024642994423629716,
      "rewards/forensics_tiou_reward": 0.4149257689714432,
      "step": 467
    },
    {
      "completion_length": 22.75,
      "epoch": 1.4362519201228878,
      "grad_norm": 5.826140880584717,
      "kl": 0.3779296875,
      "learning_rate": 6.407692307692307e-07,
      "loss": 0.0585,
      "reward": 0.7439437806606293,
      "reward_std": 0.05343588348478079,
      "reward_std_all": 0.057426790706813335,
      "rewards-max/forensics_sinkhorn_reward": 0.9658636152744293,
      "rewards-max/forensics_strict_edge_reward": 0.30339228361845016,
      "rewards-max/forensics_tiou_reward": 0.8978296220302582,
      "rewards/forensics_sinkhorn_reward": 0.6519067585468292,
      "rewards/forensics_strict_edge_reward": 0.020856795832514763,
      "rewards/forensics_tiou_reward": 0.3663366287946701,
      "step": 468
    },
    {
      "completion_length": 25.953125,
      "epoch": 1.4393241167434716,
      "grad_norm": 5.476456165313721,
      "kl": 0.34228515625,
      "learning_rate": 6.4e-07,
      "loss": 0.0505,
      "reward": 0.7391147017478943,
      "reward_std": 0.06763632223010063,
      "reward_std_all": 0.06881498917937279,
      "rewards-max/forensics_sinkhorn_reward": 0.9675306677818298,
      "rewards-max/forensics_strict_edge_reward": 0.14024515077471733,
      "rewards-max/forensics_tiou_reward": 0.9148525595664978,
      "rewards/forensics_sinkhorn_reward": 0.6677974462509155,
      "rewards/forensics_strict_edge_reward": 0.008473401423543692,
      "rewards/forensics_tiou_reward": 0.41880078613758087,
      "step": 469
    },
    {
      "completion_length": 22.0625,
      "epoch": 1.4423963133640554,
      "grad_norm": 5.191492557525635,
      "kl": 0.3818359375,
      "learning_rate": 6.392307692307692e-07,
      "loss": -0.0262,
      "reward": 0.7779964804649353,
      "reward_std": 0.04649990564212203,
      "reward_std_all": 0.0423838971182704,
      "rewards-max/forensics_sinkhorn_reward": 0.983136773109436,
      "rewards-max/forensics_strict_edge_reward": 0.4011317938566208,
      "rewards-max/forensics_tiou_reward": 0.96316859126091,
      "rewards/forensics_sinkhorn_reward": 0.7561778426170349,
      "rewards/forensics_strict_edge_reward": 0.03170734830200672,
      "rewards/forensics_tiou_reward": 0.5447236001491547,
      "step": 470
    },
    {
      "completion_length": 24.125,
      "epoch": 1.445468509984639,
      "grad_norm": 5.790704250335693,
      "kl": 0.3916015625,
      "learning_rate": 6.384615384615383e-07,
      "loss": 0.0056,
      "reward": 0.7495523393154144,
      "reward_std": 0.05165110342204571,
      "reward_std_all": 0.05477366782724857,
      "rewards-max/forensics_sinkhorn_reward": 0.9745189547538757,
      "rewards-max/forensics_strict_edge_reward": 0.4811747968196869,
      "rewards-max/forensics_tiou_reward": 0.938509613275528,
      "rewards/forensics_sinkhorn_reward": 0.7082239985466003,
      "rewards/forensics_strict_edge_reward": 0.03617037180811167,
      "rewards/forensics_tiou_reward": 0.4991879016160965,
      "step": 471
    },
    {
      "completion_length": 24.015625,
      "epoch": 1.4485407066052227,
      "grad_norm": 5.6470818519592285,
      "kl": 0.4443359375,
      "learning_rate": 6.376923076923077e-07,
      "loss": 0.1263,
      "reward": 0.76766237616539,
      "reward_std": 0.04264259245246649,
      "reward_std_all": 0.042023150250315666,
      "rewards-max/forensics_sinkhorn_reward": 0.9867743849754333,
      "rewards-max/forensics_strict_edge_reward": 0.40798234194517136,
      "rewards-max/forensics_tiou_reward": 0.9725258648395538,
      "rewards/forensics_sinkhorn_reward": 0.7298482358455658,
      "rewards/forensics_strict_edge_reward": 0.03677886072546244,
      "rewards/forensics_tiou_reward": 0.5016728788614273,
      "step": 472
    },
    {
      "completion_length": 24.625,
      "epoch": 1.4516129032258065,
      "grad_norm": 6.724281311035156,
      "kl": 0.388671875,
      "learning_rate": 6.369230769230769e-07,
      "loss": 0.046,
      "reward": 0.7375901341438293,
      "reward_std": 0.03610654268413782,
      "reward_std_all": 0.05123539175838232,
      "rewards-max/forensics_sinkhorn_reward": 0.9799267947673798,
      "rewards-max/forensics_strict_edge_reward": 0.6044360399246216,
      "rewards-max/forensics_tiou_reward": 0.9480696022510529,
      "rewards/forensics_sinkhorn_reward": 0.6965153813362122,
      "rewards/forensics_strict_edge_reward": 0.0398484542965889,
      "rewards/forensics_tiou_reward": 0.4445268511772156,
      "step": 473
    },
    {
      "completion_length": 22.71875,
      "epoch": 1.4546850998463903,
      "grad_norm": 5.787184715270996,
      "kl": 0.373046875,
      "learning_rate": 6.361538461538462e-07,
      "loss": 0.0313,
      "reward": 0.7372210025787354,
      "reward_std": 0.04923516768030822,
      "reward_std_all": 0.057238176465034485,
      "rewards-max/forensics_sinkhorn_reward": 0.961982250213623,
      "rewards-max/forensics_strict_edge_reward": 0.23651578649878502,
      "rewards-max/forensics_tiou_reward": 0.915850967168808,
      "rewards/forensics_sinkhorn_reward": 0.7183196544647217,
      "rewards/forensics_strict_edge_reward": 0.02132470381911844,
      "rewards/forensics_tiou_reward": 0.48460109531879425,
      "step": 474
    },
    {
      "completion_length": 25.5,
      "epoch": 1.4577572964669738,
      "grad_norm": 4.8541107177734375,
      "kl": 0.287109375,
      "learning_rate": 6.353846153846154e-07,
      "loss": 0.1194,
      "reward": 0.7616351246833801,
      "reward_std": 0.05027003679424524,
      "reward_std_all": 0.04915426252409816,
      "rewards-max/forensics_sinkhorn_reward": 0.9848872721195221,
      "rewards-max/forensics_strict_edge_reward": 0.6207633018493652,
      "rewards-max/forensics_tiou_reward": 0.9674331843852997,
      "rewards/forensics_sinkhorn_reward": 0.689924955368042,
      "rewards/forensics_strict_edge_reward": 0.06848649866878986,
      "rewards/forensics_tiou_reward": 0.4912414252758026,
      "step": 475
    },
    {
      "completion_length": 23.03125,
      "epoch": 1.4608294930875576,
      "grad_norm": 5.647349834442139,
      "kl": 0.3681640625,
      "learning_rate": 6.346153846153845e-07,
      "loss": 0.0004,
      "reward": 0.7703320384025574,
      "reward_std": 0.050694664940238,
      "reward_std_all": 0.047416018322110176,
      "rewards-max/forensics_sinkhorn_reward": 0.9972147941589355,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9988430142402649,
      "rewards/forensics_sinkhorn_reward": 0.7398864328861237,
      "rewards/forensics_strict_edge_reward": 0.09272958151996136,
      "rewards/forensics_tiou_reward": 0.5167163610458374,
      "step": 476
    },
    {
      "completion_length": 24.4375,
      "epoch": 1.4639016897081414,
      "grad_norm": 5.129894256591797,
      "kl": 0.2958984375,
      "learning_rate": 6.338461538461538e-07,
      "loss": -0.0279,
      "reward": 0.7625187635421753,
      "reward_std": 0.046813043765723705,
      "reward_std_all": 0.04660285171121359,
      "rewards-max/forensics_sinkhorn_reward": 0.9655071496963501,
      "rewards-max/forensics_strict_edge_reward": 0.15495047345757484,
      "rewards-max/forensics_tiou_reward": 0.9104912877082825,
      "rewards/forensics_sinkhorn_reward": 0.6963173747062683,
      "rewards/forensics_strict_edge_reward": 0.011973177781328559,
      "rewards/forensics_tiou_reward": 0.506735235452652,
      "step": 477
    },
    {
      "completion_length": 23.421875,
      "epoch": 1.466973886328725,
      "grad_norm": 5.277091026306152,
      "kl": 0.3984375,
      "learning_rate": 6.33076923076923e-07,
      "loss": 0.0747,
      "reward": 0.7757282257080078,
      "reward_std": 0.04353328421711922,
      "reward_std_all": 0.0406325226649642,
      "rewards-max/forensics_sinkhorn_reward": 0.9855459034442902,
      "rewards-max/forensics_strict_edge_reward": 0.34673839807510376,
      "rewards-max/forensics_tiou_reward": 0.9693068861961365,
      "rewards/forensics_sinkhorn_reward": 0.7187148034572601,
      "rewards/forensics_strict_edge_reward": 0.02559111826121807,
      "rewards/forensics_tiou_reward": 0.5232216417789459,
      "step": 478
    },
    {
      "completion_length": 24.515625,
      "epoch": 1.4700460829493087,
      "grad_norm": 4.847842216491699,
      "kl": 0.4140625,
      "learning_rate": 6.323076923076924e-07,
      "loss": 0.1301,
      "reward": 0.7674543559551239,
      "reward_std": 0.062341656535863876,
      "reward_std_all": 0.05670229718089104,
      "rewards-max/forensics_sinkhorn_reward": 0.970233142375946,
      "rewards-max/forensics_strict_edge_reward": 0.2665891721844673,
      "rewards-max/forensics_tiou_reward": 0.9245217144489288,
      "rewards/forensics_sinkhorn_reward": 0.7040466070175171,
      "rewards/forensics_strict_edge_reward": 0.027659681625664234,
      "rewards/forensics_tiou_reward": 0.45817890763282776,
      "step": 479
    },
    {
      "completion_length": 22.375,
      "epoch": 1.4731182795698925,
      "grad_norm": 5.4057087898254395,
      "kl": 0.3046875,
      "learning_rate": 6.315384615384615e-07,
      "loss": 0.1138,
      "reward": 0.7174850106239319,
      "reward_std": 0.0626195352524519,
      "reward_std_all": 0.07554028555750847,
      "rewards-max/forensics_sinkhorn_reward": 0.9805436432361603,
      "rewards-max/forensics_strict_edge_reward": 0.5039256066083908,
      "rewards-max/forensics_tiou_reward": 0.9558655619621277,
      "rewards/forensics_sinkhorn_reward": 0.6269932687282562,
      "rewards/forensics_strict_edge_reward": 0.04784025624394417,
      "rewards/forensics_tiou_reward": 0.4163960963487625,
      "step": 480
    },
    {
      "completion_length": 29.234375,
      "epoch": 1.4761904761904763,
      "grad_norm": 4.758365631103516,
      "kl": 0.2568359375,
      "learning_rate": 6.307692307692307e-07,
      "loss": 0.1494,
      "reward": 0.7720257639884949,
      "reward_std": 0.05150647182017565,
      "reward_std_all": 0.04732584021985531,
      "rewards-max/forensics_sinkhorn_reward": 0.9547777771949768,
      "rewards-max/forensics_strict_edge_reward": 0.3392348736524582,
      "rewards-max/forensics_tiou_reward": 0.8812384307384491,
      "rewards/forensics_sinkhorn_reward": 0.6866873204708099,
      "rewards/forensics_strict_edge_reward": 0.01707225665450096,
      "rewards/forensics_tiou_reward": 0.5234439671039581,
      "step": 481
    },
    {
      "completion_length": 25.5625,
      "epoch": 1.4792626728110598,
      "grad_norm": 4.7820940017700195,
      "kl": 0.337890625,
      "learning_rate": 6.3e-07,
      "loss": 0.0871,
      "reward": 0.7085703313350677,
      "reward_std": 0.07549523934721947,
      "reward_std_all": 0.08687299117445946,
      "rewards-max/forensics_sinkhorn_reward": 0.9739162623882294,
      "rewards-max/forensics_strict_edge_reward": 0.33664843440055847,
      "rewards-max/forensics_tiou_reward": 0.9428787529468536,
      "rewards/forensics_sinkhorn_reward": 0.6688393354415894,
      "rewards/forensics_strict_edge_reward": 0.024511702358722687,
      "rewards/forensics_tiou_reward": 0.4321751892566681,
      "step": 482
    },
    {
      "completion_length": 26.875,
      "epoch": 1.4823348694316436,
      "grad_norm": 5.8836750984191895,
      "kl": 0.337890625,
      "learning_rate": 6.292307692307692e-07,
      "loss": 0.052,
      "reward": 0.7219775915145874,
      "reward_std": 0.06295214779675007,
      "reward_std_all": 0.07511171326041222,
      "rewards-max/forensics_sinkhorn_reward": 0.9899377822875977,
      "rewards-max/forensics_strict_edge_reward": 0.3877888023853302,
      "rewards-max/forensics_tiou_reward": 0.9764019548892975,
      "rewards/forensics_sinkhorn_reward": 0.6570526957511902,
      "rewards/forensics_strict_edge_reward": 0.025715546682476997,
      "rewards/forensics_tiou_reward": 0.46164190769195557,
      "step": 483
    },
    {
      "completion_length": 25.90625,
      "epoch": 1.4854070660522274,
      "grad_norm": 5.607974529266357,
      "kl": 0.322265625,
      "learning_rate": 6.284615384615385e-07,
      "loss": 0.1025,
      "reward": 0.7391360700130463,
      "reward_std": 0.0540450606495142,
      "reward_std_all": 0.061928244307637215,
      "rewards-max/forensics_sinkhorn_reward": 0.9757338762283325,
      "rewards-max/forensics_strict_edge_reward": 0.22491609677672386,
      "rewards-max/forensics_tiou_reward": 0.9442831575870514,
      "rewards/forensics_sinkhorn_reward": 0.6857025623321533,
      "rewards/forensics_strict_edge_reward": 0.025037997867912054,
      "rewards/forensics_tiou_reward": 0.4616246372461319,
      "step": 484
    },
    {
      "completion_length": 27.78125,
      "epoch": 1.488479262672811,
      "grad_norm": 5.336965084075928,
      "kl": 0.291015625,
      "learning_rate": 6.276923076923076e-07,
      "loss": 0.0501,
      "reward": 0.7685728073120117,
      "reward_std": 0.060063157230615616,
      "reward_std_all": 0.0543248001486063,
      "rewards-max/forensics_sinkhorn_reward": 0.9718711674213409,
      "rewards-max/forensics_strict_edge_reward": 0.49431248009204865,
      "rewards-max/forensics_tiou_reward": 0.9341617822647095,
      "rewards/forensics_sinkhorn_reward": 0.7078658044338226,
      "rewards/forensics_strict_edge_reward": 0.046207211911678314,
      "rewards/forensics_tiou_reward": 0.5240099430084229,
      "step": 485
    },
    {
      "completion_length": 27.0625,
      "epoch": 1.4915514592933947,
      "grad_norm": 7.087733268737793,
      "kl": 0.2724609375,
      "learning_rate": 6.269230769230769e-07,
      "loss": 0.0454,
      "reward": 0.6866174936294556,
      "reward_std": 0.05988083407282829,
      "reward_std_all": 0.0848005935549736,
      "rewards-max/forensics_sinkhorn_reward": 0.9299691021442413,
      "rewards-max/forensics_strict_edge_reward": 0.5254523754119873,
      "rewards-max/forensics_tiou_reward": 0.8445079922676086,
      "rewards/forensics_sinkhorn_reward": 0.6199827492237091,
      "rewards/forensics_strict_edge_reward": 0.044741231482475996,
      "rewards/forensics_tiou_reward": 0.400741383433342,
      "step": 486
    },
    {
      "completion_length": 24.84375,
      "epoch": 1.4946236559139785,
      "grad_norm": 5.859115123748779,
      "kl": 0.5380859375,
      "learning_rate": 6.261538461538462e-07,
      "loss": 0.0596,
      "reward": 0.7511023879051208,
      "reward_std": 0.05267020035535097,
      "reward_std_all": 0.05420337803661823,
      "rewards-max/forensics_sinkhorn_reward": 0.9840724468231201,
      "rewards-max/forensics_strict_edge_reward": 0.6764304339885712,
      "rewards-max/forensics_tiou_reward": 0.9619944393634796,
      "rewards/forensics_sinkhorn_reward": 0.7349823713302612,
      "rewards/forensics_strict_edge_reward": 0.042579823173582554,
      "rewards/forensics_tiou_reward": 0.48594042658805847,
      "step": 487
    },
    {
      "completion_length": 25.265625,
      "epoch": 1.4976958525345623,
      "grad_norm": 4.782414436340332,
      "kl": 0.3203125,
      "learning_rate": 6.253846153846154e-07,
      "loss": 0.1045,
      "reward": 0.7853178381919861,
      "reward_std": 0.03248381428420544,
      "reward_std_all": 0.030567852780222893,
      "rewards-max/forensics_sinkhorn_reward": 0.9545325040817261,
      "rewards-max/forensics_strict_edge_reward": 0.20663544186390936,
      "rewards-max/forensics_tiou_reward": 0.8827106356620789,
      "rewards/forensics_sinkhorn_reward": 0.7216943502426147,
      "rewards/forensics_strict_edge_reward": 0.022705193201545626,
      "rewards/forensics_tiou_reward": 0.5260997414588928,
      "step": 488
    },
    {
      "completion_length": 28.3125,
      "epoch": 1.500768049155146,
      "grad_norm": 5.237250328063965,
      "kl": 0.2978515625,
      "learning_rate": 6.246153846153845e-07,
      "loss": 0.0002,
      "reward": 0.7412282824516296,
      "reward_std": 0.026864961721003056,
      "reward_std_all": 0.04511724226176739,
      "rewards-max/forensics_sinkhorn_reward": 0.9660350978374481,
      "rewards-max/forensics_strict_edge_reward": 0.3039274625480175,
      "rewards-max/forensics_tiou_reward": 0.9042383134365082,
      "rewards/forensics_sinkhorn_reward": 0.6589484810829163,
      "rewards/forensics_strict_edge_reward": 0.022213146090507507,
      "rewards/forensics_tiou_reward": 0.45292797684669495,
      "step": 489
    },
    {
      "completion_length": 25.125,
      "epoch": 1.5038402457757296,
      "grad_norm": 5.202218055725098,
      "kl": 0.3349609375,
      "learning_rate": 6.238461538461538e-07,
      "loss": 0.0362,
      "reward": 0.7649816572666168,
      "reward_std": 0.04516445193439722,
      "reward_std_all": 0.04523395374417305,
      "rewards-max/forensics_sinkhorn_reward": 0.9916036128997803,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9849211275577545,
      "rewards/forensics_sinkhorn_reward": 0.6923314332962036,
      "rewards/forensics_strict_edge_reward": 0.05082487873733044,
      "rewards/forensics_tiou_reward": 0.5255394577980042,
      "step": 490
    },
    {
      "completion_length": 23.265625,
      "epoch": 1.5069124423963134,
      "grad_norm": 5.821855545043945,
      "kl": 0.3466796875,
      "learning_rate": 6.23076923076923e-07,
      "loss": 0.0752,
      "reward": 0.7343665659427643,
      "reward_std": 0.07528031244874,
      "reward_std_all": 0.07442642003297806,
      "rewards-max/forensics_sinkhorn_reward": 0.9947956204414368,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9935774505138397,
      "rewards/forensics_sinkhorn_reward": 0.6707986891269684,
      "rewards/forensics_strict_edge_reward": 0.09814915433526039,
      "rewards/forensics_tiou_reward": 0.5116869211196899,
      "step": 491
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.509984639016897,
      "grad_norm": 5.367684364318848,
      "kl": 0.453125,
      "learning_rate": 6.223076923076924e-07,
      "loss": 0.055,
      "reward": 0.7455600202083588,
      "reward_std": 0.017930909292772412,
      "reward_std_all": 0.040474945679306984,
      "rewards-max/forensics_sinkhorn_reward": 0.9887763261795044,
      "rewards-max/forensics_strict_edge_reward": 0.670350581407547,
      "rewards-max/forensics_tiou_reward": 0.9737900197505951,
      "rewards/forensics_sinkhorn_reward": 0.6933572590351105,
      "rewards/forensics_strict_edge_reward": 0.04355747066438198,
      "rewards/forensics_tiou_reward": 0.49899590015411377,
      "step": 492
    },
    {
      "completion_length": 23.75,
      "epoch": 1.5130568356374807,
      "grad_norm": 5.31792688369751,
      "kl": 0.4228515625,
      "learning_rate": 6.215384615384615e-07,
      "loss": 0.1151,
      "reward": 0.7725540101528168,
      "reward_std": 0.04314971249550581,
      "reward_std_all": 0.041099742986261845,
      "rewards-max/forensics_sinkhorn_reward": 0.9775241911411285,
      "rewards-max/forensics_strict_edge_reward": 0.29305245727300644,
      "rewards-max/forensics_tiou_reward": 0.9346270263195038,
      "rewards/forensics_sinkhorn_reward": 0.7624415755271912,
      "rewards/forensics_strict_edge_reward": 0.021886086091399193,
      "rewards/forensics_tiou_reward": 0.5744388401508331,
      "step": 493
    },
    {
      "completion_length": 26.0,
      "epoch": 1.5161290322580645,
      "grad_norm": 3.9948291778564453,
      "kl": 0.26953125,
      "learning_rate": 6.207692307692307e-07,
      "loss": 0.0681,
      "reward": 0.7692326903343201,
      "reward_std": 0.048346202820539474,
      "reward_std_all": 0.04575980640947819,
      "rewards-max/forensics_sinkhorn_reward": 0.9887985289096832,
      "rewards-max/forensics_strict_edge_reward": 0.40175478160381317,
      "rewards-max/forensics_tiou_reward": 0.974949985742569,
      "rewards/forensics_sinkhorn_reward": 0.7530207335948944,
      "rewards/forensics_strict_edge_reward": 0.04297189973294735,
      "rewards/forensics_tiou_reward": 0.6094086766242981,
      "step": 494
    },
    {
      "completion_length": 22.484375,
      "epoch": 1.5192012288786483,
      "grad_norm": 5.75604248046875,
      "kl": 0.3193359375,
      "learning_rate": 6.2e-07,
      "loss": 0.0037,
      "reward": 0.7929360270500183,
      "reward_std": 0.01347494451329112,
      "reward_std_all": 0.01615685923025012,
      "rewards-max/forensics_sinkhorn_reward": 0.9842960238456726,
      "rewards-max/forensics_strict_edge_reward": 0.3021759018301964,
      "rewards-max/forensics_tiou_reward": 0.9594297111034393,
      "rewards/forensics_sinkhorn_reward": 0.7926268875598907,
      "rewards/forensics_strict_edge_reward": 0.042660473845899105,
      "rewards/forensics_tiou_reward": 0.5325884819030762,
      "step": 495
    },
    {
      "completion_length": 22.0625,
      "epoch": 1.522273425499232,
      "grad_norm": 7.0159502029418945,
      "kl": 0.541015625,
      "learning_rate": 6.192307692307692e-07,
      "loss": 0.0634,
      "reward": 0.7622340321540833,
      "reward_std": 0.03883867198601365,
      "reward_std_all": 0.04332081787288189,
      "rewards-max/forensics_sinkhorn_reward": 0.970670074224472,
      "rewards-max/forensics_strict_edge_reward": 0.2368861734867096,
      "rewards-max/forensics_tiou_reward": 0.9291049838066101,
      "rewards/forensics_sinkhorn_reward": 0.7126157581806183,
      "rewards/forensics_strict_edge_reward": 0.020801409147679806,
      "rewards/forensics_tiou_reward": 0.41667069494724274,
      "step": 496
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.5253456221198156,
      "grad_norm": 6.61160135269165,
      "kl": 0.3876953125,
      "learning_rate": 6.184615384615385e-07,
      "loss": 0.1451,
      "reward": 0.7422938048839569,
      "reward_std": 0.032244210597127676,
      "reward_std_all": 0.048088790848851204,
      "rewards-max/forensics_sinkhorn_reward": 0.9544742703437805,
      "rewards-max/forensics_strict_edge_reward": 0.27026205509901047,
      "rewards-max/forensics_tiou_reward": 0.8837685883045197,
      "rewards/forensics_sinkhorn_reward": 0.6341668963432312,
      "rewards/forensics_strict_edge_reward": 0.014455010183155537,
      "rewards/forensics_tiou_reward": 0.3670383393764496,
      "step": 497
    },
    {
      "completion_length": 25.21875,
      "epoch": 1.5284178187403994,
      "grad_norm": 6.752192974090576,
      "kl": 0.38671875,
      "learning_rate": 6.176923076923076e-07,
      "loss": -0.0715,
      "reward": 0.7647612690925598,
      "reward_std": 0.04771857429295778,
      "reward_std_all": 0.046495333313941956,
      "rewards-max/forensics_sinkhorn_reward": 0.9758350253105164,
      "rewards-max/forensics_strict_edge_reward": 0.3731641471385956,
      "rewards-max/forensics_tiou_reward": 0.9420734643936157,
      "rewards/forensics_sinkhorn_reward": 0.6860641241073608,
      "rewards/forensics_strict_edge_reward": 0.03733462281525135,
      "rewards/forensics_tiou_reward": 0.49155133962631226,
      "step": 498
    },
    {
      "completion_length": 23.828125,
      "epoch": 1.531490015360983,
      "grad_norm": 7.627175807952881,
      "kl": 0.40234375,
      "learning_rate": 6.169230769230768e-07,
      "loss": -0.0759,
      "reward": 0.7626542747020721,
      "reward_std": 0.054196981713175774,
      "reward_std_all": 0.0515192411839962,
      "rewards-max/forensics_sinkhorn_reward": 0.9223528206348419,
      "rewards-max/forensics_strict_edge_reward": 0.5721771121025085,
      "rewards-max/forensics_tiou_reward": 0.8197559714317322,
      "rewards/forensics_sinkhorn_reward": 0.6722472608089447,
      "rewards/forensics_strict_edge_reward": 0.05913990642875433,
      "rewards/forensics_tiou_reward": 0.4284801781177521,
      "step": 499
    },
    {
      "completion_length": 27.921875,
      "epoch": 1.5345622119815667,
      "grad_norm": 6.396727561950684,
      "kl": 0.3876953125,
      "learning_rate": 6.161538461538462e-07,
      "loss": 0.0952,
      "reward": 0.7017648816108704,
      "reward_std": 0.0629992913454771,
      "reward_std_all": 0.08426434732973576,
      "rewards-max/forensics_sinkhorn_reward": 0.9689740538597107,
      "rewards-max/forensics_strict_edge_reward": 0.37226127088069916,
      "rewards-max/forensics_tiou_reward": 0.9321928918361664,
      "rewards/forensics_sinkhorn_reward": 0.6044780313968658,
      "rewards/forensics_strict_edge_reward": 0.028021499048918486,
      "rewards/forensics_tiou_reward": 0.3938656449317932,
      "step": 500
    },
    {
      "completion_length": 26.234375,
      "epoch": 1.5376344086021505,
      "grad_norm": 5.8347554206848145,
      "kl": 0.2822265625,
      "learning_rate": 6.153846153846154e-07,
      "loss": 0.0382,
      "reward": 0.7310482859611511,
      "reward_std": 0.0459415428340435,
      "reward_std_all": 0.06446696072816849,
      "rewards-max/forensics_sinkhorn_reward": 0.9776210188865662,
      "rewards-max/forensics_strict_edge_reward": 0.5247012078762054,
      "rewards-max/forensics_tiou_reward": 0.9457184672355652,
      "rewards/forensics_sinkhorn_reward": 0.6272077560424805,
      "rewards/forensics_strict_edge_reward": 0.030849642120301723,
      "rewards/forensics_tiou_reward": 0.4554052799940109,
      "step": 501
    },
    {
      "completion_length": 28.984375,
      "epoch": 1.5407066052227343,
      "grad_norm": 5.542784690856934,
      "kl": 0.322265625,
      "learning_rate": 6.146153846153845e-07,
      "loss": 0.131,
      "reward": 0.7159314155578613,
      "reward_std": 0.058964768424630165,
      "reward_std_all": 0.0734699796885252,
      "rewards-max/forensics_sinkhorn_reward": 0.9781868755817413,
      "rewards-max/forensics_strict_edge_reward": 0.5148857682943344,
      "rewards-max/forensics_tiou_reward": 0.94474858045578,
      "rewards/forensics_sinkhorn_reward": 0.6574661731719971,
      "rewards/forensics_strict_edge_reward": 0.03657903429120779,
      "rewards/forensics_tiou_reward": 0.4585220515727997,
      "step": 502
    },
    {
      "completion_length": 28.640625,
      "epoch": 1.543778801843318,
      "grad_norm": 5.707304954528809,
      "kl": 0.4482421875,
      "learning_rate": 6.138461538461538e-07,
      "loss": 0.1108,
      "reward": 0.7540697157382965,
      "reward_std": 0.051276146434247494,
      "reward_std_all": 0.05183990299701691,
      "rewards-max/forensics_sinkhorn_reward": 0.9888112246990204,
      "rewards-max/forensics_strict_edge_reward": 0.27868296205997467,
      "rewards-max/forensics_tiou_reward": 0.9751269817352295,
      "rewards/forensics_sinkhorn_reward": 0.7309146821498871,
      "rewards/forensics_strict_edge_reward": 0.029685073532164097,
      "rewards/forensics_tiou_reward": 0.5541911125183105,
      "step": 503
    },
    {
      "completion_length": 26.078125,
      "epoch": 1.5468509984639018,
      "grad_norm": 6.49441385269165,
      "kl": 0.3115234375,
      "learning_rate": 6.13076923076923e-07,
      "loss": -0.0399,
      "reward": 0.7821120321750641,
      "reward_std": 0.03276163898408413,
      "reward_std_all": 0.03140031825751066,
      "rewards-max/forensics_sinkhorn_reward": 0.9777220189571381,
      "rewards-max/forensics_strict_edge_reward": 0.4419858157634735,
      "rewards-max/forensics_tiou_reward": 0.9501025378704071,
      "rewards/forensics_sinkhorn_reward": 0.7264493703842163,
      "rewards/forensics_strict_edge_reward": 0.03922393545508385,
      "rewards/forensics_tiou_reward": 0.5209764838218689,
      "step": 504
    },
    {
      "completion_length": 22.140625,
      "epoch": 1.5499231950844854,
      "grad_norm": 5.672663688659668,
      "kl": 0.30078125,
      "learning_rate": 6.123076923076923e-07,
      "loss": -0.0128,
      "reward": 0.7817338109016418,
      "reward_std": 0.031168093904852867,
      "reward_std_all": 0.030940987169742584,
      "rewards-max/forensics_sinkhorn_reward": 0.9810327291488647,
      "rewards-max/forensics_strict_edge_reward": 0.6154066026210785,
      "rewards-max/forensics_tiou_reward": 0.9602660834789276,
      "rewards/forensics_sinkhorn_reward": 0.7295922636985779,
      "rewards/forensics_strict_edge_reward": 0.04564382787793875,
      "rewards/forensics_tiou_reward": 0.4378209412097931,
      "step": 505
    },
    {
      "completion_length": 26.59375,
      "epoch": 1.5529953917050692,
      "grad_norm": 6.360027313232422,
      "kl": 0.3115234375,
      "learning_rate": 6.115384615384616e-07,
      "loss": 0.1161,
      "reward": 0.747561901807785,
      "reward_std": 0.05771556869149208,
      "reward_std_all": 0.057420650497078896,
      "rewards-max/forensics_sinkhorn_reward": 0.9778644144535065,
      "rewards-max/forensics_strict_edge_reward": 0.4263888746500015,
      "rewards-max/forensics_tiou_reward": 0.947879821062088,
      "rewards/forensics_sinkhorn_reward": 0.6557715535163879,
      "rewards/forensics_strict_edge_reward": 0.040302447974681854,
      "rewards/forensics_tiou_reward": 0.49420759081840515,
      "step": 506
    },
    {
      "completion_length": 27.265625,
      "epoch": 1.5560675883256527,
      "grad_norm": 6.336153507232666,
      "kl": 0.38671875,
      "learning_rate": 6.107692307692307e-07,
      "loss": 0.0713,
      "reward": 0.7330037653446198,
      "reward_std": 0.07321099191904068,
      "reward_std_all": 0.07252153381705284,
      "rewards-max/forensics_sinkhorn_reward": 0.979266494512558,
      "rewards-max/forensics_strict_edge_reward": 0.23148634284734726,
      "rewards-max/forensics_tiou_reward": 0.9425849318504333,
      "rewards/forensics_sinkhorn_reward": 0.6299214065074921,
      "rewards/forensics_strict_edge_reward": 0.01449205121025443,
      "rewards/forensics_tiou_reward": 0.444751113653183,
      "step": 507
    },
    {
      "completion_length": 29.625,
      "epoch": 1.5591397849462365,
      "grad_norm": 5.075824737548828,
      "kl": 0.3408203125,
      "learning_rate": 6.1e-07,
      "loss": 0.0642,
      "reward": 0.7702977061271667,
      "reward_std": 0.021054735407233238,
      "reward_std_all": 0.02924925461411476,
      "rewards-max/forensics_sinkhorn_reward": 0.9766325652599335,
      "rewards-max/forensics_strict_edge_reward": 0.31579052400775254,
      "rewards-max/forensics_tiou_reward": 0.9475196301937103,
      "rewards/forensics_sinkhorn_reward": 0.7023525834083557,
      "rewards/forensics_strict_edge_reward": 0.03298467554850504,
      "rewards/forensics_tiou_reward": 0.4720058739185333,
      "step": 508
    },
    {
      "completion_length": 23.65625,
      "epoch": 1.5622119815668203,
      "grad_norm": 6.705537796020508,
      "kl": 0.3759765625,
      "learning_rate": 6.092307692307692e-07,
      "loss": 0.0319,
      "reward": 0.7632612586021423,
      "reward_std": 0.06321843340992928,
      "reward_std_all": 0.058000655844807625,
      "rewards-max/forensics_sinkhorn_reward": 0.9818405508995056,
      "rewards-max/forensics_strict_edge_reward": 0.12542079389095306,
      "rewards-max/forensics_tiou_reward": 0.9556188881397247,
      "rewards/forensics_sinkhorn_reward": 0.7274759113788605,
      "rewards/forensics_strict_edge_reward": 0.01427210122346878,
      "rewards/forensics_tiou_reward": 0.5268788933753967,
      "step": 509
    },
    {
      "completion_length": 23.453125,
      "epoch": 1.565284178187404,
      "grad_norm": 5.667980194091797,
      "kl": 0.3095703125,
      "learning_rate": 6.084615384615385e-07,
      "loss": 0.0768,
      "reward": 0.7538034319877625,
      "reward_std": 0.05473666824400425,
      "reward_std_all": 0.05534621886909008,
      "rewards-max/forensics_sinkhorn_reward": 0.974674791097641,
      "rewards-max/forensics_strict_edge_reward": 0.5173898935317993,
      "rewards-max/forensics_tiou_reward": 0.9438616335391998,
      "rewards/forensics_sinkhorn_reward": 0.6909142136573792,
      "rewards/forensics_strict_edge_reward": 0.042152006179094315,
      "rewards/forensics_tiou_reward": 0.47593314945697784,
      "step": 510
    },
    {
      "completion_length": 23.734375,
      "epoch": 1.5683563748079878,
      "grad_norm": 5.103003978729248,
      "kl": 0.37255859375,
      "learning_rate": 6.076923076923076e-07,
      "loss": 0.0382,
      "reward": 0.7842536270618439,
      "reward_std": 0.023862932051997632,
      "reward_std_all": 0.024767748778685927,
      "rewards-max/forensics_sinkhorn_reward": 0.9807502031326294,
      "rewards-max/forensics_strict_edge_reward": 0.3376298248767853,
      "rewards-max/forensics_tiou_reward": 0.9554172158241272,
      "rewards/forensics_sinkhorn_reward": 0.7548202276229858,
      "rewards/forensics_strict_edge_reward": 0.02828570269048214,
      "rewards/forensics_tiou_reward": 0.502682164311409,
      "step": 511
    },
    {
      "completion_length": 23.40625,
      "epoch": 1.5714285714285714,
      "grad_norm": 4.620031356811523,
      "kl": 0.3125,
      "learning_rate": 6.069230769230768e-07,
      "loss": 0.1067,
      "reward": 0.7560073137283325,
      "reward_std": 0.03635048307478428,
      "reward_std_all": 0.04343242570757866,
      "rewards-max/forensics_sinkhorn_reward": 0.9812220335006714,
      "rewards-max/forensics_strict_edge_reward": 0.42862197756767273,
      "rewards-max/forensics_tiou_reward": 0.958309143781662,
      "rewards/forensics_sinkhorn_reward": 0.6674568951129913,
      "rewards/forensics_strict_edge_reward": 0.032298317179083824,
      "rewards/forensics_tiou_reward": 0.44928833842277527,
      "step": 512
    },
    {
      "completion_length": 26.203125,
      "epoch": 1.5745007680491552,
      "grad_norm": 6.072816371917725,
      "kl": 0.3115234375,
      "learning_rate": 6.061538461538462e-07,
      "loss": 0.1223,
      "reward": 0.7528096139431,
      "reward_std": 0.06493799202144146,
      "reward_std_all": 0.06348242983222008,
      "rewards-max/forensics_sinkhorn_reward": 0.9018512070178986,
      "rewards-max/forensics_strict_edge_reward": 0.27045610547065735,
      "rewards-max/forensics_tiou_reward": 0.7801773846149445,
      "rewards/forensics_sinkhorn_reward": 0.6639079749584198,
      "rewards/forensics_strict_edge_reward": 0.022972000762820244,
      "rewards/forensics_tiou_reward": 0.4368573725223541,
      "step": 513
    },
    {
      "completion_length": 25.671875,
      "epoch": 1.5775729646697387,
      "grad_norm": 4.605407238006592,
      "kl": 0.3466796875,
      "learning_rate": 6.053846153846154e-07,
      "loss": 0.041,
      "reward": 0.747672826051712,
      "reward_std": 0.05200003832578659,
      "reward_std_all": 0.05944391153752804,
      "rewards-max/forensics_sinkhorn_reward": 0.9779234528541565,
      "rewards-max/forensics_strict_edge_reward": 0.36884360760450363,
      "rewards-max/forensics_tiou_reward": 0.9415282607078552,
      "rewards/forensics_sinkhorn_reward": 0.7143838405609131,
      "rewards/forensics_strict_edge_reward": 0.02944638580083847,
      "rewards/forensics_tiou_reward": 0.5046527683734894,
      "step": 514
    },
    {
      "completion_length": 28.171875,
      "epoch": 1.5806451612903225,
      "grad_norm": 5.896883964538574,
      "kl": 0.3779296875,
      "learning_rate": 6.046153846153847e-07,
      "loss": 0.0192,
      "reward": 0.7119544148445129,
      "reward_std": 0.050277333706617355,
      "reward_std_all": 0.06997869908809662,
      "rewards-max/forensics_sinkhorn_reward": 0.9797253012657166,
      "rewards-max/forensics_strict_edge_reward": 0.2995039811357856,
      "rewards-max/forensics_tiou_reward": 0.950508713722229,
      "rewards/forensics_sinkhorn_reward": 0.590450257062912,
      "rewards/forensics_strict_edge_reward": 0.014275396562879905,
      "rewards/forensics_tiou_reward": 0.4175010919570923,
      "step": 515
    },
    {
      "completion_length": 25.015625,
      "epoch": 1.5837173579109063,
      "grad_norm": 5.613369464874268,
      "kl": 0.349609375,
      "learning_rate": 6.038461538461538e-07,
      "loss": 0.0564,
      "reward": 0.7545890510082245,
      "reward_std": 0.06268385984003544,
      "reward_std_all": 0.05943126045167446,
      "rewards-max/forensics_sinkhorn_reward": 0.9853186011314392,
      "rewards-max/forensics_strict_edge_reward": 0.5030753463506699,
      "rewards-max/forensics_tiou_reward": 0.9657747149467468,
      "rewards/forensics_sinkhorn_reward": 0.69094318151474,
      "rewards/forensics_strict_edge_reward": 0.06013550981879234,
      "rewards/forensics_tiou_reward": 0.5534963607788086,
      "step": 516
    },
    {
      "completion_length": 21.859375,
      "epoch": 1.58678955453149,
      "grad_norm": 4.71525764465332,
      "kl": 0.37109375,
      "learning_rate": 6.03076923076923e-07,
      "loss": 0.043,
      "reward": 0.7796429097652435,
      "reward_std": 0.03413595259189606,
      "reward_std_all": 0.03299519047141075,
      "rewards-max/forensics_sinkhorn_reward": 0.9733358323574066,
      "rewards-max/forensics_strict_edge_reward": 0.4120611250400543,
      "rewards-max/forensics_tiou_reward": 0.9378118515014648,
      "rewards/forensics_sinkhorn_reward": 0.7070290446281433,
      "rewards/forensics_strict_edge_reward": 0.0313421618193388,
      "rewards/forensics_tiou_reward": 0.46311816573143005,
      "step": 517
    },
    {
      "completion_length": 25.0625,
      "epoch": 1.5898617511520738,
      "grad_norm": 10.041158676147461,
      "kl": 0.3466796875,
      "learning_rate": 6.023076923076923e-07,
      "loss": 0.0677,
      "reward": 0.7058559358119965,
      "reward_std": 0.09666433930397034,
      "reward_std_all": 0.09577419608831406,
      "rewards-max/forensics_sinkhorn_reward": 0.9820809066295624,
      "rewards-max/forensics_strict_edge_reward": 0.5676676407456398,
      "rewards-max/forensics_tiou_reward": 0.9596118927001953,
      "rewards/forensics_sinkhorn_reward": 0.6190239191055298,
      "rewards/forensics_strict_edge_reward": 0.053161829710006714,
      "rewards/forensics_tiou_reward": 0.41149692237377167,
      "step": 518
    },
    {
      "completion_length": 22.78125,
      "epoch": 1.5929339477726574,
      "grad_norm": 5.801365375518799,
      "kl": 0.45703125,
      "learning_rate": 6.015384615384615e-07,
      "loss": 0.0436,
      "reward": 0.7605760097503662,
      "reward_std": 0.044615376740694046,
      "reward_std_all": 0.045071808621287346,
      "rewards-max/forensics_sinkhorn_reward": 0.9754161536693573,
      "rewards-max/forensics_strict_edge_reward": 0.5900419056415558,
      "rewards-max/forensics_tiou_reward": 0.9452611207962036,
      "rewards/forensics_sinkhorn_reward": 0.6896449327468872,
      "rewards/forensics_strict_edge_reward": 0.04918495565652847,
      "rewards/forensics_tiou_reward": 0.4286113828420639,
      "step": 519
    },
    {
      "completion_length": 26.265625,
      "epoch": 1.5960061443932412,
      "grad_norm": 4.793998718261719,
      "kl": 0.3701171875,
      "learning_rate": 6.007692307692307e-07,
      "loss": 0.0063,
      "reward": 0.737086147069931,
      "reward_std": 0.037789519876241684,
      "reward_std_all": 0.05608267895877361,
      "rewards-max/forensics_sinkhorn_reward": 0.9462623298168182,
      "rewards-max/forensics_strict_edge_reward": 0.26605452969670296,
      "rewards-max/forensics_tiou_reward": 0.8776739835739136,
      "rewards/forensics_sinkhorn_reward": 0.6388591527938843,
      "rewards/forensics_strict_edge_reward": 0.01675047050230205,
      "rewards/forensics_tiou_reward": 0.4058521240949631,
      "step": 520
    },
    {
      "completion_length": 26.359375,
      "epoch": 1.5990783410138247,
      "grad_norm": 5.116741180419922,
      "kl": 0.34765625,
      "learning_rate": 6e-07,
      "loss": 0.0547,
      "reward": 0.7608191072940826,
      "reward_std": 0.04575154557824135,
      "reward_std_all": 0.04645892605185509,
      "rewards-max/forensics_sinkhorn_reward": 0.9710914194583893,
      "rewards-max/forensics_strict_edge_reward": 0.28496674448251724,
      "rewards-max/forensics_tiou_reward": 0.9210672080516815,
      "rewards/forensics_sinkhorn_reward": 0.6797505915164948,
      "rewards/forensics_strict_edge_reward": 0.014796408591791987,
      "rewards/forensics_tiou_reward": 0.42763765156269073,
      "step": 521
    },
    {
      "completion_length": 22.078125,
      "epoch": 1.6021505376344085,
      "grad_norm": 4.8686394691467285,
      "kl": 0.3896484375,
      "learning_rate": 5.992307692307692e-07,
      "loss": 0.0536,
      "reward": 0.7313961982727051,
      "reward_std": 0.04321767948567867,
      "reward_std_all": 0.05840879958122969,
      "rewards-max/forensics_sinkhorn_reward": 0.9682502746582031,
      "rewards-max/forensics_strict_edge_reward": 0.28303438052535057,
      "rewards-max/forensics_tiou_reward": 0.9113249480724335,
      "rewards/forensics_sinkhorn_reward": 0.6531081795692444,
      "rewards/forensics_strict_edge_reward": 0.03221721155568957,
      "rewards/forensics_tiou_reward": 0.43546992540359497,
      "step": 522
    },
    {
      "completion_length": 26.15625,
      "epoch": 1.6052227342549923,
      "grad_norm": 5.575462341308594,
      "kl": 0.3642578125,
      "learning_rate": 5.984615384615385e-07,
      "loss": -0.0259,
      "reward": 0.7279048264026642,
      "reward_std": 0.03741587791591883,
      "reward_std_all": 0.057713257148861885,
      "rewards-max/forensics_sinkhorn_reward": 0.9791211187839508,
      "rewards-max/forensics_strict_edge_reward": 0.2990078628063202,
      "rewards-max/forensics_tiou_reward": 0.9495203793048859,
      "rewards/forensics_sinkhorn_reward": 0.674063503742218,
      "rewards/forensics_strict_edge_reward": 0.03687002323567867,
      "rewards/forensics_tiou_reward": 0.4949598163366318,
      "step": 523
    },
    {
      "completion_length": 21.65625,
      "epoch": 1.608294930875576,
      "grad_norm": 5.80586576461792,
      "kl": 0.419921875,
      "learning_rate": 5.976923076923077e-07,
      "loss": -0.0501,
      "reward": 0.7936035692691803,
      "reward_std": 0.019052845425903797,
      "reward_std_all": 0.01869061589241028,
      "rewards-max/forensics_sinkhorn_reward": 0.9838200211524963,
      "rewards-max/forensics_strict_edge_reward": 0.538444846868515,
      "rewards-max/forensics_tiou_reward": 0.9604997634887695,
      "rewards/forensics_sinkhorn_reward": 0.7887231707572937,
      "rewards/forensics_strict_edge_reward": 0.08748277276754379,
      "rewards/forensics_tiou_reward": 0.5892106294631958,
      "step": 524
    },
    {
      "completion_length": 21.03125,
      "epoch": 1.6113671274961598,
      "grad_norm": 5.584583759307861,
      "kl": 0.30859375,
      "learning_rate": 5.969230769230768e-07,
      "loss": 0.0058,
      "reward": 0.7303845882415771,
      "reward_std": 0.07292747125029564,
      "reward_std_all": 0.07800974696874619,
      "rewards-max/forensics_sinkhorn_reward": 0.900935024023056,
      "rewards-max/forensics_strict_edge_reward": 0.02401515014935285,
      "rewards-max/forensics_tiou_reward": 0.7577084600925446,
      "rewards/forensics_sinkhorn_reward": 0.6585625410079956,
      "rewards/forensics_strict_edge_reward": 0.001764068285410758,
      "rewards/forensics_tiou_reward": 0.3658946603536606,
      "step": 525
    },
    {
      "completion_length": 27.03125,
      "epoch": 1.6144393241167436,
      "grad_norm": 6.3262939453125,
      "kl": 0.3876953125,
      "learning_rate": 5.961538461538461e-07,
      "loss": 0.0395,
      "reward": 0.7920597195625305,
      "reward_std": 0.020280523225665092,
      "reward_std_all": 0.02049604058265686,
      "rewards-max/forensics_sinkhorn_reward": 0.9703563749790192,
      "rewards-max/forensics_strict_edge_reward": 0.15600645309314132,
      "rewards-max/forensics_tiou_reward": 0.927686870098114,
      "rewards/forensics_sinkhorn_reward": 0.7501617670059204,
      "rewards/forensics_strict_edge_reward": 0.011144424672238529,
      "rewards/forensics_tiou_reward": 0.4849037230014801,
      "step": 526
    },
    {
      "completion_length": 23.109375,
      "epoch": 1.6175115207373272,
      "grad_norm": 4.923805236816406,
      "kl": 0.30078125,
      "learning_rate": 5.953846153846154e-07,
      "loss": 0.1318,
      "reward": 0.7763845920562744,
      "reward_std": 0.022021030075848103,
      "reward_std_all": 0.027439434081315994,
      "rewards-max/forensics_sinkhorn_reward": 0.9838521480560303,
      "rewards-max/forensics_strict_edge_reward": 0.3904794454574585,
      "rewards-max/forensics_tiou_reward": 0.9645612835884094,
      "rewards/forensics_sinkhorn_reward": 0.7221252918243408,
      "rewards/forensics_strict_edge_reward": 0.036181473173201084,
      "rewards/forensics_tiou_reward": 0.5116247981786728,
      "step": 527
    },
    {
      "completion_length": 22.78125,
      "epoch": 1.620583717357911,
      "grad_norm": 5.271144390106201,
      "kl": 0.3798828125,
      "learning_rate": 5.946153846153847e-07,
      "loss": 0.1646,
      "reward": 0.7413913607597351,
      "reward_std": 0.055655477568507195,
      "reward_std_all": 0.0613996759057045,
      "rewards-max/forensics_sinkhorn_reward": 0.9800819754600525,
      "rewards-max/forensics_strict_edge_reward": 0.4747692793607712,
      "rewards-max/forensics_tiou_reward": 0.9480196833610535,
      "rewards/forensics_sinkhorn_reward": 0.6678541004657745,
      "rewards/forensics_strict_edge_reward": 0.04806220345199108,
      "rewards/forensics_tiou_reward": 0.47692446410655975,
      "step": 528
    },
    {
      "completion_length": 23.640625,
      "epoch": 1.6236559139784945,
      "grad_norm": 6.956597805023193,
      "kl": 0.3037109375,
      "learning_rate": 5.938461538461538e-07,
      "loss": -0.0177,
      "reward": 0.7561820149421692,
      "reward_std": 0.058022527024149895,
      "reward_std_all": 0.055933400988578796,
      "rewards-max/forensics_sinkhorn_reward": 0.9737933576107025,
      "rewards-max/forensics_strict_edge_reward": 0.3973057419061661,
      "rewards-max/forensics_tiou_reward": 0.9359787106513977,
      "rewards/forensics_sinkhorn_reward": 0.7228090167045593,
      "rewards/forensics_strict_edge_reward": 0.040380255319178104,
      "rewards/forensics_tiou_reward": 0.4982886016368866,
      "step": 529
    },
    {
      "completion_length": 23.5,
      "epoch": 1.6267281105990783,
      "grad_norm": 4.794943809509277,
      "kl": 0.3232421875,
      "learning_rate": 5.93076923076923e-07,
      "loss": -0.0108,
      "reward": 0.7893618941307068,
      "reward_std": 0.01814545225352049,
      "reward_std_all": 0.019308692310005426,
      "rewards-max/forensics_sinkhorn_reward": 0.9752293229103088,
      "rewards-max/forensics_strict_edge_reward": 0.31398798525333405,
      "rewards-max/forensics_tiou_reward": 0.9323554635047913,
      "rewards/forensics_sinkhorn_reward": 0.7457992732524872,
      "rewards/forensics_strict_edge_reward": 0.043928357772529125,
      "rewards/forensics_tiou_reward": 0.5392298400402069,
      "step": 530
    },
    {
      "completion_length": 26.890625,
      "epoch": 1.629800307219662,
      "grad_norm": 5.631935119628906,
      "kl": 0.3251953125,
      "learning_rate": 5.923076923076923e-07,
      "loss": 0.0909,
      "reward": 0.7274686992168427,
      "reward_std": 0.06542899459600449,
      "reward_std_all": 0.06900795176625252,
      "rewards-max/forensics_sinkhorn_reward": 0.9759573042392731,
      "rewards-max/forensics_strict_edge_reward": 0.4396780878305435,
      "rewards-max/forensics_tiou_reward": 0.9438716173171997,
      "rewards/forensics_sinkhorn_reward": 0.625476598739624,
      "rewards/forensics_strict_edge_reward": 0.024991903454065323,
      "rewards/forensics_tiou_reward": 0.4100148379802704,
      "step": 531
    },
    {
      "completion_length": 26.0625,
      "epoch": 1.6328725038402458,
      "grad_norm": 5.200063705444336,
      "kl": 0.4873046875,
      "learning_rate": 5.915384615384615e-07,
      "loss": 0.0628,
      "reward": 0.7488273680210114,
      "reward_std": 0.035292237997055054,
      "reward_std_all": 0.044809489510953426,
      "rewards-max/forensics_sinkhorn_reward": 0.9900142550468445,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.981589138507843,
      "rewards/forensics_sinkhorn_reward": 0.7026686370372772,
      "rewards/forensics_strict_edge_reward": 0.047719551250338554,
      "rewards/forensics_tiou_reward": 0.458983451128006,
      "step": 532
    },
    {
      "completion_length": 26.265625,
      "epoch": 1.6359447004608296,
      "grad_norm": 6.199139595031738,
      "kl": 0.3466796875,
      "learning_rate": 5.907692307692307e-07,
      "loss": 0.0644,
      "reward": 0.7662113010883331,
      "reward_std": 0.047226106747984886,
      "reward_std_all": 0.045614223927259445,
      "rewards-max/forensics_sinkhorn_reward": 0.9928396642208099,
      "rewards-max/forensics_strict_edge_reward": 0.5790443122386932,
      "rewards-max/forensics_tiou_reward": 0.9846548140048981,
      "rewards/forensics_sinkhorn_reward": 0.7310513556003571,
      "rewards/forensics_strict_edge_reward": 0.04126300849020481,
      "rewards/forensics_tiou_reward": 0.5052470266819,
      "step": 533
    },
    {
      "completion_length": 22.78125,
      "epoch": 1.6390168970814132,
      "grad_norm": 5.027120590209961,
      "kl": 0.28759765625,
      "learning_rate": 5.9e-07,
      "loss": -0.039,
      "reward": 0.7451409101486206,
      "reward_std": 0.021745986305177212,
      "reward_std_all": 0.03971888870000839,
      "rewards-max/forensics_sinkhorn_reward": 0.9816907346248627,
      "rewards-max/forensics_strict_edge_reward": 0.19958370923995972,
      "rewards-max/forensics_tiou_reward": 0.9486008882522583,
      "rewards/forensics_sinkhorn_reward": 0.7061793208122253,
      "rewards/forensics_strict_edge_reward": 0.016365804709494114,
      "rewards/forensics_tiou_reward": 0.4592537581920624,
      "step": 534
    },
    {
      "completion_length": 25.5,
      "epoch": 1.642089093701997,
      "grad_norm": 6.2139482498168945,
      "kl": 0.4794921875,
      "learning_rate": 5.892307692307692e-07,
      "loss": 0.085,
      "reward": 0.7222417593002319,
      "reward_std": 0.06752812303602695,
      "reward_std_all": 0.07459197752177715,
      "rewards-max/forensics_sinkhorn_reward": 0.9776875674724579,
      "rewards-max/forensics_strict_edge_reward": 0.480148583650589,
      "rewards-max/forensics_tiou_reward": 0.9330918192863464,
      "rewards/forensics_sinkhorn_reward": 0.6407077610492706,
      "rewards/forensics_strict_edge_reward": 0.049716345965862274,
      "rewards/forensics_tiou_reward": 0.434604674577713,
      "step": 535
    },
    {
      "completion_length": 26.4375,
      "epoch": 1.6451612903225805,
      "grad_norm": 5.45266056060791,
      "kl": 0.3486328125,
      "learning_rate": 5.884615384615385e-07,
      "loss": 0.0721,
      "reward": 0.7384983599185944,
      "reward_std": 0.07593357935547829,
      "reward_std_all": 0.073113564401865,
      "rewards-max/forensics_sinkhorn_reward": 0.9676595330238342,
      "rewards-max/forensics_strict_edge_reward": 0.36850982904434204,
      "rewards-max/forensics_tiou_reward": 0.9153431355953217,
      "rewards/forensics_sinkhorn_reward": 0.6485729217529297,
      "rewards/forensics_strict_edge_reward": 0.023782705888152122,
      "rewards/forensics_tiou_reward": 0.46628984808921814,
      "step": 536
    },
    {
      "completion_length": 23.703125,
      "epoch": 1.6482334869431643,
      "grad_norm": 5.209942817687988,
      "kl": 0.443359375,
      "learning_rate": 5.876923076923077e-07,
      "loss": -0.018,
      "reward": 0.7743256092071533,
      "reward_std": 0.04145381413400173,
      "reward_std_all": 0.03962506912648678,
      "rewards-max/forensics_sinkhorn_reward": 0.981300413608551,
      "rewards-max/forensics_strict_edge_reward": 0.4790885001420975,
      "rewards-max/forensics_tiou_reward": 0.9594281315803528,
      "rewards/forensics_sinkhorn_reward": 0.7212466895580292,
      "rewards/forensics_strict_edge_reward": 0.04088435973972082,
      "rewards/forensics_tiou_reward": 0.5149890631437302,
      "step": 537
    },
    {
      "completion_length": 25.96875,
      "epoch": 1.651305683563748,
      "grad_norm": 5.940717697143555,
      "kl": 0.5615234375,
      "learning_rate": 5.869230769230768e-07,
      "loss": 0.1094,
      "reward": 0.7472900152206421,
      "reward_std": 0.03054447891190648,
      "reward_std_all": 0.04683329910039902,
      "rewards-max/forensics_sinkhorn_reward": 0.9672617316246033,
      "rewards-max/forensics_strict_edge_reward": 0.3465723395347595,
      "rewards-max/forensics_tiou_reward": 0.9306184351444244,
      "rewards/forensics_sinkhorn_reward": 0.6974330246448517,
      "rewards/forensics_strict_edge_reward": 0.020940148271620274,
      "rewards/forensics_tiou_reward": 0.50143101811409,
      "step": 538
    },
    {
      "completion_length": 28.34375,
      "epoch": 1.6543778801843319,
      "grad_norm": 4.86178731918335,
      "kl": 0.44921875,
      "learning_rate": 5.861538461538461e-07,
      "loss": 0.0765,
      "reward": 0.715291440486908,
      "reward_std": 0.07016776129603386,
      "reward_std_all": 0.07857784628868103,
      "rewards-max/forensics_sinkhorn_reward": 0.9858018457889557,
      "rewards-max/forensics_strict_edge_reward": 0.220241479575634,
      "rewards-max/forensics_tiou_reward": 0.9673080742359161,
      "rewards/forensics_sinkhorn_reward": 0.6516528427600861,
      "rewards/forensics_strict_edge_reward": 0.022304059006273746,
      "rewards/forensics_tiou_reward": 0.4317893236875534,
      "step": 539
    },
    {
      "completion_length": 29.265625,
      "epoch": 1.6574500768049156,
      "grad_norm": 4.855945587158203,
      "kl": 0.4296875,
      "learning_rate": 5.853846153846154e-07,
      "loss": 0.2198,
      "reward": 0.7628938853740692,
      "reward_std": 0.05535350367426872,
      "reward_std_all": 0.052220672369003296,
      "rewards-max/forensics_sinkhorn_reward": 0.9791187942028046,
      "rewards-max/forensics_strict_edge_reward": 0.5516124665737152,
      "rewards-max/forensics_tiou_reward": 0.9534289836883545,
      "rewards/forensics_sinkhorn_reward": 0.6774370074272156,
      "rewards/forensics_strict_edge_reward": 0.06223068758845329,
      "rewards/forensics_tiou_reward": 0.520197868347168,
      "step": 540
    },
    {
      "completion_length": 24.625,
      "epoch": 1.6605222734254992,
      "grad_norm": 5.309898853302002,
      "kl": 0.4169921875,
      "learning_rate": 5.846153846153847e-07,
      "loss": 0.0986,
      "reward": 0.7631979286670685,
      "reward_std": 0.05589957907795906,
      "reward_std_all": 0.05254729278385639,
      "rewards-max/forensics_sinkhorn_reward": 0.9923299551010132,
      "rewards-max/forensics_strict_edge_reward": 0.7789089381694794,
      "rewards-max/forensics_tiou_reward": 0.9875605404376984,
      "rewards/forensics_sinkhorn_reward": 0.7127178311347961,
      "rewards/forensics_strict_edge_reward": 0.0545542910695076,
      "rewards/forensics_tiou_reward": 0.5586830079555511,
      "step": 541
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.663594470046083,
      "grad_norm": 4.83980655670166,
      "kl": 0.4287109375,
      "learning_rate": 5.838461538461538e-07,
      "loss": 0.0167,
      "reward": 0.7541109919548035,
      "reward_std": 0.06439826264977455,
      "reward_std_all": 0.06103387102484703,
      "rewards-max/forensics_sinkhorn_reward": 0.9904683828353882,
      "rewards-max/forensics_strict_edge_reward": 0.4962490648031235,
      "rewards-max/forensics_tiou_reward": 0.9789248704910278,
      "rewards/forensics_sinkhorn_reward": 0.7420276403427124,
      "rewards/forensics_strict_edge_reward": 0.04267875477671623,
      "rewards/forensics_tiou_reward": 0.588449090719223,
      "step": 542
    },
    {
      "completion_length": 24.765625,
      "epoch": 1.6666666666666665,
      "grad_norm": 5.984306335449219,
      "kl": 0.4169921875,
      "learning_rate": 5.83076923076923e-07,
      "loss": 0.1551,
      "reward": 0.7427643239498138,
      "reward_std": 0.08673398569226265,
      "reward_std_all": 0.08098529651761055,
      "rewards-max/forensics_sinkhorn_reward": 0.9147821962833405,
      "rewards-max/forensics_strict_edge_reward": 0.09708453249186277,
      "rewards-max/forensics_tiou_reward": 0.7879888415336609,
      "rewards/forensics_sinkhorn_reward": 0.6486531794071198,
      "rewards/forensics_strict_edge_reward": 0.006436544412281364,
      "rewards/forensics_tiou_reward": 0.40958285331726074,
      "step": 543
    },
    {
      "completion_length": 28.453125,
      "epoch": 1.6697388632872503,
      "grad_norm": 5.55346155166626,
      "kl": 0.4072265625,
      "learning_rate": 5.823076923076923e-07,
      "loss": 0.1136,
      "reward": 0.7375372052192688,
      "reward_std": 0.052188897505402565,
      "reward_std_all": 0.05851539969444275,
      "rewards-max/forensics_sinkhorn_reward": 0.9592294692993164,
      "rewards-max/forensics_strict_edge_reward": 0.18100298941135406,
      "rewards-max/forensics_tiou_reward": 0.89053875207901,
      "rewards/forensics_sinkhorn_reward": 0.6500475406646729,
      "rewards/forensics_strict_edge_reward": 0.014654445927590132,
      "rewards/forensics_tiou_reward": 0.460944801568985,
      "step": 544
    },
    {
      "completion_length": 25.015625,
      "epoch": 1.672811059907834,
      "grad_norm": 5.247895240783691,
      "kl": 0.302734375,
      "learning_rate": 5.815384615384615e-07,
      "loss": 0.1067,
      "reward": 0.7509293854236603,
      "reward_std": 0.0799848921597004,
      "reward_std_all": 0.07322725281119347,
      "rewards-max/forensics_sinkhorn_reward": 0.9250948131084442,
      "rewards-max/forensics_strict_edge_reward": 0.10335423424839973,
      "rewards-max/forensics_tiou_reward": 0.8289446532726288,
      "rewards/forensics_sinkhorn_reward": 0.6439631581306458,
      "rewards/forensics_strict_edge_reward": 0.008201833814382553,
      "rewards/forensics_tiou_reward": 0.4179396480321884,
      "step": 545
    },
    {
      "completion_length": 26.046875,
      "epoch": 1.6758832565284179,
      "grad_norm": 5.194231033325195,
      "kl": 0.3076171875,
      "learning_rate": 5.807692307692307e-07,
      "loss": 0.0603,
      "reward": 0.7415262758731842,
      "reward_std": 0.08127761818468571,
      "reward_std_all": 0.07632699236273766,
      "rewards-max/forensics_sinkhorn_reward": 0.96274733543396,
      "rewards-max/forensics_strict_edge_reward": 0.3280272036790848,
      "rewards-max/forensics_tiou_reward": 0.9087026417255402,
      "rewards/forensics_sinkhorn_reward": 0.6547622680664062,
      "rewards/forensics_strict_edge_reward": 0.028218654450029135,
      "rewards/forensics_tiou_reward": 0.5251676440238953,
      "step": 546
    },
    {
      "completion_length": 24.171875,
      "epoch": 1.6789554531490016,
      "grad_norm": 5.128241539001465,
      "kl": 0.3310546875,
      "learning_rate": 5.8e-07,
      "loss": -0.0033,
      "reward": 0.7776873707771301,
      "reward_std": 0.029310548678040504,
      "reward_std_all": 0.030481668189167976,
      "rewards-max/forensics_sinkhorn_reward": 0.9565781950950623,
      "rewards-max/forensics_strict_edge_reward": 0.3414454311132431,
      "rewards-max/forensics_tiou_reward": 0.9014991223812103,
      "rewards/forensics_sinkhorn_reward": 0.7062909603118896,
      "rewards/forensics_strict_edge_reward": 0.036216418258845806,
      "rewards/forensics_tiou_reward": 0.4986877143383026,
      "step": 547
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.6820276497695854,
      "grad_norm": 6.459956169128418,
      "kl": 0.3310546875,
      "learning_rate": 5.792307692307692e-07,
      "loss": 0.2207,
      "reward": 0.7346845865249634,
      "reward_std": 0.07685772702097893,
      "reward_std_all": 0.07656655088067055,
      "rewards-max/forensics_sinkhorn_reward": 0.9286321997642517,
      "rewards-max/forensics_strict_edge_reward": 0.09119362663477659,
      "rewards-max/forensics_tiou_reward": 0.8296638131141663,
      "rewards/forensics_sinkhorn_reward": 0.6312164962291718,
      "rewards/forensics_strict_edge_reward": 0.005651429935824126,
      "rewards/forensics_tiou_reward": 0.40098996460437775,
      "step": 548
    },
    {
      "completion_length": 25.109375,
      "epoch": 1.685099846390169,
      "grad_norm": 4.465525150299072,
      "kl": 0.3271484375,
      "learning_rate": 5.784615384615385e-07,
      "loss": 0.0387,
      "reward": 0.7630963921546936,
      "reward_std": 0.03192708268761635,
      "reward_std_all": 0.03900148160755634,
      "rewards-max/forensics_sinkhorn_reward": 0.9884304702281952,
      "rewards-max/forensics_strict_edge_reward": 0.5252516865730286,
      "rewards-max/forensics_tiou_reward": 0.9685648083686829,
      "rewards/forensics_sinkhorn_reward": 0.7061655521392822,
      "rewards/forensics_strict_edge_reward": 0.039623684249818325,
      "rewards/forensics_tiou_reward": 0.46366186439991,
      "step": 549
    },
    {
      "completion_length": 24.6875,
      "epoch": 1.6881720430107527,
      "grad_norm": 8.22679328918457,
      "kl": 0.2958984375,
      "learning_rate": 5.776923076923077e-07,
      "loss": -0.0023,
      "reward": 0.766627311706543,
      "reward_std": 0.043298461474478245,
      "reward_std_all": 0.04272250272333622,
      "rewards-max/forensics_sinkhorn_reward": 0.9786506593227386,
      "rewards-max/forensics_strict_edge_reward": 0.5050012171268463,
      "rewards-max/forensics_tiou_reward": 0.9508808553218842,
      "rewards/forensics_sinkhorn_reward": 0.7370678186416626,
      "rewards/forensics_strict_edge_reward": 0.0389044638723135,
      "rewards/forensics_tiou_reward": 0.4830775260925293,
      "step": 550
    },
    {
      "completion_length": 23.28125,
      "epoch": 1.6912442396313363,
      "grad_norm": 5.630136966705322,
      "kl": 0.330078125,
      "learning_rate": 5.769230769230768e-07,
      "loss": 0.0541,
      "reward": 0.7802795469760895,
      "reward_std": 0.030888275243341923,
      "reward_std_all": 0.03082169685512781,
      "rewards-max/forensics_sinkhorn_reward": 0.9710076451301575,
      "rewards-max/forensics_strict_edge_reward": 0.34577055275440216,
      "rewards-max/forensics_tiou_reward": 0.9316183924674988,
      "rewards/forensics_sinkhorn_reward": 0.7264794409275055,
      "rewards/forensics_strict_edge_reward": 0.03847440890967846,
      "rewards/forensics_tiou_reward": 0.5315370261669159,
      "step": 551
    },
    {
      "completion_length": 24.1875,
      "epoch": 1.69431643625192,
      "grad_norm": 6.399780750274658,
      "kl": 0.3515625,
      "learning_rate": 5.761538461538461e-07,
      "loss": 0.0042,
      "reward": 0.7926977574825287,
      "reward_std": 0.021971918176859617,
      "reward_std_all": 0.021361561492085457,
      "rewards-max/forensics_sinkhorn_reward": 0.9791691899299622,
      "rewards-max/forensics_strict_edge_reward": 0.3908739238977432,
      "rewards-max/forensics_tiou_reward": 0.9511233568191528,
      "rewards/forensics_sinkhorn_reward": 0.781592458486557,
      "rewards/forensics_strict_edge_reward": 0.03222522884607315,
      "rewards/forensics_tiou_reward": 0.5736795365810394,
      "step": 552
    },
    {
      "completion_length": 22.484375,
      "epoch": 1.6973886328725039,
      "grad_norm": 8.528595924377441,
      "kl": 0.35546875,
      "learning_rate": 5.753846153846153e-07,
      "loss": 0.0566,
      "reward": 0.7803236246109009,
      "reward_std": 0.030069400556385517,
      "reward_std_all": 0.030109363608062267,
      "rewards-max/forensics_sinkhorn_reward": 0.8981707394123077,
      "rewards-max/forensics_strict_edge_reward": 0.07752389460802078,
      "rewards-max/forensics_tiou_reward": 0.7540079355239868,
      "rewards/forensics_sinkhorn_reward": 0.7276817858219147,
      "rewards/forensics_strict_edge_reward": 0.0042216926231049,
      "rewards/forensics_tiou_reward": 0.42758525907993317,
      "step": 553
    },
    {
      "completion_length": 26.953125,
      "epoch": 1.7004608294930876,
      "grad_norm": 7.056140899658203,
      "kl": 0.3671875,
      "learning_rate": 5.746153846153847e-07,
      "loss": 0.0038,
      "reward": 0.7976947724819183,
      "reward_std": 0.009280869271606207,
      "reward_std_all": 0.010841267881914973,
      "rewards-max/forensics_sinkhorn_reward": 0.9653243720531464,
      "rewards-max/forensics_strict_edge_reward": 0.40153393894433975,
      "rewards-max/forensics_tiou_reward": 0.9139664769172668,
      "rewards/forensics_sinkhorn_reward": 0.7580825686454773,
      "rewards/forensics_strict_edge_reward": 0.037928421050310135,
      "rewards/forensics_tiou_reward": 0.5043841302394867,
      "step": 554
    },
    {
      "completion_length": 26.953125,
      "epoch": 1.7035330261136714,
      "grad_norm": 5.051644325256348,
      "kl": 0.302734375,
      "learning_rate": 5.738461538461539e-07,
      "loss": 0.1076,
      "reward": 0.7594937086105347,
      "reward_std": 0.0565992770716548,
      "reward_std_all": 0.053983697667717934,
      "rewards-max/forensics_sinkhorn_reward": 0.9670401215553284,
      "rewards-max/forensics_strict_edge_reward": 0.4104880876839161,
      "rewards-max/forensics_tiou_reward": 0.9188442826271057,
      "rewards/forensics_sinkhorn_reward": 0.6756306886672974,
      "rewards/forensics_strict_edge_reward": 0.028028397355228662,
      "rewards/forensics_tiou_reward": 0.42460107803344727,
      "step": 555
    },
    {
      "completion_length": 24.859375,
      "epoch": 1.706605222734255,
      "grad_norm": 9.300514221191406,
      "kl": 0.353515625,
      "learning_rate": 5.73076923076923e-07,
      "loss": 0.0938,
      "reward": 0.7181307077407837,
      "reward_std": 0.08466774225234985,
      "reward_std_all": 0.0899047739803791,
      "rewards-max/forensics_sinkhorn_reward": 0.9931061863899231,
      "rewards-max/forensics_strict_edge_reward": 0.26962868869304657,
      "rewards-max/forensics_tiou_reward": 0.9842645525932312,
      "rewards/forensics_sinkhorn_reward": 0.6205215156078339,
      "rewards/forensics_strict_edge_reward": 0.02451472682878375,
      "rewards/forensics_tiou_reward": 0.41677623987197876,
      "step": 556
    },
    {
      "completion_length": 22.890625,
      "epoch": 1.7096774193548387,
      "grad_norm": 4.953405380249023,
      "kl": 0.349609375,
      "learning_rate": 5.723076923076923e-07,
      "loss": -0.0628,
      "reward": 0.7589689195156097,
      "reward_std": 0.04218098148703575,
      "reward_std_all": 0.044913507997989655,
      "rewards-max/forensics_sinkhorn_reward": 0.9663752317428589,
      "rewards-max/forensics_strict_edge_reward": 0.3564630448818207,
      "rewards-max/forensics_tiou_reward": 0.9225166141986847,
      "rewards/forensics_sinkhorn_reward": 0.7373690903186798,
      "rewards/forensics_strict_edge_reward": 0.02392559591680765,
      "rewards/forensics_tiou_reward": 0.4888036847114563,
      "step": 557
    },
    {
      "completion_length": 21.296875,
      "epoch": 1.7127496159754223,
      "grad_norm": 11.73309326171875,
      "kl": 0.998046875,
      "learning_rate": 5.715384615384615e-07,
      "loss": 0.0496,
      "reward": 0.7390838265419006,
      "reward_std": 0.05543125979602337,
      "reward_std_all": 0.06659392081201077,
      "rewards-max/forensics_sinkhorn_reward": 0.9897312223911285,
      "rewards-max/forensics_strict_edge_reward": 0.6123424768447876,
      "rewards-max/forensics_tiou_reward": 0.9788979291915894,
      "rewards/forensics_sinkhorn_reward": 0.7320366501808167,
      "rewards/forensics_strict_edge_reward": 0.055898185819387436,
      "rewards/forensics_tiou_reward": 0.4832073897123337,
      "step": 558
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.715821812596006,
      "grad_norm": 5.030884265899658,
      "kl": 0.3837890625,
      "learning_rate": 5.707692307692307e-07,
      "loss": 0.0668,
      "reward": 0.7479560673236847,
      "reward_std": 0.0360380532220006,
      "reward_std_all": 0.04915503226220608,
      "rewards-max/forensics_sinkhorn_reward": 0.9936374127864838,
      "rewards-max/forensics_strict_edge_reward": 0.33921515941619873,
      "rewards-max/forensics_tiou_reward": 0.9868581295013428,
      "rewards/forensics_sinkhorn_reward": 0.6913602948188782,
      "rewards/forensics_strict_edge_reward": 0.04486241703853011,
      "rewards/forensics_tiou_reward": 0.4794818013906479,
      "step": 559
    },
    {
      "completion_length": 27.859375,
      "epoch": 1.7188940092165899,
      "grad_norm": 5.460862636566162,
      "kl": 0.33984375,
      "learning_rate": 5.699999999999999e-07,
      "loss": 0.0782,
      "reward": 0.7285134196281433,
      "reward_std": 0.05455722287297249,
      "reward_std_all": 0.0665105190128088,
      "rewards-max/forensics_sinkhorn_reward": 0.9421485960483551,
      "rewards-max/forensics_strict_edge_reward": 0.3761972188949585,
      "rewards-max/forensics_tiou_reward": 0.8703864216804504,
      "rewards/forensics_sinkhorn_reward": 0.6556586027145386,
      "rewards/forensics_strict_edge_reward": 0.02287632878869772,
      "rewards/forensics_tiou_reward": 0.4252951443195343,
      "step": 560
    },
    {
      "completion_length": 24.171875,
      "epoch": 1.7219662058371736,
      "grad_norm": 6.106491565704346,
      "kl": 0.3623046875,
      "learning_rate": 5.692307692307692e-07,
      "loss": 0.0644,
      "reward": 0.7615661919116974,
      "reward_std": 0.05895922891795635,
      "reward_std_all": 0.05547734908759594,
      "rewards-max/forensics_sinkhorn_reward": 0.9772482514381409,
      "rewards-max/forensics_strict_edge_reward": 0.29412222653627396,
      "rewards-max/forensics_tiou_reward": 0.9390967786312103,
      "rewards/forensics_sinkhorn_reward": 0.6958720982074738,
      "rewards/forensics_strict_edge_reward": 0.028517596423625946,
      "rewards/forensics_tiou_reward": 0.47644124925136566,
      "step": 561
    },
    {
      "completion_length": 26.015625,
      "epoch": 1.7250384024577574,
      "grad_norm": 6.155138969421387,
      "kl": 0.328125,
      "learning_rate": 5.684615384615385e-07,
      "loss": -0.0229,
      "reward": 0.7723680138587952,
      "reward_std": 0.03408723417669535,
      "reward_std_all": 0.0356166313868016,
      "rewards-max/forensics_sinkhorn_reward": 0.9723333418369293,
      "rewards-max/forensics_strict_edge_reward": 0.44267232716083527,
      "rewards-max/forensics_tiou_reward": 0.9366407990455627,
      "rewards/forensics_sinkhorn_reward": 0.7354342639446259,
      "rewards/forensics_strict_edge_reward": 0.031970882788300514,
      "rewards/forensics_tiou_reward": 0.5225422382354736,
      "step": 562
    },
    {
      "completion_length": 23.640625,
      "epoch": 1.728110599078341,
      "grad_norm": 6.591681003570557,
      "kl": 0.3984375,
      "learning_rate": 5.676923076923077e-07,
      "loss": 0.005,
      "reward": 0.7662014961242676,
      "reward_std": 0.03186935931444168,
      "reward_std_all": 0.0360336359590292,
      "rewards-max/forensics_sinkhorn_reward": 0.9777005314826965,
      "rewards-max/forensics_strict_edge_reward": 0.42239537835121155,
      "rewards-max/forensics_tiou_reward": 0.9469994902610779,
      "rewards/forensics_sinkhorn_reward": 0.7108339965343475,
      "rewards/forensics_strict_edge_reward": 0.03868464194238186,
      "rewards/forensics_tiou_reward": 0.46924422681331635,
      "step": 563
    },
    {
      "completion_length": 25.921875,
      "epoch": 1.7311827956989247,
      "grad_norm": 6.293186187744141,
      "kl": 0.345703125,
      "learning_rate": 5.669230769230769e-07,
      "loss": 0.0424,
      "reward": 0.7320941984653473,
      "reward_std": 0.08994228392839432,
      "reward_std_all": 0.0852532759308815,
      "rewards-max/forensics_sinkhorn_reward": 0.9247129857540131,
      "rewards-max/forensics_strict_edge_reward": 0.13442042469978333,
      "rewards-max/forensics_tiou_reward": 0.8150245547294617,
      "rewards/forensics_sinkhorn_reward": 0.6717274487018585,
      "rewards/forensics_strict_edge_reward": 0.006886923685669899,
      "rewards/forensics_tiou_reward": 0.429564893245697,
      "step": 564
    },
    {
      "completion_length": 26.25,
      "epoch": 1.7342549923195083,
      "grad_norm": 4.2072858810424805,
      "kl": 0.3564453125,
      "learning_rate": 5.661538461538461e-07,
      "loss": 0.0648,
      "reward": 0.731645941734314,
      "reward_std": 0.04868502542376518,
      "reward_std_all": 0.06289991736412048,
      "rewards-max/forensics_sinkhorn_reward": 0.990532785654068,
      "rewards-max/forensics_strict_edge_reward": 0.45910312235355377,
      "rewards-max/forensics_tiou_reward": 0.9746630489826202,
      "rewards/forensics_sinkhorn_reward": 0.7211478650569916,
      "rewards/forensics_strict_edge_reward": 0.039345210418105125,
      "rewards/forensics_tiou_reward": 0.5548164546489716,
      "step": 565
    },
    {
      "completion_length": 24.265625,
      "epoch": 1.737327188940092,
      "grad_norm": 7.26626443862915,
      "kl": 0.5126953125,
      "learning_rate": 5.653846153846153e-07,
      "loss": 0.0242,
      "reward": 0.736196368932724,
      "reward_std": 0.046559739857912064,
      "reward_std_all": 0.057427506893873215,
      "rewards-max/forensics_sinkhorn_reward": 0.9604342877864838,
      "rewards-max/forensics_strict_edge_reward": 0.2201457917690277,
      "rewards-max/forensics_tiou_reward": 0.9066850543022156,
      "rewards/forensics_sinkhorn_reward": 0.693427711725235,
      "rewards/forensics_strict_edge_reward": 0.018902286887168884,
      "rewards/forensics_tiou_reward": 0.46324168145656586,
      "step": 566
    },
    {
      "completion_length": 22.4375,
      "epoch": 1.7403993855606759,
      "grad_norm": 7.166266441345215,
      "kl": 0.4892578125,
      "learning_rate": 5.646153846153847e-07,
      "loss": 0.0055,
      "reward": 0.7663189470767975,
      "reward_std": 0.044800445437431335,
      "reward_std_all": 0.043994395062327385,
      "rewards-max/forensics_sinkhorn_reward": 0.9834643304347992,
      "rewards-max/forensics_strict_edge_reward": 0.2401108630001545,
      "rewards-max/forensics_tiou_reward": 0.962719202041626,
      "rewards/forensics_sinkhorn_reward": 0.7324226498603821,
      "rewards/forensics_strict_edge_reward": 0.015672451816499233,
      "rewards/forensics_tiou_reward": 0.5505891442298889,
      "step": 567
    },
    {
      "completion_length": 25.140625,
      "epoch": 1.7434715821812596,
      "grad_norm": 5.669638633728027,
      "kl": 0.3759765625,
      "learning_rate": 5.638461538461539e-07,
      "loss": 0.0281,
      "reward": 0.7894324660301208,
      "reward_std": 0.02142077311873436,
      "reward_std_all": 0.021880868822336197,
      "rewards-max/forensics_sinkhorn_reward": 0.9820902347564697,
      "rewards-max/forensics_strict_edge_reward": 0.5382090210914612,
      "rewards-max/forensics_tiou_reward": 0.9555456936359406,
      "rewards/forensics_sinkhorn_reward": 0.7329898774623871,
      "rewards/forensics_strict_edge_reward": 0.037127383053302765,
      "rewards/forensics_tiou_reward": 0.5150200128555298,
      "step": 568
    },
    {
      "completion_length": 28.296875,
      "epoch": 1.7465437788018434,
      "grad_norm": 5.40162467956543,
      "kl": 0.29296875,
      "learning_rate": 5.63076923076923e-07,
      "loss": 0.0248,
      "reward": 0.7712427377700806,
      "reward_std": 0.03703276067972183,
      "reward_std_all": 0.0373521251603961,
      "rewards-max/forensics_sinkhorn_reward": 0.9379671514034271,
      "rewards-max/forensics_strict_edge_reward": 0.14368989691138268,
      "rewards-max/forensics_tiou_reward": 0.8452436029911041,
      "rewards/forensics_sinkhorn_reward": 0.7011669278144836,
      "rewards/forensics_strict_edge_reward": 0.00797124154632911,
      "rewards/forensics_tiou_reward": 0.4703487753868103,
      "step": 569
    },
    {
      "completion_length": 26.578125,
      "epoch": 1.7496159754224272,
      "grad_norm": 5.133552551269531,
      "kl": 0.427734375,
      "learning_rate": 5.623076923076923e-07,
      "loss": 0.0457,
      "reward": 0.766593337059021,
      "reward_std": 0.052114058285951614,
      "reward_std_all": 0.04900266043841839,
      "rewards-max/forensics_sinkhorn_reward": 0.9860094785690308,
      "rewards-max/forensics_strict_edge_reward": 0.5210062116384506,
      "rewards-max/forensics_tiou_reward": 0.9673907458782196,
      "rewards/forensics_sinkhorn_reward": 0.7535064816474915,
      "rewards/forensics_strict_edge_reward": 0.055710126645863056,
      "rewards/forensics_tiou_reward": 0.5749333798885345,
      "step": 570
    },
    {
      "completion_length": 27.015625,
      "epoch": 1.7526881720430108,
      "grad_norm": 4.913238048553467,
      "kl": 0.3603515625,
      "learning_rate": 5.615384615384615e-07,
      "loss": -0.0533,
      "reward": 0.8021073639392853,
      "reward_std": 0.00529122119769454,
      "reward_std_all": 0.006986838299781084,
      "rewards-max/forensics_sinkhorn_reward": 0.9681263566017151,
      "rewards-max/forensics_strict_edge_reward": 0.2303518857806921,
      "rewards-max/forensics_tiou_reward": 0.9183687269687653,
      "rewards/forensics_sinkhorn_reward": 0.7463996708393097,
      "rewards/forensics_strict_edge_reward": 0.017623850610107183,
      "rewards/forensics_tiou_reward": 0.5885926485061646,
      "step": 571
    },
    {
      "completion_length": 28.0625,
      "epoch": 1.7557603686635943,
      "grad_norm": 4.764396667480469,
      "kl": 0.4677734375,
      "learning_rate": 5.607692307692308e-07,
      "loss": 0.1215,
      "reward": 0.6803793907165527,
      "reward_std": 0.07100304961204529,
      "reward_std_all": 0.09827128425240517,
      "rewards-max/forensics_sinkhorn_reward": 0.9609432518482208,
      "rewards-max/forensics_strict_edge_reward": 0.2844267636537552,
      "rewards-max/forensics_tiou_reward": 0.91482874751091,
      "rewards/forensics_sinkhorn_reward": 0.6000319123268127,
      "rewards/forensics_strict_edge_reward": 0.023120553232729435,
      "rewards/forensics_tiou_reward": 0.39611440896987915,
      "step": 572
    },
    {
      "completion_length": 27.65625,
      "epoch": 1.758832565284178,
      "grad_norm": 4.622543811798096,
      "kl": 0.3759765625,
      "learning_rate": 5.6e-07,
      "loss": -0.0136,
      "reward": 0.7599619328975677,
      "reward_std": 0.05585963651537895,
      "reward_std_all": 0.05440334230661392,
      "rewards-max/forensics_sinkhorn_reward": 0.9819332659244537,
      "rewards-max/forensics_strict_edge_reward": 0.5935080647468567,
      "rewards-max/forensics_tiou_reward": 0.9588516056537628,
      "rewards/forensics_sinkhorn_reward": 0.731616199016571,
      "rewards/forensics_strict_edge_reward": 0.0617627389729023,
      "rewards/forensics_tiou_reward": 0.5545973181724548,
      "step": 573
    },
    {
      "completion_length": 25.09375,
      "epoch": 1.7619047619047619,
      "grad_norm": 6.326516628265381,
      "kl": 0.359375,
      "learning_rate": 5.592307692307691e-07,
      "loss": -0.0565,
      "reward": 0.7575125396251678,
      "reward_std": 0.024745337665081024,
      "reward_std_all": 0.034700275864452124,
      "rewards-max/forensics_sinkhorn_reward": 0.9774389564990997,
      "rewards-max/forensics_strict_edge_reward": 0.280071496963501,
      "rewards-max/forensics_tiou_reward": 0.942735493183136,
      "rewards/forensics_sinkhorn_reward": 0.7155863046646118,
      "rewards/forensics_strict_edge_reward": 0.03428647480905056,
      "rewards/forensics_tiou_reward": 0.48854123055934906,
      "step": 574
    },
    {
      "completion_length": 26.515625,
      "epoch": 1.7649769585253456,
      "grad_norm": 4.916798114776611,
      "kl": 0.357421875,
      "learning_rate": 5.584615384615385e-07,
      "loss": 0.0944,
      "reward": 0.7709445357322693,
      "reward_std": 0.042523717507719994,
      "reward_std_all": 0.04141637962311506,
      "rewards-max/forensics_sinkhorn_reward": 0.9856636822223663,
      "rewards-max/forensics_strict_edge_reward": 0.5397666096687317,
      "rewards-max/forensics_tiou_reward": 0.967954009771347,
      "rewards/forensics_sinkhorn_reward": 0.7165905237197876,
      "rewards/forensics_strict_edge_reward": 0.04882039874792099,
      "rewards/forensics_tiou_reward": 0.43958739936351776,
      "step": 575
    },
    {
      "completion_length": 25.734375,
      "epoch": 1.7680491551459294,
      "grad_norm": 5.34441614151001,
      "kl": 0.3486328125,
      "learning_rate": 5.576923076923077e-07,
      "loss": 0.0552,
      "reward": 0.7279796898365021,
      "reward_std": 0.07530231773853302,
      "reward_std_all": 0.07540259137749672,
      "rewards-max/forensics_sinkhorn_reward": 0.981563538312912,
      "rewards-max/forensics_strict_edge_reward": 0.27430494874715805,
      "rewards-max/forensics_tiou_reward": 0.9537293910980225,
      "rewards/forensics_sinkhorn_reward": 0.6947629451751709,
      "rewards/forensics_strict_edge_reward": 0.017607659101486206,
      "rewards/forensics_tiou_reward": 0.47289495170116425,
      "step": 576
    },
    {
      "completion_length": 27.390625,
      "epoch": 1.7711213517665132,
      "grad_norm": 7.309980869293213,
      "kl": 0.3603515625,
      "learning_rate": 5.569230769230769e-07,
      "loss": 0.1242,
      "reward": 0.676907867193222,
      "reward_std": 0.1043400838971138,
      "reward_std_all": 0.11101548001170158,
      "rewards-max/forensics_sinkhorn_reward": 0.9810923933982849,
      "rewards-max/forensics_strict_edge_reward": 0.5320244655013084,
      "rewards-max/forensics_tiou_reward": 0.955316960811615,
      "rewards/forensics_sinkhorn_reward": 0.5919920802116394,
      "rewards/forensics_strict_edge_reward": 0.03946157079190016,
      "rewards/forensics_tiou_reward": 0.4165385067462921,
      "step": 577
    },
    {
      "completion_length": 26.328125,
      "epoch": 1.7741935483870968,
      "grad_norm": 5.2018232345581055,
      "kl": 0.4052734375,
      "learning_rate": 5.561538461538461e-07,
      "loss": -0.0031,
      "reward": 0.7583701312541962,
      "reward_std": 0.029120638966560364,
      "reward_std_all": 0.03980395942926407,
      "rewards-max/forensics_sinkhorn_reward": 0.9592950940132141,
      "rewards-max/forensics_strict_edge_reward": 0.32053010910749435,
      "rewards-max/forensics_tiou_reward": 0.9012317061424255,
      "rewards/forensics_sinkhorn_reward": 0.7235084772109985,
      "rewards/forensics_strict_edge_reward": 0.030420418828725815,
      "rewards/forensics_tiou_reward": 0.47330842912197113,
      "step": 578
    },
    {
      "completion_length": 23.34375,
      "epoch": 1.7772657450076805,
      "grad_norm": 8.607769012451172,
      "kl": 0.609375,
      "learning_rate": 5.553846153846153e-07,
      "loss": 0.2218,
      "reward": 0.7360228002071381,
      "reward_std": 0.07671482488512993,
      "reward_std_all": 0.07417459040880203,
      "rewards-max/forensics_sinkhorn_reward": 0.9584429562091827,
      "rewards-max/forensics_strict_edge_reward": 0.268801212310791,
      "rewards-max/forensics_tiou_reward": 0.8841147124767303,
      "rewards/forensics_sinkhorn_reward": 0.6551516354084015,
      "rewards/forensics_strict_edge_reward": 0.020132320933043957,
      "rewards/forensics_tiou_reward": 0.39590156078338623,
      "step": 579
    },
    {
      "completion_length": 28.671875,
      "epoch": 1.780337941628264,
      "grad_norm": 5.0069355964660645,
      "kl": 0.35546875,
      "learning_rate": 5.546153846153846e-07,
      "loss": 0.1083,
      "reward": 0.7361172437667847,
      "reward_std": 0.06424953788518906,
      "reward_std_all": 0.06637866050004959,
      "rewards-max/forensics_sinkhorn_reward": 0.9859313070774078,
      "rewards-max/forensics_strict_edge_reward": 0.6425930857658386,
      "rewards-max/forensics_tiou_reward": 0.9677030742168427,
      "rewards/forensics_sinkhorn_reward": 0.6514681279659271,
      "rewards/forensics_strict_edge_reward": 0.036142606288194656,
      "rewards/forensics_tiou_reward": 0.4896819293498993,
      "step": 580
    },
    {
      "completion_length": 21.6875,
      "epoch": 1.7834101382488479,
      "grad_norm": 7.629713535308838,
      "kl": 0.5498046875,
      "learning_rate": 5.538461538461539e-07,
      "loss": 0.1567,
      "reward": 0.7512091398239136,
      "reward_std": 0.07009714283049107,
      "reward_std_all": 0.06557900365442038,
      "rewards-max/forensics_sinkhorn_reward": 0.9854287803173065,
      "rewards-max/forensics_strict_edge_reward": 0.34762609750032425,
      "rewards-max/forensics_tiou_reward": 0.9678796529769897,
      "rewards/forensics_sinkhorn_reward": 0.7148447632789612,
      "rewards/forensics_strict_edge_reward": 0.027396169491112232,
      "rewards/forensics_tiou_reward": 0.5297186821699142,
      "step": 581
    },
    {
      "completion_length": 27.84375,
      "epoch": 1.7864823348694316,
      "grad_norm": 5.874691486358643,
      "kl": 0.40234375,
      "learning_rate": 5.53076923076923e-07,
      "loss": 0.0731,
      "reward": 0.7509539127349854,
      "reward_std": 0.047314876690506935,
      "reward_std_all": 0.05114097148180008,
      "rewards-max/forensics_sinkhorn_reward": 0.9788172841072083,
      "rewards-max/forensics_strict_edge_reward": 0.4029631242156029,
      "rewards-max/forensics_tiou_reward": 0.9518814980983734,
      "rewards/forensics_sinkhorn_reward": 0.7019426226615906,
      "rewards/forensics_strict_edge_reward": 0.0224419217556715,
      "rewards/forensics_tiou_reward": 0.4861934185028076,
      "step": 582
    },
    {
      "completion_length": 25.265625,
      "epoch": 1.7895545314900154,
      "grad_norm": 6.869665622711182,
      "kl": 0.3984375,
      "learning_rate": 5.523076923076923e-07,
      "loss": 0.0905,
      "reward": 0.7541599869728088,
      "reward_std": 0.04612052999436855,
      "reward_std_all": 0.04926062002778053,
      "rewards-max/forensics_sinkhorn_reward": 0.9430858790874481,
      "rewards-max/forensics_strict_edge_reward": 0.2658524736762047,
      "rewards-max/forensics_tiou_reward": 0.8684514760971069,
      "rewards/forensics_sinkhorn_reward": 0.6871906816959381,
      "rewards/forensics_strict_edge_reward": 0.016052872873842716,
      "rewards/forensics_tiou_reward": 0.40998929738998413,
      "step": 583
    },
    {
      "completion_length": 25.859375,
      "epoch": 1.7926267281105992,
      "grad_norm": 5.918100833892822,
      "kl": 0.427734375,
      "learning_rate": 5.515384615384615e-07,
      "loss": -0.0168,
      "reward": 0.7700028717517853,
      "reward_std": 0.02868689503520727,
      "reward_std_all": 0.03211789391934872,
      "rewards-max/forensics_sinkhorn_reward": 0.9940957427024841,
      "rewards-max/forensics_strict_edge_reward": 0.5683161467313766,
      "rewards-max/forensics_tiou_reward": 0.9908126294612885,
      "rewards/forensics_sinkhorn_reward": 0.7253724336624146,
      "rewards/forensics_strict_edge_reward": 0.03024101071059704,
      "rewards/forensics_tiou_reward": 0.5414304435253143,
      "step": 584
    },
    {
      "completion_length": 23.140625,
      "epoch": 1.7956989247311828,
      "grad_norm": 4.723190784454346,
      "kl": 0.3349609375,
      "learning_rate": 5.507692307692308e-07,
      "loss": 0.0667,
      "reward": 0.7314387857913971,
      "reward_std": 0.04616040922701359,
      "reward_std_all": 0.06164884939789772,
      "rewards-max/forensics_sinkhorn_reward": 0.9151073098182678,
      "rewards-max/forensics_strict_edge_reward": 0.15301930159330368,
      "rewards-max/forensics_tiou_reward": 0.7966504991054535,
      "rewards/forensics_sinkhorn_reward": 0.670076310634613,
      "rewards/forensics_strict_edge_reward": 0.00945393624715507,
      "rewards/forensics_tiou_reward": 0.40261662006378174,
      "step": 585
    },
    {
      "completion_length": 23.078125,
      "epoch": 1.7987711213517665,
      "grad_norm": 6.854187488555908,
      "kl": 0.37890625,
      "learning_rate": 5.5e-07,
      "loss": 0.0313,
      "reward": 0.7537185251712799,
      "reward_std": 0.05544709786772728,
      "reward_std_all": 0.05498493276536465,
      "rewards-max/forensics_sinkhorn_reward": 0.9362694919109344,
      "rewards-max/forensics_strict_edge_reward": 0.1640457883477211,
      "rewards-max/forensics_tiou_reward": 0.8318842649459839,
      "rewards/forensics_sinkhorn_reward": 0.6855120658874512,
      "rewards/forensics_strict_edge_reward": 0.014043943025171757,
      "rewards/forensics_tiou_reward": 0.4508960396051407,
      "step": 586
    },
    {
      "completion_length": 22.578125,
      "epoch": 1.80184331797235,
      "grad_norm": 5.837581157684326,
      "kl": 0.3662109375,
      "learning_rate": 5.492307692307691e-07,
      "loss": 0.0057,
      "reward": 0.7504096031188965,
      "reward_std": 0.04126041755080223,
      "reward_std_all": 0.04768672212958336,
      "rewards-max/forensics_sinkhorn_reward": 0.9551413059234619,
      "rewards-max/forensics_strict_edge_reward": 0.4014057479798794,
      "rewards-max/forensics_tiou_reward": 0.8907827138900757,
      "rewards/forensics_sinkhorn_reward": 0.7213674485683441,
      "rewards/forensics_strict_edge_reward": 0.023105953820049763,
      "rewards/forensics_tiou_reward": 0.5073554962873459,
      "step": 587
    },
    {
      "completion_length": 26.640625,
      "epoch": 1.8049155145929339,
      "grad_norm": 5.574072360992432,
      "kl": 0.3837890625,
      "learning_rate": 5.484615384615385e-07,
      "loss": -0.0167,
      "reward": 0.7753615975379944,
      "reward_std": 0.033216895535588264,
      "reward_std_all": 0.03385043842718005,
      "rewards-max/forensics_sinkhorn_reward": 0.9618517756462097,
      "rewards-max/forensics_strict_edge_reward": 0.2024296298623085,
      "rewards-max/forensics_tiou_reward": 0.9150914251804352,
      "rewards/forensics_sinkhorn_reward": 0.715320497751236,
      "rewards/forensics_strict_edge_reward": 0.011196027509868145,
      "rewards/forensics_tiou_reward": 0.5221841186285019,
      "step": 588
    },
    {
      "completion_length": 24.59375,
      "epoch": 1.8079877112135176,
      "grad_norm": 4.498987674713135,
      "kl": 0.4384765625,
      "learning_rate": 5.476923076923077e-07,
      "loss": 0.054,
      "reward": 0.7577056288719177,
      "reward_std": 0.039866700768470764,
      "reward_std_all": 0.04853974934667349,
      "rewards-max/forensics_sinkhorn_reward": 0.9834258854389191,
      "rewards-max/forensics_strict_edge_reward": 0.4231441020965576,
      "rewards-max/forensics_tiou_reward": 0.9596178531646729,
      "rewards/forensics_sinkhorn_reward": 0.7226925790309906,
      "rewards/forensics_strict_edge_reward": 0.0455029159784317,
      "rewards/forensics_tiou_reward": 0.5016906261444092,
      "step": 589
    },
    {
      "completion_length": 25.0,
      "epoch": 1.8110599078341014,
      "grad_norm": 4.904358863830566,
      "kl": 0.404296875,
      "learning_rate": 5.469230769230769e-07,
      "loss": 0.0455,
      "reward": 0.7541618645191193,
      "reward_std": 0.038493343628942966,
      "reward_std_all": 0.04762344807386398,
      "rewards-max/forensics_sinkhorn_reward": 0.9809154868125916,
      "rewards-max/forensics_strict_edge_reward": 0.3484425164060667,
      "rewards-max/forensics_tiou_reward": 0.9540792405605316,
      "rewards/forensics_sinkhorn_reward": 0.7436382174491882,
      "rewards/forensics_strict_edge_reward": 0.031756882584886625,
      "rewards/forensics_tiou_reward": 0.543804943561554,
      "step": 590
    },
    {
      "completion_length": 24.875,
      "epoch": 1.8141321044546852,
      "grad_norm": 5.284805774688721,
      "kl": 0.39453125,
      "learning_rate": 5.461538461538461e-07,
      "loss": 0.0568,
      "reward": 0.7451129257678986,
      "reward_std": 0.05905763804912567,
      "reward_std_all": 0.0603405199944973,
      "rewards-max/forensics_sinkhorn_reward": 0.9821312129497528,
      "rewards-max/forensics_strict_edge_reward": 0.3552819490432739,
      "rewards-max/forensics_tiou_reward": 0.9568477272987366,
      "rewards/forensics_sinkhorn_reward": 0.6751239001750946,
      "rewards/forensics_strict_edge_reward": 0.03735855966806412,
      "rewards/forensics_tiou_reward": 0.4814952611923218,
      "step": 591
    },
    {
      "completion_length": 26.0,
      "epoch": 1.817204301075269,
      "grad_norm": 5.252058982849121,
      "kl": 0.443359375,
      "learning_rate": 5.453846153846153e-07,
      "loss": -0.0027,
      "reward": 0.7688230276107788,
      "reward_std": 0.052944084629416466,
      "reward_std_all": 0.0498075895011425,
      "rewards-max/forensics_sinkhorn_reward": 0.9917486310005188,
      "rewards-max/forensics_strict_edge_reward": 0.6360974311828613,
      "rewards-max/forensics_tiou_reward": 0.9829314947128296,
      "rewards/forensics_sinkhorn_reward": 0.7093817591667175,
      "rewards/forensics_strict_edge_reward": 0.07593419216573238,
      "rewards/forensics_tiou_reward": 0.5422843992710114,
      "step": 592
    },
    {
      "completion_length": 27.453125,
      "epoch": 1.8202764976958525,
      "grad_norm": 5.857450008392334,
      "kl": 0.296875,
      "learning_rate": 5.446153846153846e-07,
      "loss": 0.1308,
      "reward": 0.7000607252120972,
      "reward_std": 0.06839840114116669,
      "reward_std_all": 0.08430559933185577,
      "rewards-max/forensics_sinkhorn_reward": 0.9869903028011322,
      "rewards-max/forensics_strict_edge_reward": 0.5770299434661865,
      "rewards-max/forensics_tiou_reward": 0.9723982214927673,
      "rewards/forensics_sinkhorn_reward": 0.6581472754478455,
      "rewards/forensics_strict_edge_reward": 0.0432637520134449,
      "rewards/forensics_tiou_reward": 0.43168359994888306,
      "step": 593
    },
    {
      "completion_length": 24.1875,
      "epoch": 1.823348694316436,
      "grad_norm": 5.980468273162842,
      "kl": 0.580078125,
      "learning_rate": 5.438461538461538e-07,
      "loss": 0.1188,
      "reward": 0.7482396960258484,
      "reward_std": 0.03858768008649349,
      "reward_std_all": 0.04920095484703779,
      "rewards-max/forensics_sinkhorn_reward": 0.9882633984088898,
      "rewards-max/forensics_strict_edge_reward": 0.5369522273540497,
      "rewards-max/forensics_tiou_reward": 0.9699194133281708,
      "rewards/forensics_sinkhorn_reward": 0.6819364726543427,
      "rewards/forensics_strict_edge_reward": 0.05514325946569443,
      "rewards/forensics_tiou_reward": 0.45313575863838196,
      "step": 594
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.8264208909370199,
      "grad_norm": 5.0670342445373535,
      "kl": 0.310546875,
      "learning_rate": 5.430769230769231e-07,
      "loss": 0.0275,
      "reward": 0.7558062970638275,
      "reward_std": 0.05273890250828117,
      "reward_std_all": 0.05218907305970788,
      "rewards-max/forensics_sinkhorn_reward": 0.9694467782974243,
      "rewards-max/forensics_strict_edge_reward": 0.32014521956443787,
      "rewards-max/forensics_tiou_reward": 0.9205317795276642,
      "rewards/forensics_sinkhorn_reward": 0.7013765573501587,
      "rewards/forensics_strict_edge_reward": 0.030393357388675213,
      "rewards/forensics_tiou_reward": 0.46152758598327637,
      "step": 595
    },
    {
      "completion_length": 21.640625,
      "epoch": 1.8294930875576036,
      "grad_norm": 5.970157623291016,
      "kl": 0.3515625,
      "learning_rate": 5.423076923076923e-07,
      "loss": 0.0775,
      "reward": 0.7130811810493469,
      "reward_std": 0.05742263421416283,
      "reward_std_all": 0.07019926980137825,
      "rewards-max/forensics_sinkhorn_reward": 0.9921624660491943,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9878877699375153,
      "rewards/forensics_sinkhorn_reward": 0.6590893268585205,
      "rewards/forensics_strict_edge_reward": 0.046515319496393204,
      "rewards/forensics_tiou_reward": 0.4225716292858124,
      "step": 596
    },
    {
      "completion_length": 23.03125,
      "epoch": 1.8325652841781874,
      "grad_norm": 5.382920265197754,
      "kl": 0.359375,
      "learning_rate": 5.415384615384615e-07,
      "loss": -0.0109,
      "reward": 0.7580365538597107,
      "reward_std": 0.04709540028125048,
      "reward_std_all": 0.04780692793428898,
      "rewards-max/forensics_sinkhorn_reward": 0.9598555266857147,
      "rewards-max/forensics_strict_edge_reward": 0.15977148711681366,
      "rewards-max/forensics_tiou_reward": 0.9025238454341888,
      "rewards/forensics_sinkhorn_reward": 0.7391291260719299,
      "rewards/forensics_strict_edge_reward": 0.02061480563133955,
      "rewards/forensics_tiou_reward": 0.5121236443519592,
      "step": 597
    },
    {
      "completion_length": 21.359375,
      "epoch": 1.8356374807987712,
      "grad_norm": 5.977749824523926,
      "kl": 0.283203125,
      "learning_rate": 5.407692307692308e-07,
      "loss": -0.0016,
      "reward": 0.7617774307727814,
      "reward_std": 0.0572917852550745,
      "reward_std_all": 0.05385562591254711,
      "rewards-max/forensics_sinkhorn_reward": 0.986343264579773,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9687729477882385,
      "rewards/forensics_sinkhorn_reward": 0.7481010556221008,
      "rewards/forensics_strict_edge_reward": 0.059231373481452465,
      "rewards/forensics_tiou_reward": 0.558635339140892,
      "step": 598
    },
    {
      "completion_length": 26.28125,
      "epoch": 1.838709677419355,
      "grad_norm": 6.898449897766113,
      "kl": 0.384765625,
      "learning_rate": 5.4e-07,
      "loss": 0.1329,
      "reward": 0.6765479147434235,
      "reward_std": 0.08053235150873661,
      "reward_std_all": 0.10310353338718414,
      "rewards-max/forensics_sinkhorn_reward": 0.9735952913761139,
      "rewards-max/forensics_strict_edge_reward": 0.17994184605777264,
      "rewards-max/forensics_tiou_reward": 0.9298528432846069,
      "rewards/forensics_sinkhorn_reward": 0.5601877272129059,
      "rewards/forensics_strict_edge_reward": 0.017282649758271873,
      "rewards/forensics_tiou_reward": 0.33942176401615143,
      "step": 599
    },
    {
      "completion_length": 26.4375,
      "epoch": 1.8417818740399385,
      "grad_norm": 4.811708927154541,
      "kl": 0.341796875,
      "learning_rate": 5.392307692307691e-07,
      "loss": 0.0979,
      "reward": 0.7586191594600677,
      "reward_std": 0.05103913228958845,
      "reward_std_all": 0.050996556878089905,
      "rewards-max/forensics_sinkhorn_reward": 0.9867087006568909,
      "rewards-max/forensics_strict_edge_reward": 0.6115466058254242,
      "rewards-max/forensics_tiou_reward": 0.96941277384758,
      "rewards/forensics_sinkhorn_reward": 0.7300638854503632,
      "rewards/forensics_strict_edge_reward": 0.05374574288725853,
      "rewards/forensics_tiou_reward": 0.5366408228874207,
      "step": 600
    },
    {
      "completion_length": 21.796875,
      "epoch": 1.8448540706605223,
      "grad_norm": 5.284385681152344,
      "kl": 0.4013671875,
      "learning_rate": 5.384615384615384e-07,
      "loss": 0.0037,
      "reward": 0.7672060132026672,
      "reward_std": 0.031399606028571725,
      "reward_std_all": 0.03808620944619179,
      "rewards-max/forensics_sinkhorn_reward": 0.9929703772068024,
      "rewards-max/forensics_strict_edge_reward": 0.42219409346580505,
      "rewards-max/forensics_tiou_reward": 0.9867000579833984,
      "rewards/forensics_sinkhorn_reward": 0.7403473258018494,
      "rewards/forensics_strict_edge_reward": 0.029725125059485435,
      "rewards/forensics_tiou_reward": 0.4824844002723694,
      "step": 601
    },
    {
      "completion_length": 24.78125,
      "epoch": 1.8479262672811059,
      "grad_norm": 5.4776411056518555,
      "kl": 0.453125,
      "learning_rate": 5.376923076923077e-07,
      "loss": 0.0516,
      "reward": 0.7658559083938599,
      "reward_std": 0.049838099628686905,
      "reward_std_all": 0.04789822921156883,
      "rewards-max/forensics_sinkhorn_reward": 0.989162802696228,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9773522317409515,
      "rewards/forensics_sinkhorn_reward": 0.7614896297454834,
      "rewards/forensics_strict_edge_reward": 0.07272733747959137,
      "rewards/forensics_tiou_reward": 0.5671737492084503,
      "step": 602
    },
    {
      "completion_length": 23.78125,
      "epoch": 1.8509984639016897,
      "grad_norm": 5.12232780456543,
      "kl": 0.4111328125,
      "learning_rate": 5.369230769230769e-07,
      "loss": 0.0062,
      "reward": 0.7755849063396454,
      "reward_std": 0.034591817297041416,
      "reward_std_all": 0.03462024964392185,
      "rewards-max/forensics_sinkhorn_reward": 0.9909354150295258,
      "rewards-max/forensics_strict_edge_reward": 0.6032977849245071,
      "rewards-max/forensics_tiou_reward": 0.9810613691806793,
      "rewards/forensics_sinkhorn_reward": 0.726276695728302,
      "rewards/forensics_strict_edge_reward": 0.05517140403389931,
      "rewards/forensics_tiou_reward": 0.5018905997276306,
      "step": 603
    },
    {
      "completion_length": 24.96875,
      "epoch": 1.8540706605222734,
      "grad_norm": 5.1481733322143555,
      "kl": 0.42578125,
      "learning_rate": 5.361538461538462e-07,
      "loss": 0.027,
      "reward": 0.7982050776481628,
      "reward_std": 0.012782192789018154,
      "reward_std_all": 0.013360568322241306,
      "rewards-max/forensics_sinkhorn_reward": 0.9815183877944946,
      "rewards-max/forensics_strict_edge_reward": 0.5965884327888489,
      "rewards-max/forensics_tiou_reward": 0.9592312276363373,
      "rewards/forensics_sinkhorn_reward": 0.7982138097286224,
      "rewards/forensics_strict_edge_reward": 0.05051709897816181,
      "rewards/forensics_tiou_reward": 0.5555747151374817,
      "step": 604
    },
    {
      "completion_length": 22.703125,
      "epoch": 1.8571428571428572,
      "grad_norm": 5.457108020782471,
      "kl": 0.3349609375,
      "learning_rate": 5.353846153846153e-07,
      "loss": -0.0638,
      "reward": 0.799620658159256,
      "reward_std": 0.006987983477301896,
      "reward_std_all": 0.0088457849342376,
      "rewards-max/forensics_sinkhorn_reward": 0.9749033153057098,
      "rewards-max/forensics_strict_edge_reward": 0.35375696420669556,
      "rewards-max/forensics_tiou_reward": 0.9405202865600586,
      "rewards/forensics_sinkhorn_reward": 0.8053503036499023,
      "rewards/forensics_strict_edge_reward": 0.02121151238679886,
      "rewards/forensics_tiou_reward": 0.6150842905044556,
      "step": 605
    },
    {
      "completion_length": 24.078125,
      "epoch": 1.860215053763441,
      "grad_norm": 4.743715286254883,
      "kl": 0.3935546875,
      "learning_rate": 5.346153846153846e-07,
      "loss": -0.0957,
      "reward": 0.7146271467208862,
      "reward_std": 0.07182619161903858,
      "reward_std_all": 0.08121351711452007,
      "rewards-max/forensics_sinkhorn_reward": 0.9121436774730682,
      "rewards-max/forensics_strict_edge_reward": 0.3196670934557915,
      "rewards-max/forensics_tiou_reward": 0.7890157401561737,
      "rewards/forensics_sinkhorn_reward": 0.6639517247676849,
      "rewards/forensics_strict_edge_reward": 0.02424901258200407,
      "rewards/forensics_tiou_reward": 0.3859715759754181,
      "step": 606
    },
    {
      "completion_length": 24.46875,
      "epoch": 1.8632872503840245,
      "grad_norm": 4.285360813140869,
      "kl": 0.326171875,
      "learning_rate": 5.338461538461538e-07,
      "loss": -0.0431,
      "reward": 0.7884227633476257,
      "reward_std": 0.014697398990392685,
      "reward_std_all": 0.017662924714386463,
      "rewards-max/forensics_sinkhorn_reward": 0.9660170376300812,
      "rewards-max/forensics_strict_edge_reward": 0.22118929401040077,
      "rewards-max/forensics_tiou_reward": 0.9105506241321564,
      "rewards/forensics_sinkhorn_reward": 0.7731727361679077,
      "rewards/forensics_strict_edge_reward": 0.02340845437720418,
      "rewards/forensics_tiou_reward": 0.5333472192287445,
      "step": 607
    },
    {
      "completion_length": 26.28125,
      "epoch": 1.8663594470046083,
      "grad_norm": 4.2079176902771,
      "kl": 0.337890625,
      "learning_rate": 5.33076923076923e-07,
      "loss": -0.0113,
      "reward": 0.7548907101154327,
      "reward_std": 0.02120962319895625,
      "reward_std_all": 0.036860024090856314,
      "rewards-max/forensics_sinkhorn_reward": 0.9892847836017609,
      "rewards-max/forensics_strict_edge_reward": 0.7055691182613373,
      "rewards-max/forensics_tiou_reward": 0.9783600270748138,
      "rewards/forensics_sinkhorn_reward": 0.7519591748714447,
      "rewards/forensics_strict_edge_reward": 0.04038732685148716,
      "rewards/forensics_tiou_reward": 0.5794649720191956,
      "step": 608
    },
    {
      "completion_length": 25.953125,
      "epoch": 1.8694316436251919,
      "grad_norm": 6.441254615783691,
      "kl": 0.3291015625,
      "learning_rate": 5.323076923076923e-07,
      "loss": -0.0692,
      "reward": 0.7598937749862671,
      "reward_std": 0.048687687143683434,
      "reward_std_all": 0.04859470948576927,
      "rewards-max/forensics_sinkhorn_reward": 0.9776464700698853,
      "rewards-max/forensics_strict_edge_reward": 0.47380809485912323,
      "rewards-max/forensics_tiou_reward": 0.9470115602016449,
      "rewards/forensics_sinkhorn_reward": 0.7070873379707336,
      "rewards/forensics_strict_edge_reward": 0.030840665102005005,
      "rewards/forensics_tiou_reward": 0.5154140591621399,
      "step": 609
    },
    {
      "completion_length": 25.578125,
      "epoch": 1.8725038402457757,
      "grad_norm": 5.194805145263672,
      "kl": 0.380859375,
      "learning_rate": 5.315384615384615e-07,
      "loss": 0.0257,
      "reward": 0.7355412840843201,
      "reward_std": 0.03957399958744645,
      "reward_std_all": 0.054565554484725,
      "rewards-max/forensics_sinkhorn_reward": 0.973261296749115,
      "rewards-max/forensics_strict_edge_reward": 0.36141736642457545,
      "rewards-max/forensics_tiou_reward": 0.9346733093261719,
      "rewards/forensics_sinkhorn_reward": 0.6561560928821564,
      "rewards/forensics_strict_edge_reward": 0.033896749024279416,
      "rewards/forensics_tiou_reward": 0.4472033679485321,
      "step": 610
    },
    {
      "completion_length": 21.46875,
      "epoch": 1.8755760368663594,
      "grad_norm": 5.378125190734863,
      "kl": 0.3154296875,
      "learning_rate": 5.307692307692308e-07,
      "loss": 0.0319,
      "reward": 0.7805054187774658,
      "reward_std": 0.029840916395187378,
      "reward_std_all": 0.029795042239129543,
      "rewards-max/forensics_sinkhorn_reward": 0.9845223128795624,
      "rewards-max/forensics_strict_edge_reward": 0.6119300425052643,
      "rewards-max/forensics_tiou_reward": 0.9666694402694702,
      "rewards/forensics_sinkhorn_reward": 0.7579284310340881,
      "rewards/forensics_strict_edge_reward": 0.05125177279114723,
      "rewards/forensics_tiou_reward": 0.5230026841163635,
      "step": 611
    },
    {
      "completion_length": 22.296875,
      "epoch": 1.8786482334869432,
      "grad_norm": 7.348824501037598,
      "kl": 0.4609375,
      "learning_rate": 5.3e-07,
      "loss": 0.1232,
      "reward": 0.7446945607662201,
      "reward_std": 0.046797050163149834,
      "reward_std_all": 0.05308189895004034,
      "rewards-max/forensics_sinkhorn_reward": 0.9810687303543091,
      "rewards-max/forensics_strict_edge_reward": 0.5034982860088348,
      "rewards-max/forensics_tiou_reward": 0.957036018371582,
      "rewards/forensics_sinkhorn_reward": 0.6761187314987183,
      "rewards/forensics_strict_edge_reward": 0.03641210123896599,
      "rewards/forensics_tiou_reward": 0.4614957571029663,
      "step": 612
    },
    {
      "completion_length": 23.625,
      "epoch": 1.881720430107527,
      "grad_norm": 6.104162216186523,
      "kl": 0.341796875,
      "learning_rate": 5.292307692307692e-07,
      "loss": 0.0536,
      "reward": 0.7833430469036102,
      "reward_std": 0.024619121570140123,
      "reward_std_all": 0.025650168769061565,
      "rewards-max/forensics_sinkhorn_reward": 0.9833144843578339,
      "rewards-max/forensics_strict_edge_reward": 0.16178249660879374,
      "rewards-max/forensics_tiou_reward": 0.9597864151000977,
      "rewards/forensics_sinkhorn_reward": 0.765765905380249,
      "rewards/forensics_strict_edge_reward": 0.009062019176781178,
      "rewards/forensics_tiou_reward": 0.5381810963153839,
      "step": 613
    },
    {
      "completion_length": 24.75,
      "epoch": 1.8847926267281108,
      "grad_norm": 6.346436500549316,
      "kl": 0.3896484375,
      "learning_rate": 5.284615384615384e-07,
      "loss": 0.0356,
      "reward": 0.7596611082553864,
      "reward_std": 0.04925056733191013,
      "reward_std_all": 0.04920102283358574,
      "rewards-max/forensics_sinkhorn_reward": 0.9844229221343994,
      "rewards-max/forensics_strict_edge_reward": 0.36248424649238586,
      "rewards-max/forensics_tiou_reward": 0.9584486186504364,
      "rewards/forensics_sinkhorn_reward": 0.722842812538147,
      "rewards/forensics_strict_edge_reward": 0.030236564576625824,
      "rewards/forensics_tiou_reward": 0.5060622096061707,
      "step": 614
    },
    {
      "completion_length": 25.921875,
      "epoch": 1.8878648233486943,
      "grad_norm": 5.462343215942383,
      "kl": 0.46484375,
      "learning_rate": 5.276923076923076e-07,
      "loss": 0.1265,
      "reward": 0.7435231804847717,
      "reward_std": 0.06635607779026031,
      "reward_std_all": 0.06531072035431862,
      "rewards-max/forensics_sinkhorn_reward": 0.9515491127967834,
      "rewards-max/forensics_strict_edge_reward": 0.1387500250712037,
      "rewards-max/forensics_tiou_reward": 0.8783390522003174,
      "rewards/forensics_sinkhorn_reward": 0.7016010880470276,
      "rewards/forensics_strict_edge_reward": 0.014840061659924686,
      "rewards/forensics_tiou_reward": 0.425168514251709,
      "step": 615
    },
    {
      "completion_length": 26.28125,
      "epoch": 1.8909370199692779,
      "grad_norm": 6.5068840980529785,
      "kl": 0.6220703125,
      "learning_rate": 5.269230769230769e-07,
      "loss": 0.103,
      "reward": 0.7776455283164978,
      "reward_std": 0.03654483985155821,
      "reward_std_all": 0.03526275046169758,
      "rewards-max/forensics_sinkhorn_reward": 0.9809489846229553,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9590694606304169,
      "rewards/forensics_sinkhorn_reward": 0.7671748697757721,
      "rewards/forensics_strict_edge_reward": 0.05702442303299904,
      "rewards/forensics_tiou_reward": 0.5646205842494965,
      "step": 616
    },
    {
      "completion_length": 23.796875,
      "epoch": 1.8940092165898617,
      "grad_norm": 5.741851806640625,
      "kl": 0.365234375,
      "learning_rate": 5.261538461538462e-07,
      "loss": 0.0943,
      "reward": 0.749507486820221,
      "reward_std": 0.07206932082772255,
      "reward_std_all": 0.06994083151221275,
      "rewards-max/forensics_sinkhorn_reward": 0.9667642712593079,
      "rewards-max/forensics_strict_edge_reward": 0.559723362326622,
      "rewards-max/forensics_tiou_reward": 0.9216552674770355,
      "rewards/forensics_sinkhorn_reward": 0.7228781580924988,
      "rewards/forensics_strict_edge_reward": 0.05170309543609619,
      "rewards/forensics_tiou_reward": 0.5294796526432037,
      "step": 617
    },
    {
      "completion_length": 26.125,
      "epoch": 1.8970814132104454,
      "grad_norm": 7.243576526641846,
      "kl": 0.38671875,
      "learning_rate": 5.253846153846153e-07,
      "loss": -0.0436,
      "reward": 0.7756312489509583,
      "reward_std": 0.05122583732008934,
      "reward_std_all": 0.04651864431798458,
      "rewards-max/forensics_sinkhorn_reward": 0.9735682308673859,
      "rewards-max/forensics_strict_edge_reward": 0.13556911796331406,
      "rewards-max/forensics_tiou_reward": 0.9361623823642731,
      "rewards/forensics_sinkhorn_reward": 0.7617468237876892,
      "rewards/forensics_strict_edge_reward": 0.015256236307322979,
      "rewards/forensics_tiou_reward": 0.5515077412128448,
      "step": 618
    },
    {
      "completion_length": 23.078125,
      "epoch": 1.9001536098310292,
      "grad_norm": 7.1103596687316895,
      "kl": 0.3701171875,
      "learning_rate": 5.246153846153846e-07,
      "loss": -0.0008,
      "reward": 0.7560586631298065,
      "reward_std": 0.035696434788405895,
      "reward_std_all": 0.04124573431909084,
      "rewards-max/forensics_sinkhorn_reward": 0.9739470183849335,
      "rewards-max/forensics_strict_edge_reward": 0.42584793269634247,
      "rewards-max/forensics_tiou_reward": 0.9368732869625092,
      "rewards/forensics_sinkhorn_reward": 0.7089923620223999,
      "rewards/forensics_strict_edge_reward": 0.031681555323302746,
      "rewards/forensics_tiou_reward": 0.4741818606853485,
      "step": 619
    },
    {
      "completion_length": 25.09375,
      "epoch": 1.903225806451613,
      "grad_norm": 5.062434673309326,
      "kl": 0.30078125,
      "learning_rate": 5.238461538461538e-07,
      "loss": 0.1045,
      "reward": 0.7698961198329926,
      "reward_std": 0.048976694233715534,
      "reward_std_all": 0.04632169008255005,
      "rewards-max/forensics_sinkhorn_reward": 0.9753879010677338,
      "rewards-max/forensics_strict_edge_reward": 0.2554169371724129,
      "rewards-max/forensics_tiou_reward": 0.9397018253803253,
      "rewards/forensics_sinkhorn_reward": 0.6709122657775879,
      "rewards/forensics_strict_edge_reward": 0.0182630461640656,
      "rewards/forensics_tiou_reward": 0.5003257840871811,
      "step": 620
    },
    {
      "completion_length": 24.703125,
      "epoch": 1.9062980030721968,
      "grad_norm": 5.88073205947876,
      "kl": 0.3427734375,
      "learning_rate": 5.23076923076923e-07,
      "loss": 0.0222,
      "reward": 0.7824758291244507,
      "reward_std": 0.0325619475916028,
      "reward_std_all": 0.03122641146183014,
      "rewards-max/forensics_sinkhorn_reward": 0.9829620718955994,
      "rewards-max/forensics_strict_edge_reward": 0.268824752420187,
      "rewards-max/forensics_tiou_reward": 0.9532471597194672,
      "rewards/forensics_sinkhorn_reward": 0.7965605556964874,
      "rewards/forensics_strict_edge_reward": 0.03827937040477991,
      "rewards/forensics_tiou_reward": 0.5920525789260864,
      "step": 621
    },
    {
      "completion_length": 24.625,
      "epoch": 1.9093701996927803,
      "grad_norm": 4.979229927062988,
      "kl": 0.3818359375,
      "learning_rate": 5.223076923076924e-07,
      "loss": 0.0286,
      "reward": 0.7529061436653137,
      "reward_std": 0.035616835579276085,
      "reward_std_all": 0.044903626665472984,
      "rewards-max/forensics_sinkhorn_reward": 0.952574223279953,
      "rewards-max/forensics_strict_edge_reward": 0.17360299918800592,
      "rewards-max/forensics_tiou_reward": 0.8764016330242157,
      "rewards/forensics_sinkhorn_reward": 0.7229812145233154,
      "rewards/forensics_strict_edge_reward": 0.011771399294957519,
      "rewards/forensics_tiou_reward": 0.504505842924118,
      "step": 622
    },
    {
      "completion_length": 25.828125,
      "epoch": 1.912442396313364,
      "grad_norm": 5.43746280670166,
      "kl": 0.4453125,
      "learning_rate": 5.215384615384615e-07,
      "loss": -0.0073,
      "reward": 0.7636568546295166,
      "reward_std": 0.060052162036299706,
      "reward_std_all": 0.055411484092473984,
      "rewards-max/forensics_sinkhorn_reward": 0.9923249781131744,
      "rewards-max/forensics_strict_edge_reward": 0.5814422816038132,
      "rewards-max/forensics_tiou_reward": 0.9867495596408844,
      "rewards/forensics_sinkhorn_reward": 0.7710324227809906,
      "rewards/forensics_strict_edge_reward": 0.041596646420657635,
      "rewards/forensics_tiou_reward": 0.5779532194137573,
      "step": 623
    },
    {
      "completion_length": 27.890625,
      "epoch": 1.9155145929339477,
      "grad_norm": 5.815533638000488,
      "kl": 0.3388671875,
      "learning_rate": 5.207692307692308e-07,
      "loss": 0.0817,
      "reward": 0.7365503013134003,
      "reward_std": 0.06998554989695549,
      "reward_std_all": 0.06973868981003761,
      "rewards-max/forensics_sinkhorn_reward": 0.9883557260036469,
      "rewards-max/forensics_strict_edge_reward": 0.4231441020965576,
      "rewards-max/forensics_tiou_reward": 0.9730247259140015,
      "rewards/forensics_sinkhorn_reward": 0.6424052119255066,
      "rewards/forensics_strict_edge_reward": 0.030168931931257248,
      "rewards/forensics_tiou_reward": 0.4547918438911438,
      "step": 624
    },
    {
      "completion_length": 31.90625,
      "epoch": 1.9185867895545314,
      "grad_norm": 5.011209487915039,
      "kl": 0.3671875,
      "learning_rate": 5.2e-07,
      "loss": 0.0698,
      "reward": 0.7090529799461365,
      "reward_std": 0.05737231019884348,
      "reward_std_all": 0.07773041352629662,
      "rewards-max/forensics_sinkhorn_reward": 0.955566018819809,
      "rewards-max/forensics_strict_edge_reward": 0.12756267562508583,
      "rewards-max/forensics_tiou_reward": 0.897691160440445,
      "rewards/forensics_sinkhorn_reward": 0.6493092179298401,
      "rewards/forensics_strict_edge_reward": 0.011875365860760212,
      "rewards/forensics_tiou_reward": 0.47261491417884827,
      "step": 625
    },
    {
      "completion_length": 24.421875,
      "epoch": 1.9216589861751152,
      "grad_norm": 5.142571449279785,
      "kl": 0.3701171875,
      "learning_rate": 5.192307692307692e-07,
      "loss": 0.0276,
      "reward": 0.7740842700004578,
      "reward_std": 0.05459245853126049,
      "reward_std_all": 0.049256423488259315,
      "rewards-max/forensics_sinkhorn_reward": 0.9808200001716614,
      "rewards-max/forensics_strict_edge_reward": 0.22590027749538422,
      "rewards-max/forensics_tiou_reward": 0.9565664529800415,
      "rewards/forensics_sinkhorn_reward": 0.7707669138908386,
      "rewards/forensics_strict_edge_reward": 0.018661753740161657,
      "rewards/forensics_tiou_reward": 0.5718384981155396,
      "step": 626
    },
    {
      "completion_length": 26.375,
      "epoch": 1.924731182795699,
      "grad_norm": 5.22344970703125,
      "kl": 0.322265625,
      "learning_rate": 5.184615384615384e-07,
      "loss": 0.0137,
      "reward": 0.7788506746292114,
      "reward_std": 0.03804885409772396,
      "reward_std_all": 0.0363033888861537,
      "rewards-max/forensics_sinkhorn_reward": 0.9830294549465179,
      "rewards-max/forensics_strict_edge_reward": 0.3647812753915787,
      "rewards-max/forensics_tiou_reward": 0.960789829492569,
      "rewards/forensics_sinkhorn_reward": 0.7087415754795074,
      "rewards/forensics_strict_edge_reward": 0.025857185944914818,
      "rewards/forensics_tiou_reward": 0.5974686145782471,
      "step": 627
    },
    {
      "completion_length": 25.609375,
      "epoch": 1.9278033794162828,
      "grad_norm": 4.261199474334717,
      "kl": 0.322265625,
      "learning_rate": 5.176923076923076e-07,
      "loss": 0.0103,
      "reward": 0.7869894802570343,
      "reward_std": 0.023145086131989956,
      "reward_std_all": 0.02348837163299322,
      "rewards-max/forensics_sinkhorn_reward": 0.9621773660182953,
      "rewards-max/forensics_strict_edge_reward": 0.24313345551490784,
      "rewards-max/forensics_tiou_reward": 0.9097220003604889,
      "rewards/forensics_sinkhorn_reward": 0.7874064743518829,
      "rewards/forensics_strict_edge_reward": 0.023352337535470724,
      "rewards/forensics_tiou_reward": 0.5657999515533447,
      "step": 628
    },
    {
      "completion_length": 24.140625,
      "epoch": 1.9308755760368663,
      "grad_norm": 6.817155838012695,
      "kl": 0.3759765625,
      "learning_rate": 5.16923076923077e-07,
      "loss": -0.0119,
      "reward": 0.7503069043159485,
      "reward_std": 0.03298209421336651,
      "reward_std_all": 0.04552118293941021,
      "rewards-max/forensics_sinkhorn_reward": 0.9846028089523315,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9656705260276794,
      "rewards/forensics_sinkhorn_reward": 0.712382972240448,
      "rewards/forensics_strict_edge_reward": 0.04658036306500435,
      "rewards/forensics_tiou_reward": 0.5103558897972107,
      "step": 629
    },
    {
      "completion_length": 27.8125,
      "epoch": 1.93394777265745,
      "grad_norm": 5.289923191070557,
      "kl": 0.423828125,
      "learning_rate": 5.161538461538462e-07,
      "loss": -0.0061,
      "reward": 0.7308186292648315,
      "reward_std": 0.06365714129060507,
      "reward_std_all": 0.06962372176349163,
      "rewards-max/forensics_sinkhorn_reward": 0.9770331382751465,
      "rewards-max/forensics_strict_edge_reward": 0.3690408021211624,
      "rewards-max/forensics_tiou_reward": 0.9500894248485565,
      "rewards/forensics_sinkhorn_reward": 0.6569536030292511,
      "rewards/forensics_strict_edge_reward": 0.0321237463504076,
      "rewards/forensics_tiou_reward": 0.4789688140153885,
      "step": 630
    },
    {
      "completion_length": 27.375,
      "epoch": 1.9370199692780337,
      "grad_norm": 6.722125053405762,
      "kl": 0.4091796875,
      "learning_rate": 5.153846153846153e-07,
      "loss": 0.1058,
      "reward": 0.7201560437679291,
      "reward_std": 0.04050253768218681,
      "reward_std_all": 0.0650899838656187,
      "rewards-max/forensics_sinkhorn_reward": 0.9649922549724579,
      "rewards-max/forensics_strict_edge_reward": 0.30686262249946594,
      "rewards-max/forensics_tiou_reward": 0.9161774218082428,
      "rewards/forensics_sinkhorn_reward": 0.6578367352485657,
      "rewards/forensics_strict_edge_reward": 0.019108748994767666,
      "rewards/forensics_tiou_reward": 0.5013791769742966,
      "step": 631
    },
    {
      "completion_length": 22.984375,
      "epoch": 1.9400921658986174,
      "grad_norm": 5.29190731048584,
      "kl": 0.421875,
      "learning_rate": 5.146153846153846e-07,
      "loss": 0.0463,
      "reward": 0.7482712268829346,
      "reward_std": 0.044385751709342,
      "reward_std_all": 0.049326105043292046,
      "rewards-max/forensics_sinkhorn_reward": 0.9380453526973724,
      "rewards-max/forensics_strict_edge_reward": 0.1581155825406313,
      "rewards-max/forensics_tiou_reward": 0.8424150943756104,
      "rewards/forensics_sinkhorn_reward": 0.6644107699394226,
      "rewards/forensics_strict_edge_reward": 0.0071834619011497125,
      "rewards/forensics_tiou_reward": 0.3944410979747772,
      "step": 632
    },
    {
      "completion_length": 23.25,
      "epoch": 1.9431643625192012,
      "grad_norm": 6.19614839553833,
      "kl": 0.375,
      "learning_rate": 5.138461538461538e-07,
      "loss": 0.0469,
      "reward": 0.7070446610450745,
      "reward_std": 0.06810343265533447,
      "reward_std_all": 0.08242588862776756,
      "rewards-max/forensics_sinkhorn_reward": 0.9522417187690735,
      "rewards-max/forensics_strict_edge_reward": 0.13787118904292583,
      "rewards-max/forensics_tiou_reward": 0.8796470165252686,
      "rewards/forensics_sinkhorn_reward": 0.6267065405845642,
      "rewards/forensics_strict_edge_reward": 0.017617180477827787,
      "rewards/forensics_tiou_reward": 0.38561928272247314,
      "step": 633
    },
    {
      "completion_length": 23.84375,
      "epoch": 1.946236559139785,
      "grad_norm": 5.265752792358398,
      "kl": 0.361328125,
      "learning_rate": 5.13076923076923e-07,
      "loss": -0.0049,
      "reward": 0.7828393876552582,
      "reward_std": 0.03523298352956772,
      "reward_std_all": 0.03328322060406208,
      "rewards-max/forensics_sinkhorn_reward": 0.9896545708179474,
      "rewards-max/forensics_strict_edge_reward": 0.7936499714851379,
      "rewards-max/forensics_tiou_reward": 0.9808257520198822,
      "rewards/forensics_sinkhorn_reward": 0.7669607400894165,
      "rewards/forensics_strict_edge_reward": 0.061900217086076736,
      "rewards/forensics_tiou_reward": 0.4978463500738144,
      "step": 634
    },
    {
      "completion_length": 27.28125,
      "epoch": 1.9493087557603688,
      "grad_norm": 6.3183794021606445,
      "kl": 0.421875,
      "learning_rate": 5.123076923076923e-07,
      "loss": -0.0421,
      "reward": 0.742938220500946,
      "reward_std": 0.041427724063396454,
      "reward_std_all": 0.05295524373650551,
      "rewards-max/forensics_sinkhorn_reward": 0.9812864661216736,
      "rewards-max/forensics_strict_edge_reward": 0.44775912165641785,
      "rewards-max/forensics_tiou_reward": 0.9578795731067657,
      "rewards/forensics_sinkhorn_reward": 0.6871626675128937,
      "rewards/forensics_strict_edge_reward": 0.027077436447143555,
      "rewards/forensics_tiou_reward": 0.42496873438358307,
      "step": 635
    },
    {
      "completion_length": 22.875,
      "epoch": 1.9523809523809523,
      "grad_norm": 6.362329959869385,
      "kl": 0.3349609375,
      "learning_rate": 5.115384615384615e-07,
      "loss": 0.0276,
      "reward": 0.7883692979812622,
      "reward_std": 0.030267623253166676,
      "reward_std_all": 0.027788780629634857,
      "rewards-max/forensics_sinkhorn_reward": 0.9845125675201416,
      "rewards-max/forensics_strict_edge_reward": 0.4260037988424301,
      "rewards-max/forensics_tiou_reward": 0.9661080539226532,
      "rewards/forensics_sinkhorn_reward": 0.7527580559253693,
      "rewards/forensics_strict_edge_reward": 0.0503925527445972,
      "rewards/forensics_tiou_reward": 0.5848005414009094,
      "step": 636
    },
    {
      "completion_length": 28.75,
      "epoch": 1.955453149001536,
      "grad_norm": 5.427719593048096,
      "kl": 0.3994140625,
      "learning_rate": 5.107692307692308e-07,
      "loss": 0.1415,
      "reward": 0.7419049143791199,
      "reward_std": 0.06642723828554153,
      "reward_std_all": 0.06781450659036636,
      "rewards-max/forensics_sinkhorn_reward": 0.9331259429454803,
      "rewards-max/forensics_strict_edge_reward": 0.3288623634725809,
      "rewards-max/forensics_tiou_reward": 0.8362787067890167,
      "rewards/forensics_sinkhorn_reward": 0.6519513428211212,
      "rewards/forensics_strict_edge_reward": 0.023145576007664204,
      "rewards/forensics_tiou_reward": 0.4333091229200363,
      "step": 637
    },
    {
      "completion_length": 28.375,
      "epoch": 1.9585253456221197,
      "grad_norm": 5.407547950744629,
      "kl": 0.4775390625,
      "learning_rate": 5.1e-07,
      "loss": 0.0428,
      "reward": 0.7807862162590027,
      "reward_std": 0.0292350216768682,
      "reward_std_all": 0.02972117392346263,
      "rewards-max/forensics_sinkhorn_reward": 0.9494788944721222,
      "rewards-max/forensics_strict_edge_reward": 0.18178267497569323,
      "rewards-max/forensics_tiou_reward": 0.8747754395008087,
      "rewards/forensics_sinkhorn_reward": 0.7193675637245178,
      "rewards/forensics_strict_edge_reward": 0.013848980859620497,
      "rewards/forensics_tiou_reward": 0.5044484734535217,
      "step": 638
    },
    {
      "completion_length": 25.8125,
      "epoch": 1.9615975422427034,
      "grad_norm": 5.524186134338379,
      "kl": 0.2998046875,
      "learning_rate": 5.092307692307692e-07,
      "loss": 0.0253,
      "reward": 0.7920220196247101,
      "reward_std": 0.023186085047200322,
      "reward_std_all": 0.022356663830578327,
      "rewards-max/forensics_sinkhorn_reward": 0.9837207794189453,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9642495810985565,
      "rewards/forensics_sinkhorn_reward": 0.7870199084281921,
      "rewards/forensics_strict_edge_reward": 0.06922566145658493,
      "rewards/forensics_tiou_reward": 0.5633634030818939,
      "step": 639
    },
    {
      "completion_length": 21.71875,
      "epoch": 1.9646697388632872,
      "grad_norm": 7.098333358764648,
      "kl": 0.4287109375,
      "learning_rate": 5.084615384615384e-07,
      "loss": -0.0123,
      "reward": 0.7766428589820862,
      "reward_std": 0.038834819570183754,
      "reward_std_all": 0.03718637302517891,
      "rewards-max/forensics_sinkhorn_reward": 0.9832746088504791,
      "rewards-max/forensics_strict_edge_reward": 0.4753422290086746,
      "rewards-max/forensics_tiou_reward": 0.9603437185287476,
      "rewards/forensics_sinkhorn_reward": 0.7453856766223907,
      "rewards/forensics_strict_edge_reward": 0.05276171304285526,
      "rewards/forensics_tiou_reward": 0.5429944396018982,
      "step": 640
    },
    {
      "completion_length": 23.3125,
      "epoch": 1.967741935483871,
      "grad_norm": 6.698963165283203,
      "kl": 0.30859375,
      "learning_rate": 5.076923076923076e-07,
      "loss": -0.058,
      "reward": 0.7764175534248352,
      "reward_std": 0.032457662746310234,
      "reward_std_all": 0.032521527260541916,
      "rewards-max/forensics_sinkhorn_reward": 0.9781407117843628,
      "rewards-max/forensics_strict_edge_reward": 0.38808465003967285,
      "rewards-max/forensics_tiou_reward": 0.9527751207351685,
      "rewards/forensics_sinkhorn_reward": 0.7585732936859131,
      "rewards/forensics_strict_edge_reward": 0.06341751664876938,
      "rewards/forensics_tiou_reward": 0.5253631621599197,
      "step": 641
    },
    {
      "completion_length": 24.390625,
      "epoch": 1.9708141321044548,
      "grad_norm": 6.197840690612793,
      "kl": 0.462890625,
      "learning_rate": 5.06923076923077e-07,
      "loss": -0.0493,
      "reward": 0.7873499691486359,
      "reward_std": 0.023676077369600534,
      "reward_std_all": 0.023738424759358168,
      "rewards-max/forensics_sinkhorn_reward": 0.9878226816654205,
      "rewards-max/forensics_strict_edge_reward": 0.5599738955497742,
      "rewards-max/forensics_tiou_reward": 0.9724946618080139,
      "rewards/forensics_sinkhorn_reward": 0.7827869951725006,
      "rewards/forensics_strict_edge_reward": 0.08377857133746147,
      "rewards/forensics_tiou_reward": 0.5816984176635742,
      "step": 642
    },
    {
      "completion_length": 23.296875,
      "epoch": 1.9738863287250386,
      "grad_norm": 9.134550094604492,
      "kl": 0.36328125,
      "learning_rate": 5.061538461538462e-07,
      "loss": 0.0434,
      "reward": 0.7625733613967896,
      "reward_std": 0.052275411784648895,
      "reward_std_all": 0.05055813118815422,
      "rewards-max/forensics_sinkhorn_reward": 0.9513196349143982,
      "rewards-max/forensics_strict_edge_reward": 0.1466790735721588,
      "rewards-max/forensics_tiou_reward": 0.8800055980682373,
      "rewards/forensics_sinkhorn_reward": 0.6732776761054993,
      "rewards/forensics_strict_edge_reward": 0.009064373094588518,
      "rewards/forensics_tiou_reward": 0.4353553354740143,
      "step": 643
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.976958525345622,
      "grad_norm": 5.356232166290283,
      "kl": 0.3173828125,
      "learning_rate": 5.053846153846154e-07,
      "loss": -0.0266,
      "reward": 0.7525729238986969,
      "reward_std": 0.019398442469537258,
      "reward_std_all": 0.0355590283870697,
      "rewards-max/forensics_sinkhorn_reward": 0.9792956411838531,
      "rewards-max/forensics_strict_edge_reward": 0.4737038314342499,
      "rewards-max/forensics_tiou_reward": 0.942063182592392,
      "rewards/forensics_sinkhorn_reward": 0.7451864182949066,
      "rewards/forensics_strict_edge_reward": 0.030532106757164,
      "rewards/forensics_tiou_reward": 0.5377323627471924,
      "step": 644
    },
    {
      "completion_length": 29.390625,
      "epoch": 1.9800307219662059,
      "grad_norm": 6.374257564544678,
      "kl": 0.431640625,
      "learning_rate": 5.046153846153846e-07,
      "loss": 0.0603,
      "reward": 0.7479479014873505,
      "reward_std": 0.058973388746380806,
      "reward_std_all": 0.05905279889702797,
      "rewards-max/forensics_sinkhorn_reward": 0.9739878177642822,
      "rewards-max/forensics_strict_edge_reward": 0.5900419056415558,
      "rewards-max/forensics_tiou_reward": 0.9450136423110962,
      "rewards/forensics_sinkhorn_reward": 0.6951267421245575,
      "rewards/forensics_strict_edge_reward": 0.06935230456292629,
      "rewards/forensics_tiou_reward": 0.5650194585323334,
      "step": 645
    },
    {
      "completion_length": 26.515625,
      "epoch": 1.9831029185867894,
      "grad_norm": 5.650396823883057,
      "kl": 0.3828125,
      "learning_rate": 5.038461538461538e-07,
      "loss": 0.1336,
      "reward": 0.7821750044822693,
      "reward_std": 0.02861960418522358,
      "reward_std_all": 0.02835945412516594,
      "rewards-max/forensics_sinkhorn_reward": 0.9833140671253204,
      "rewards-max/forensics_strict_edge_reward": 0.42263756692409515,
      "rewards-max/forensics_tiou_reward": 0.9558443129062653,
      "rewards/forensics_sinkhorn_reward": 0.7303488552570343,
      "rewards/forensics_strict_edge_reward": 0.03916489891707897,
      "rewards/forensics_tiou_reward": 0.5506555438041687,
      "step": 646
    },
    {
      "completion_length": 25.53125,
      "epoch": 1.9861751152073732,
      "grad_norm": 5.122254371643066,
      "kl": 0.27685546875,
      "learning_rate": 5.03076923076923e-07,
      "loss": -0.0521,
      "reward": 0.77137690782547,
      "reward_std": 0.03288980294018984,
      "reward_std_all": 0.03494279272854328,
      "rewards-max/forensics_sinkhorn_reward": 0.9762158691883087,
      "rewards-max/forensics_strict_edge_reward": 0.31747324764728546,
      "rewards-max/forensics_tiou_reward": 0.9370267987251282,
      "rewards/forensics_sinkhorn_reward": 0.749546229839325,
      "rewards/forensics_strict_edge_reward": 0.025966588407754898,
      "rewards/forensics_tiou_reward": 0.5314816832542419,
      "step": 647
    },
    {
      "completion_length": 26.84375,
      "epoch": 1.989247311827957,
      "grad_norm": 5.343527317047119,
      "kl": 0.466796875,
      "learning_rate": 5.023076923076923e-07,
      "loss": 0.0222,
      "reward": 0.7428678274154663,
      "reward_std": 0.08018408343195915,
      "reward_std_all": 0.07561270520091057,
      "rewards-max/forensics_sinkhorn_reward": 0.9737726151943207,
      "rewards-max/forensics_strict_edge_reward": 0.26586400903761387,
      "rewards-max/forensics_tiou_reward": 0.9390472769737244,
      "rewards/forensics_sinkhorn_reward": 0.6803601086139679,
      "rewards/forensics_strict_edge_reward": 0.03599012363702059,
      "rewards/forensics_tiou_reward": 0.5257508754730225,
      "step": 648
    },
    {
      "completion_length": 27.109375,
      "epoch": 1.9923195084485408,
      "grad_norm": 5.0102009773254395,
      "kl": 0.2998046875,
      "learning_rate": 5.015384615384614e-07,
      "loss": -0.0328,
      "reward": 0.7446140646934509,
      "reward_std": 0.03303886856883764,
      "reward_std_all": 0.04878545831888914,
      "rewards-max/forensics_sinkhorn_reward": 0.9781855344772339,
      "rewards-max/forensics_strict_edge_reward": 0.27970148622989655,
      "rewards-max/forensics_tiou_reward": 0.9501744508743286,
      "rewards/forensics_sinkhorn_reward": 0.7005345821380615,
      "rewards/forensics_strict_edge_reward": 0.035057776141911745,
      "rewards/forensics_tiou_reward": 0.5352218449115753,
      "step": 649
    },
    {
      "completion_length": 22.921875,
      "epoch": 1.9953917050691246,
      "grad_norm": 6.066019058227539,
      "kl": 0.3916015625,
      "learning_rate": 5.007692307692308e-07,
      "loss": 0.1162,
      "reward": 0.7256382405757904,
      "reward_std": 0.057794949039816856,
      "reward_std_all": 0.0682445615530014,
      "rewards-max/forensics_sinkhorn_reward": 0.9915446937084198,
      "rewards-max/forensics_strict_edge_reward": 0.19096612185239792,
      "rewards-max/forensics_tiou_reward": 0.9811938405036926,
      "rewards/forensics_sinkhorn_reward": 0.633162796497345,
      "rewards/forensics_strict_edge_reward": 0.022632166743278503,
      "rewards/forensics_tiou_reward": 0.45522989332675934,
      "step": 650
    },
    {
      "completion_length": 25.4375,
      "epoch": 1.998463901689708,
      "grad_norm": 7.020522594451904,
      "kl": 0.369140625,
      "learning_rate": 5e-07,
      "loss": 0.0087,
      "reward": 0.7662260830402374,
      "reward_std": 0.04449391085654497,
      "reward_std_all": 0.044467743486166,
      "rewards-max/forensics_sinkhorn_reward": 0.9923118054866791,
      "rewards-max/forensics_strict_edge_reward": 0.40228159725666046,
      "rewards-max/forensics_tiou_reward": 0.9844931960105896,
      "rewards/forensics_sinkhorn_reward": 0.7496404051780701,
      "rewards/forensics_strict_edge_reward": 0.04433223232626915,
      "rewards/forensics_tiou_reward": 0.5599821209907532,
      "step": 651
    },
    {
      "completion_length": 24.600000381469727,
      "epoch": 2.0,
      "grad_norm": 4.356888294219971,
      "kl": 0.6015625,
      "learning_rate": 4.992307692307692e-07,
      "loss": 0.0352,
      "reward": 0.6749808192253113,
      "reward_std": 0.08142394572496414,
      "reward_std_all": 0.12023214250802994,
      "rewards-max/forensics_sinkhorn_reward": 0.9275057911872864,
      "rewards-max/forensics_strict_edge_reward": 0.2018965184688568,
      "rewards-max/forensics_tiou_reward": 0.8391443490982056,
      "rewards/forensics_sinkhorn_reward": 0.5998515486717224,
      "rewards/forensics_strict_edge_reward": 0.06804047524929047,
      "rewards/forensics_tiou_reward": 0.39640626311302185,
      "step": 652
    },
    {
      "completion_length": 21.8125,
      "epoch": 2.003072196620584,
      "grad_norm": 5.850307941436768,
      "kl": 0.34765625,
      "learning_rate": 4.984615384615384e-07,
      "loss": -0.0641,
      "reward": 0.7643648386001587,
      "reward_std": 0.039182813838124275,
      "reward_std_all": 0.04530028533190489,
      "rewards-max/forensics_sinkhorn_reward": 0.9849264323711395,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9707494378089905,
      "rewards/forensics_sinkhorn_reward": 0.6798997819423676,
      "rewards/forensics_strict_edge_reward": 0.052043044939637184,
      "rewards/forensics_tiou_reward": 0.4935269355773926,
      "step": 653
    },
    {
      "completion_length": 24.25,
      "epoch": 2.0061443932411676,
      "grad_norm": 6.364166259765625,
      "kl": 0.3818359375,
      "learning_rate": 4.976923076923077e-07,
      "loss": 0.0397,
      "reward": 0.7752638757228851,
      "reward_std": 0.04629095457494259,
      "reward_std_all": 0.04279227927327156,
      "rewards-max/forensics_sinkhorn_reward": 0.9947661757469177,
      "rewards-max/forensics_strict_edge_reward": 0.5793185830116272,
      "rewards-max/forensics_tiou_reward": 0.9894151985645294,
      "rewards/forensics_sinkhorn_reward": 0.7578893005847931,
      "rewards/forensics_strict_edge_reward": 0.04726458713412285,
      "rewards/forensics_tiou_reward": 0.5219897031784058,
      "step": 654
    },
    {
      "completion_length": 28.484375,
      "epoch": 2.0092165898617513,
      "grad_norm": 4.416747093200684,
      "kl": 0.529296875,
      "learning_rate": 4.969230769230769e-07,
      "loss": 0.1064,
      "reward": 0.7580888271331787,
      "reward_std": 0.04637854639440775,
      "reward_std_all": 0.05245702713727951,
      "rewards-max/forensics_sinkhorn_reward": 0.9701920449733734,
      "rewards-max/forensics_strict_edge_reward": 0.1955089271068573,
      "rewards-max/forensics_tiou_reward": 0.9058466255664825,
      "rewards/forensics_sinkhorn_reward": 0.7084530889987946,
      "rewards/forensics_strict_edge_reward": 0.016594558721408248,
      "rewards/forensics_tiou_reward": 0.5174103081226349,
      "step": 655
    },
    {
      "completion_length": 24.484375,
      "epoch": 2.0122887864823347,
      "grad_norm": 6.414154529571533,
      "kl": 0.4267578125,
      "learning_rate": 4.961538461538462e-07,
      "loss": 0.0534,
      "reward": 0.7661494612693787,
      "reward_std": 0.05368353798985481,
      "reward_std_all": 0.050608277320861816,
      "rewards-max/forensics_sinkhorn_reward": 0.9837840497493744,
      "rewards-max/forensics_strict_edge_reward": 0.5192448049783707,
      "rewards-max/forensics_tiou_reward": 0.9631819725036621,
      "rewards/forensics_sinkhorn_reward": 0.7339686155319214,
      "rewards/forensics_strict_edge_reward": 0.05338452197611332,
      "rewards/forensics_tiou_reward": 0.5589892417192459,
      "step": 656
    },
    {
      "completion_length": 25.515625,
      "epoch": 2.0153609831029184,
      "grad_norm": 4.524357318878174,
      "kl": 0.3720703125,
      "learning_rate": 4.953846153846154e-07,
      "loss": 0.1311,
      "reward": 0.7488756477832794,
      "reward_std": 0.019443731755018234,
      "reward_std_all": 0.04008463956415653,
      "rewards-max/forensics_sinkhorn_reward": 0.996032327413559,
      "rewards-max/forensics_strict_edge_reward": 0.796825110912323,
      "rewards-max/forensics_tiou_reward": 0.9939227402210236,
      "rewards/forensics_sinkhorn_reward": 0.7144583463668823,
      "rewards/forensics_strict_edge_reward": 0.0691758431494236,
      "rewards/forensics_tiou_reward": 0.5439845323562622,
      "step": 657
    },
    {
      "completion_length": 26.234375,
      "epoch": 2.0184331797235022,
      "grad_norm": 4.738323211669922,
      "kl": 0.404296875,
      "learning_rate": 4.946153846153846e-07,
      "loss": 0.0855,
      "reward": 0.766854465007782,
      "reward_std": 0.0403280658647418,
      "reward_std_all": 0.04195873625576496,
      "rewards-max/forensics_sinkhorn_reward": 0.973727822303772,
      "rewards-max/forensics_strict_edge_reward": 0.38279129564762115,
      "rewards-max/forensics_tiou_reward": 0.9398378431797028,
      "rewards/forensics_sinkhorn_reward": 0.7113297879695892,
      "rewards/forensics_strict_edge_reward": 0.03413805738091469,
      "rewards/forensics_tiou_reward": 0.5060068666934967,
      "step": 658
    },
    {
      "completion_length": 24.296875,
      "epoch": 2.021505376344086,
      "grad_norm": 5.740506649017334,
      "kl": 0.3388671875,
      "learning_rate": 4.938461538461538e-07,
      "loss": 0.0309,
      "reward": 0.7539239227771759,
      "reward_std": 0.04566907323896885,
      "reward_std_all": 0.04852093290537596,
      "rewards-max/forensics_sinkhorn_reward": 0.9773313105106354,
      "rewards-max/forensics_strict_edge_reward": 0.26213184371590614,
      "rewards-max/forensics_tiou_reward": 0.947479635477066,
      "rewards/forensics_sinkhorn_reward": 0.7134034633636475,
      "rewards/forensics_strict_edge_reward": 0.023948210524395108,
      "rewards/forensics_tiou_reward": 0.5346387922763824,
      "step": 659
    },
    {
      "completion_length": 24.421875,
      "epoch": 2.02457757296467,
      "grad_norm": 5.981856822967529,
      "kl": 0.3447265625,
      "learning_rate": 4.93076923076923e-07,
      "loss": 0.038,
      "reward": 0.7553362846374512,
      "reward_std": 0.04191174264997244,
      "reward_std_all": 0.0489758076146245,
      "rewards-max/forensics_sinkhorn_reward": 0.9830419421195984,
      "rewards-max/forensics_strict_edge_reward": 0.30642860382795334,
      "rewards-max/forensics_tiou_reward": 0.9613743722438812,
      "rewards/forensics_sinkhorn_reward": 0.6964603662490845,
      "rewards/forensics_strict_edge_reward": 0.04426236730068922,
      "rewards/forensics_tiou_reward": 0.4717090427875519,
      "step": 660
    },
    {
      "completion_length": 26.828125,
      "epoch": 2.0276497695852536,
      "grad_norm": 4.920258045196533,
      "kl": 0.412109375,
      "learning_rate": 4.923076923076923e-07,
      "loss": -0.0656,
      "reward": 0.7586578726768494,
      "reward_std": 0.028592298738658428,
      "reward_std_all": 0.037899866700172424,
      "rewards-max/forensics_sinkhorn_reward": 0.9659167230129242,
      "rewards-max/forensics_strict_edge_reward": 0.40708814561367035,
      "rewards-max/forensics_tiou_reward": 0.9215222597122192,
      "rewards/forensics_sinkhorn_reward": 0.7296228110790253,
      "rewards/forensics_strict_edge_reward": 0.02573551796376705,
      "rewards/forensics_tiou_reward": 0.48175449669361115,
      "step": 661
    },
    {
      "completion_length": 22.5,
      "epoch": 2.0307219662058373,
      "grad_norm": 6.134034156799316,
      "kl": 0.5361328125,
      "learning_rate": 4.915384615384615e-07,
      "loss": 0.0551,
      "reward": 0.7695485055446625,
      "reward_std": 0.024560962803661823,
      "reward_std_all": 0.033374859020113945,
      "rewards-max/forensics_sinkhorn_reward": 0.9868277907371521,
      "rewards-max/forensics_strict_edge_reward": 0.49102427065372467,
      "rewards-max/forensics_tiou_reward": 0.9698914289474487,
      "rewards/forensics_sinkhorn_reward": 0.7620588839054108,
      "rewards/forensics_strict_edge_reward": 0.06469373777508736,
      "rewards/forensics_tiou_reward": 0.5085772126913071,
      "step": 662
    },
    {
      "completion_length": 25.234375,
      "epoch": 2.033794162826421,
      "grad_norm": 7.426711559295654,
      "kl": 0.4111328125,
      "learning_rate": 4.907692307692308e-07,
      "loss": -0.023,
      "reward": 0.7970339059829712,
      "reward_std": 0.01190023822709918,
      "reward_std_all": 0.012997314799576998,
      "rewards-max/forensics_sinkhorn_reward": 0.9818932712078094,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9595144987106323,
      "rewards/forensics_sinkhorn_reward": 0.7795784771442413,
      "rewards/forensics_strict_edge_reward": 0.03650965727865696,
      "rewards/forensics_tiou_reward": 0.6031120717525482,
      "step": 663
    },
    {
      "completion_length": 27.828125,
      "epoch": 2.0368663594470044,
      "grad_norm": 4.959684371948242,
      "kl": 0.3330078125,
      "learning_rate": 4.9e-07,
      "loss": 0.0568,
      "reward": 0.780320942401886,
      "reward_std": 0.03562572970986366,
      "reward_std_all": 0.034098897129297256,
      "rewards-max/forensics_sinkhorn_reward": 0.9475997984409332,
      "rewards-max/forensics_strict_edge_reward": 0.218518509529531,
      "rewards-max/forensics_tiou_reward": 0.8638647794723511,
      "rewards/forensics_sinkhorn_reward": 0.7476129531860352,
      "rewards/forensics_strict_edge_reward": 0.03554254793561995,
      "rewards/forensics_tiou_reward": 0.53925821185112,
      "step": 664
    },
    {
      "completion_length": 27.453125,
      "epoch": 2.0399385560675882,
      "grad_norm": 5.492393970489502,
      "kl": 0.3349609375,
      "learning_rate": 4.892307692307692e-07,
      "loss": 0.0402,
      "reward": 0.8007759153842926,
      "reward_std": 0.007682787254452705,
      "reward_std_all": 0.008946883957833052,
      "rewards-max/forensics_sinkhorn_reward": 0.9719325304031372,
      "rewards-max/forensics_strict_edge_reward": 0.44935600459575653,
      "rewards-max/forensics_tiou_reward": 0.9329127371311188,
      "rewards/forensics_sinkhorn_reward": 0.7654235363006592,
      "rewards/forensics_strict_edge_reward": 0.03986490610986948,
      "rewards/forensics_tiou_reward": 0.5776987671852112,
      "step": 665
    },
    {
      "completion_length": 22.734375,
      "epoch": 2.043010752688172,
      "grad_norm": 6.202325344085693,
      "kl": 0.330078125,
      "learning_rate": 4.884615384615384e-07,
      "loss": 0.0249,
      "reward": 0.7872697710990906,
      "reward_std": 0.03113527223467827,
      "reward_std_all": 0.02939486037939787,
      "rewards-max/forensics_sinkhorn_reward": 0.9526722729206085,
      "rewards-max/forensics_strict_edge_reward": 0.21724942326545715,
      "rewards-max/forensics_tiou_reward": 0.8754147887229919,
      "rewards/forensics_sinkhorn_reward": 0.7773796916007996,
      "rewards/forensics_strict_edge_reward": 0.02133611124008894,
      "rewards/forensics_tiou_reward": 0.5397410988807678,
      "step": 666
    },
    {
      "completion_length": 26.40625,
      "epoch": 2.046082949308756,
      "grad_norm": 6.604079246520996,
      "kl": 0.306640625,
      "learning_rate": 4.876923076923077e-07,
      "loss": 0.0456,
      "reward": 0.7537721395492554,
      "reward_std": 0.038995920680463314,
      "reward_std_all": 0.04696582816541195,
      "rewards-max/forensics_sinkhorn_reward": 0.9783952534198761,
      "rewards-max/forensics_strict_edge_reward": 0.4200657457113266,
      "rewards-max/forensics_tiou_reward": 0.9484127163887024,
      "rewards/forensics_sinkhorn_reward": 0.6954617500305176,
      "rewards/forensics_strict_edge_reward": 0.03511765971779823,
      "rewards/forensics_tiou_reward": 0.5476028621196747,
      "step": 667
    },
    {
      "completion_length": 23.109375,
      "epoch": 2.0491551459293396,
      "grad_norm": 7.167043685913086,
      "kl": 0.3427734375,
      "learning_rate": 4.869230769230769e-07,
      "loss": -0.1394,
      "reward": 0.7439512610435486,
      "reward_std": 0.02534118271432817,
      "reward_std_all": 0.04412594507448375,
      "rewards-max/forensics_sinkhorn_reward": 0.9935370087623596,
      "rewards-max/forensics_strict_edge_reward": 0.4042135924100876,
      "rewards-max/forensics_tiou_reward": 0.9848968088626862,
      "rewards/forensics_sinkhorn_reward": 0.7623320519924164,
      "rewards/forensics_strict_edge_reward": 0.05242989771068096,
      "rewards/forensics_tiou_reward": 0.6043020784854889,
      "step": 668
    },
    {
      "completion_length": 24.359375,
      "epoch": 2.0522273425499233,
      "grad_norm": 5.914920806884766,
      "kl": 0.3798828125,
      "learning_rate": 4.861538461538461e-07,
      "loss": 0.0324,
      "reward": 0.7571735978126526,
      "reward_std": 0.05927208624780178,
      "reward_std_all": 0.0569151695817709,
      "rewards-max/forensics_sinkhorn_reward": 0.9545551240444183,
      "rewards-max/forensics_strict_edge_reward": 0.14824033062905073,
      "rewards-max/forensics_tiou_reward": 0.8839415013790131,
      "rewards/forensics_sinkhorn_reward": 0.7033790946006775,
      "rewards/forensics_strict_edge_reward": 0.017333280527964234,
      "rewards/forensics_tiou_reward": 0.5042391270399094,
      "step": 669
    },
    {
      "completion_length": 24.75,
      "epoch": 2.055299539170507,
      "grad_norm": 4.316748142242432,
      "kl": 0.36328125,
      "learning_rate": 4.853846153846154e-07,
      "loss": -0.0291,
      "reward": 0.7489911317825317,
      "reward_std": 0.05475510563701391,
      "reward_std_all": 0.057192813605070114,
      "rewards-max/forensics_sinkhorn_reward": 0.9831563830375671,
      "rewards-max/forensics_strict_edge_reward": 0.36869725584983826,
      "rewards-max/forensics_tiou_reward": 0.9624574482440948,
      "rewards/forensics_sinkhorn_reward": 0.7503616213798523,
      "rewards/forensics_strict_edge_reward": 0.028787094168365,
      "rewards/forensics_tiou_reward": 0.5438220798969269,
      "step": 670
    },
    {
      "completion_length": 23.234375,
      "epoch": 2.0583717357910905,
      "grad_norm": 6.5695929527282715,
      "kl": 0.384765625,
      "learning_rate": 4.846153846153846e-07,
      "loss": 0.0709,
      "reward": 0.7334251999855042,
      "reward_std": 0.05028811073862016,
      "reward_std_all": 0.060278813354671,
      "rewards-max/forensics_sinkhorn_reward": 0.9845395088195801,
      "rewards-max/forensics_strict_edge_reward": 0.33328156918287277,
      "rewards-max/forensics_tiou_reward": 0.963844895362854,
      "rewards/forensics_sinkhorn_reward": 0.6723844110965729,
      "rewards/forensics_strict_edge_reward": 0.02645891346037388,
      "rewards/forensics_tiou_reward": 0.4946649968624115,
      "step": 671
    },
    {
      "completion_length": 31.296875,
      "epoch": 2.0614439324116742,
      "grad_norm": 5.627969741821289,
      "kl": 0.49609375,
      "learning_rate": 4.838461538461539e-07,
      "loss": 0.0653,
      "reward": 0.7686486542224884,
      "reward_std": 0.026003798469901085,
      "reward_std_all": 0.034920476377010345,
      "rewards-max/forensics_sinkhorn_reward": 0.9796696305274963,
      "rewards-max/forensics_strict_edge_reward": 0.41550540924072266,
      "rewards-max/forensics_tiou_reward": 0.9518709480762482,
      "rewards/forensics_sinkhorn_reward": 0.6692953705787659,
      "rewards/forensics_strict_edge_reward": 0.05592622235417366,
      "rewards/forensics_tiou_reward": 0.5955275595188141,
      "step": 672
    },
    {
      "completion_length": 28.6875,
      "epoch": 2.064516129032258,
      "grad_norm": 6.223988056182861,
      "kl": 0.62890625,
      "learning_rate": 4.83076923076923e-07,
      "loss": 0.0322,
      "reward": 0.7614077925682068,
      "reward_std": 0.04944496043026447,
      "reward_std_all": 0.0485920924693346,
      "rewards-max/forensics_sinkhorn_reward": 0.979823499917984,
      "rewards-max/forensics_strict_edge_reward": 0.4300691783428192,
      "rewards-max/forensics_tiou_reward": 0.9487760961055756,
      "rewards/forensics_sinkhorn_reward": 0.7190090119838715,
      "rewards/forensics_strict_edge_reward": 0.05871333088725805,
      "rewards/forensics_tiou_reward": 0.5551413595676422,
      "step": 673
    },
    {
      "completion_length": 23.421875,
      "epoch": 2.067588325652842,
      "grad_norm": 7.152769565582275,
      "kl": 0.345703125,
      "learning_rate": 4.823076923076923e-07,
      "loss": 0.075,
      "reward": 0.7568017840385437,
      "reward_std": 0.02355615422129631,
      "reward_std_all": 0.038362372666597366,
      "rewards-max/forensics_sinkhorn_reward": 0.9873636960983276,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9734640121459961,
      "rewards/forensics_sinkhorn_reward": 0.7405332922935486,
      "rewards/forensics_strict_edge_reward": 0.04962766170501709,
      "rewards/forensics_tiou_reward": 0.5704962909221649,
      "step": 674
    },
    {
      "completion_length": 26.203125,
      "epoch": 2.0706605222734256,
      "grad_norm": 5.986942291259766,
      "kl": 0.380859375,
      "learning_rate": 4.815384615384615e-07,
      "loss": -0.0012,
      "reward": 0.7612225413322449,
      "reward_std": 0.055759633891284466,
      "reward_std_all": 0.05338619090616703,
      "rewards-max/forensics_sinkhorn_reward": 0.9640110433101654,
      "rewards-max/forensics_strict_edge_reward": 0.37837012112140656,
      "rewards-max/forensics_tiou_reward": 0.9113307893276215,
      "rewards/forensics_sinkhorn_reward": 0.7253877818584442,
      "rewards/forensics_strict_edge_reward": 0.03277084790170193,
      "rewards/forensics_tiou_reward": 0.5369684100151062,
      "step": 675
    },
    {
      "completion_length": 24.5625,
      "epoch": 2.0737327188940093,
      "grad_norm": 5.077764987945557,
      "kl": 0.369140625,
      "learning_rate": 4.807692307692307e-07,
      "loss": 0.0637,
      "reward": 0.7863685488700867,
      "reward_std": 0.023161411634646356,
      "reward_std_all": 0.023440455086529255,
      "rewards-max/forensics_sinkhorn_reward": 0.9871971607208252,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.974840372800827,
      "rewards/forensics_sinkhorn_reward": 0.7588730454444885,
      "rewards/forensics_strict_edge_reward": 0.04803985543549061,
      "rewards/forensics_tiou_reward": 0.6066237688064575,
      "step": 676
    },
    {
      "completion_length": 26.078125,
      "epoch": 2.076804915514593,
      "grad_norm": 5.8844709396362305,
      "kl": 0.3701171875,
      "learning_rate": 4.8e-07,
      "loss": 0.0381,
      "reward": 0.7905628085136414,
      "reward_std": 0.021855640909052454,
      "reward_std_all": 0.021984432358294725,
      "rewards-max/forensics_sinkhorn_reward": 0.9688591957092285,
      "rewards-max/forensics_strict_edge_reward": 0.32489311695098877,
      "rewards-max/forensics_tiou_reward": 0.9269002377986908,
      "rewards/forensics_sinkhorn_reward": 0.7442829012870789,
      "rewards/forensics_strict_edge_reward": 0.033029692247509956,
      "rewards/forensics_tiou_reward": 0.5964508056640625,
      "step": 677
    },
    {
      "completion_length": 26.84375,
      "epoch": 2.0798771121351765,
      "grad_norm": 5.995879173278809,
      "kl": 0.4384765625,
      "learning_rate": 4.792307692307692e-07,
      "loss": 0.0299,
      "reward": 0.7808982729911804,
      "reward_std": 0.011900406796485186,
      "reward_std_all": 0.021050387993454933,
      "rewards-max/forensics_sinkhorn_reward": 0.9707761704921722,
      "rewards-max/forensics_strict_edge_reward": 0.5711847841739655,
      "rewards-max/forensics_tiou_reward": 0.9367630183696747,
      "rewards/forensics_sinkhorn_reward": 0.7832882702350616,
      "rewards/forensics_strict_edge_reward": 0.05126417987048626,
      "rewards/forensics_tiou_reward": 0.5947562754154205,
      "step": 678
    },
    {
      "completion_length": 24.203125,
      "epoch": 2.0829493087557602,
      "grad_norm": 7.375401020050049,
      "kl": 0.546875,
      "learning_rate": 4.784615384615384e-07,
      "loss": -0.0372,
      "reward": 0.7065413594245911,
      "reward_std": 0.024205224588513374,
      "reward_std_all": 0.06115472503006458,
      "rewards-max/forensics_sinkhorn_reward": 0.9839263260364532,
      "rewards-max/forensics_strict_edge_reward": 0.6664355397224426,
      "rewards-max/forensics_tiou_reward": 0.9676329791545868,
      "rewards/forensics_sinkhorn_reward": 0.6135035455226898,
      "rewards/forensics_strict_edge_reward": 0.0374621469527483,
      "rewards/forensics_tiou_reward": 0.3828900158405304,
      "step": 679
    },
    {
      "completion_length": 29.78125,
      "epoch": 2.086021505376344,
      "grad_norm": 5.398895263671875,
      "kl": 0.640625,
      "learning_rate": 4.776923076923077e-07,
      "loss": 0.0609,
      "reward": 0.7343849539756775,
      "reward_std": 0.042564040049910545,
      "reward_std_all": 0.05720517970621586,
      "rewards-max/forensics_sinkhorn_reward": 0.9719526469707489,
      "rewards-max/forensics_strict_edge_reward": 0.6343722939491272,
      "rewards-max/forensics_tiou_reward": 0.9374479651451111,
      "rewards/forensics_sinkhorn_reward": 0.7146551012992859,
      "rewards/forensics_strict_edge_reward": 0.06682638637721539,
      "rewards/forensics_tiou_reward": 0.5584060549736023,
      "step": 680
    },
    {
      "completion_length": 26.078125,
      "epoch": 2.089093701996928,
      "grad_norm": 5.911357402801514,
      "kl": 0.4404296875,
      "learning_rate": 4.769230769230769e-07,
      "loss": -0.0423,
      "reward": 0.7973918914794922,
      "reward_std": 0.015506343683227897,
      "reward_std_all": 0.01523403450846672,
      "rewards-max/forensics_sinkhorn_reward": 0.987350195646286,
      "rewards-max/forensics_strict_edge_reward": 0.7007113695144653,
      "rewards-max/forensics_tiou_reward": 0.9753707349300385,
      "rewards/forensics_sinkhorn_reward": 0.803606390953064,
      "rewards/forensics_strict_edge_reward": 0.07968040555715561,
      "rewards/forensics_tiou_reward": 0.6328320503234863,
      "step": 681
    },
    {
      "completion_length": 23.046875,
      "epoch": 2.0921658986175116,
      "grad_norm": 5.837103843688965,
      "kl": 0.3916015625,
      "learning_rate": 4.7615384615384613e-07,
      "loss": -0.0335,
      "reward": 0.7886475026607513,
      "reward_std": 0.01615692093037069,
      "reward_std_all": 0.01808269042521715,
      "rewards-max/forensics_sinkhorn_reward": 0.9931335151195526,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9842805564403534,
      "rewards/forensics_sinkhorn_reward": 0.7471268177032471,
      "rewards/forensics_strict_edge_reward": 0.04934132471680641,
      "rewards/forensics_tiou_reward": 0.5214062333106995,
      "step": 682
    },
    {
      "completion_length": 23.40625,
      "epoch": 2.0952380952380953,
      "grad_norm": 5.5141921043396,
      "kl": 0.4111328125,
      "learning_rate": 4.753846153846154e-07,
      "loss": -0.1104,
      "reward": 0.7912353873252869,
      "reward_std": 0.015006027184426785,
      "reward_std_all": 0.01632486446760595,
      "rewards-max/forensics_sinkhorn_reward": 0.9925924241542816,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.988190621137619,
      "rewards/forensics_sinkhorn_reward": 0.8034141361713409,
      "rewards/forensics_strict_edge_reward": 0.10058403015136719,
      "rewards/forensics_tiou_reward": 0.6026861071586609,
      "step": 683
    },
    {
      "completion_length": 23.1875,
      "epoch": 2.098310291858679,
      "grad_norm": 6.063951015472412,
      "kl": 0.3525390625,
      "learning_rate": 4.7461538461538456e-07,
      "loss": -0.0049,
      "reward": 0.7716509699821472,
      "reward_std": 0.04174876771867275,
      "reward_std_all": 0.04078436829149723,
      "rewards-max/forensics_sinkhorn_reward": 0.9788666665554047,
      "rewards-max/forensics_strict_edge_reward": 0.5335612744092941,
      "rewards-max/forensics_tiou_reward": 0.9504210650920868,
      "rewards/forensics_sinkhorn_reward": 0.7726814448833466,
      "rewards/forensics_strict_edge_reward": 0.048647690564394,
      "rewards/forensics_tiou_reward": 0.5561416745185852,
      "step": 684
    },
    {
      "completion_length": 22.53125,
      "epoch": 2.1013824884792625,
      "grad_norm": 6.290798664093018,
      "kl": 0.431640625,
      "learning_rate": 4.738461538461538e-07,
      "loss": -0.0062,
      "reward": 0.7587621808052063,
      "reward_std": 0.04121367819607258,
      "reward_std_all": 0.04393933527171612,
      "rewards-max/forensics_sinkhorn_reward": 0.9869960844516754,
      "rewards-max/forensics_strict_edge_reward": 0.6131598502397537,
      "rewards-max/forensics_tiou_reward": 0.971260279417038,
      "rewards/forensics_sinkhorn_reward": 0.7415936887264252,
      "rewards/forensics_strict_edge_reward": 0.04876251146197319,
      "rewards/forensics_tiou_reward": 0.5648490786552429,
      "step": 685
    },
    {
      "completion_length": 24.3125,
      "epoch": 2.1044546850998462,
      "grad_norm": 6.654916763305664,
      "kl": 0.3916015625,
      "learning_rate": 4.7307692307692304e-07,
      "loss": -0.0088,
      "reward": 0.7859704196453094,
      "reward_std": 0.027521838434040546,
      "reward_std_all": 0.02709764800965786,
      "rewards-max/forensics_sinkhorn_reward": 0.9589486122131348,
      "rewards-max/forensics_strict_edge_reward": 0.08249473571777344,
      "rewards-max/forensics_tiou_reward": 0.8961530923843384,
      "rewards/forensics_sinkhorn_reward": 0.7685360312461853,
      "rewards/forensics_strict_edge_reward": 0.0065298390109092,
      "rewards/forensics_tiou_reward": 0.5940371155738831,
      "step": 686
    },
    {
      "completion_length": 23.328125,
      "epoch": 2.10752688172043,
      "grad_norm": 5.318211555480957,
      "kl": 0.482421875,
      "learning_rate": 4.723076923076923e-07,
      "loss": -0.0142,
      "reward": 0.7730036675930023,
      "reward_std": 0.03134383913129568,
      "reward_std_all": 0.03269049618393183,
      "rewards-max/forensics_sinkhorn_reward": 0.9659673273563385,
      "rewards-max/forensics_strict_edge_reward": 0.3480503261089325,
      "rewards-max/forensics_tiou_reward": 0.9220889210700989,
      "rewards/forensics_sinkhorn_reward": 0.7385827302932739,
      "rewards/forensics_strict_edge_reward": 0.0336031150072813,
      "rewards/forensics_tiou_reward": 0.5596320331096649,
      "step": 687
    },
    {
      "completion_length": 27.453125,
      "epoch": 2.110599078341014,
      "grad_norm": 6.495015621185303,
      "kl": 0.3525390625,
      "learning_rate": 4.715384615384615e-07,
      "loss": 0.1233,
      "reward": 0.7558504939079285,
      "reward_std": 0.05524751543998718,
      "reward_std_all": 0.05396660789847374,
      "rewards-max/forensics_sinkhorn_reward": 0.9837889969348907,
      "rewards-max/forensics_strict_edge_reward": 0.41391946375370026,
      "rewards-max/forensics_tiou_reward": 0.9599722921848297,
      "rewards/forensics_sinkhorn_reward": 0.7261167764663696,
      "rewards/forensics_strict_edge_reward": 0.045673564076423645,
      "rewards/forensics_tiou_reward": 0.5800547301769257,
      "step": 688
    },
    {
      "completion_length": 25.21875,
      "epoch": 2.1136712749615976,
      "grad_norm": 6.241086483001709,
      "kl": 0.3154296875,
      "learning_rate": 4.7076923076923073e-07,
      "loss": -0.0089,
      "reward": 0.6939437985420227,
      "reward_std": 0.03725033067166805,
      "reward_std_all": 0.07487799786031246,
      "rewards-max/forensics_sinkhorn_reward": 0.996629923582077,
      "rewards-max/forensics_strict_edge_reward": 0.7797744870185852,
      "rewards-max/forensics_tiou_reward": 0.9975532591342926,
      "rewards/forensics_sinkhorn_reward": 0.7045718133449554,
      "rewards/forensics_strict_edge_reward": 0.07665173709392548,
      "rewards/forensics_tiou_reward": 0.5468261241912842,
      "step": 689
    },
    {
      "completion_length": 26.734375,
      "epoch": 2.1167434715821813,
      "grad_norm": 6.4001336097717285,
      "kl": 0.3740234375,
      "learning_rate": 4.6999999999999995e-07,
      "loss": -0.0627,
      "reward": 0.7793076038360596,
      "reward_std": 0.012029777397401631,
      "reward_std_all": 0.02055376721546054,
      "rewards-max/forensics_sinkhorn_reward": 0.9923577308654785,
      "rewards-max/forensics_strict_edge_reward": 0.6302271485328674,
      "rewards-max/forensics_tiou_reward": 0.9849896728992462,
      "rewards/forensics_sinkhorn_reward": 0.7502721846103668,
      "rewards/forensics_strict_edge_reward": 0.032294053584337234,
      "rewards/forensics_tiou_reward": 0.5420851409435272,
      "step": 690
    },
    {
      "completion_length": 22.46875,
      "epoch": 2.119815668202765,
      "grad_norm": 8.92097282409668,
      "kl": 0.5009765625,
      "learning_rate": 4.692307692307692e-07,
      "loss": 0.0138,
      "reward": 0.801032304763794,
      "reward_std": 0.00789451354648918,
      "reward_std_all": 0.009161403286270797,
      "rewards-max/forensics_sinkhorn_reward": 0.9792738258838654,
      "rewards-max/forensics_strict_edge_reward": 0.296694740653038,
      "rewards-max/forensics_tiou_reward": 0.9463472962379456,
      "rewards/forensics_sinkhorn_reward": 0.8096863627433777,
      "rewards/forensics_strict_edge_reward": 0.04945967858657241,
      "rewards/forensics_tiou_reward": 0.5739782750606537,
      "step": 691
    },
    {
      "completion_length": 24.703125,
      "epoch": 2.122887864823349,
      "grad_norm": 6.4421586990356445,
      "kl": 0.3916015625,
      "learning_rate": 4.684615384615385e-07,
      "loss": -0.0512,
      "reward": 0.757574737071991,
      "reward_std": 0.03999593015760183,
      "reward_std_all": 0.04427051357924938,
      "rewards-max/forensics_sinkhorn_reward": 0.9901678264141083,
      "rewards-max/forensics_strict_edge_reward": 0.5385358333587646,
      "rewards-max/forensics_tiou_reward": 0.9776418507099152,
      "rewards/forensics_sinkhorn_reward": 0.7160784602165222,
      "rewards/forensics_strict_edge_reward": 0.038715142756700516,
      "rewards/forensics_tiou_reward": 0.5187192857265472,
      "step": 692
    },
    {
      "completion_length": 23.625,
      "epoch": 2.1259600614439322,
      "grad_norm": 6.772078514099121,
      "kl": 0.3583984375,
      "learning_rate": 4.6769230769230764e-07,
      "loss": 0.0482,
      "reward": 0.7728385031223297,
      "reward_std": 0.036900844424963,
      "reward_std_all": 0.03662450239062309,
      "rewards-max/forensics_sinkhorn_reward": 0.9520013630390167,
      "rewards-max/forensics_strict_edge_reward": 0.45670856535434723,
      "rewards-max/forensics_tiou_reward": 0.8807641267776489,
      "rewards/forensics_sinkhorn_reward": 0.7415980100631714,
      "rewards/forensics_strict_edge_reward": 0.040810076519846916,
      "rewards/forensics_tiou_reward": 0.5278016924858093,
      "step": 693
    },
    {
      "completion_length": 26.6875,
      "epoch": 2.129032258064516,
      "grad_norm": 6.153506755828857,
      "kl": 0.43359375,
      "learning_rate": 4.669230769230769e-07,
      "loss": 0.1028,
      "reward": 0.7291146814823151,
      "reward_std": 0.08701204881072044,
      "reward_std_all": 0.08609882742166519,
      "rewards-max/forensics_sinkhorn_reward": 0.9690062999725342,
      "rewards-max/forensics_strict_edge_reward": 0.402368500828743,
      "rewards-max/forensics_tiou_reward": 0.9319826066493988,
      "rewards/forensics_sinkhorn_reward": 0.666508138179779,
      "rewards/forensics_strict_edge_reward": 0.026730588637292385,
      "rewards/forensics_tiou_reward": 0.4603613466024399,
      "step": 694
    },
    {
      "completion_length": 24.71875,
      "epoch": 2.1321044546851,
      "grad_norm": 5.74907922744751,
      "kl": 0.3623046875,
      "learning_rate": 4.661538461538461e-07,
      "loss": 0.1527,
      "reward": 0.7469924688339233,
      "reward_std": 0.07281244732439518,
      "reward_std_all": 0.06907154247164726,
      "rewards-max/forensics_sinkhorn_reward": 0.9802219271659851,
      "rewards-max/forensics_strict_edge_reward": 0.6868928670883179,
      "rewards-max/forensics_tiou_reward": 0.9582491219043732,
      "rewards/forensics_sinkhorn_reward": 0.686160683631897,
      "rewards/forensics_strict_edge_reward": 0.039270924404263496,
      "rewards/forensics_tiou_reward": 0.5066381543874741,
      "step": 695
    },
    {
      "completion_length": 26.125,
      "epoch": 2.1351766513056836,
      "grad_norm": 8.038463592529297,
      "kl": 0.4384765625,
      "learning_rate": 4.653846153846154e-07,
      "loss": 0.1018,
      "reward": 0.7655731737613678,
      "reward_std": 0.03228858020156622,
      "reward_std_all": 0.036687953397631645,
      "rewards-max/forensics_sinkhorn_reward": 0.9895890057086945,
      "rewards-max/forensics_strict_edge_reward": 0.6483739018440247,
      "rewards-max/forensics_tiou_reward": 0.9803153872489929,
      "rewards/forensics_sinkhorn_reward": 0.7264804542064667,
      "rewards/forensics_strict_edge_reward": 0.05399483349174261,
      "rewards/forensics_tiou_reward": 0.5051645189523697,
      "step": 696
    },
    {
      "completion_length": 27.171875,
      "epoch": 2.1382488479262673,
      "grad_norm": 5.8183207511901855,
      "kl": 0.3857421875,
      "learning_rate": 4.646153846153846e-07,
      "loss": -0.0222,
      "reward": 0.7840546667575836,
      "reward_std": 0.03664377611130476,
      "reward_std_all": 0.03374896198511124,
      "rewards-max/forensics_sinkhorn_reward": 0.9838432669639587,
      "rewards-max/forensics_strict_edge_reward": 0.48896080255508423,
      "rewards-max/forensics_tiou_reward": 0.9617001116275787,
      "rewards/forensics_sinkhorn_reward": 0.7803496718406677,
      "rewards/forensics_strict_edge_reward": 0.03865283913910389,
      "rewards/forensics_tiou_reward": 0.5665394067764282,
      "step": 697
    },
    {
      "completion_length": 21.984375,
      "epoch": 2.141321044546851,
      "grad_norm": 5.606300354003906,
      "kl": 0.408203125,
      "learning_rate": 4.638461538461538e-07,
      "loss": 0.0639,
      "reward": 0.7629055082798004,
      "reward_std": 0.05546704865992069,
      "reward_std_all": 0.05240819975733757,
      "rewards-max/forensics_sinkhorn_reward": 0.9786538481712341,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.952557384967804,
      "rewards/forensics_sinkhorn_reward": 0.7558738887310028,
      "rewards/forensics_strict_edge_reward": 0.04491519369184971,
      "rewards/forensics_tiou_reward": 0.5644398182630539,
      "step": 698
    },
    {
      "completion_length": 26.796875,
      "epoch": 2.144393241167435,
      "grad_norm": 8.184003829956055,
      "kl": 0.3330078125,
      "learning_rate": 4.6307692307692303e-07,
      "loss": -0.0141,
      "reward": 0.8031177520751953,
      "reward_std": 0.00505822931881994,
      "reward_std_all": 0.006630523130297661,
      "rewards-max/forensics_sinkhorn_reward": 0.986700028181076,
      "rewards-max/forensics_strict_edge_reward": 0.42959781736135483,
      "rewards-max/forensics_tiou_reward": 0.968488335609436,
      "rewards/forensics_sinkhorn_reward": 0.8349284529685974,
      "rewards/forensics_strict_edge_reward": 0.036742741242051125,
      "rewards/forensics_tiou_reward": 0.6668771207332611,
      "step": 699
    },
    {
      "completion_length": 22.5625,
      "epoch": 2.1474654377880182,
      "grad_norm": 5.741703033447266,
      "kl": 0.3779296875,
      "learning_rate": 4.623076923076923e-07,
      "loss": 0.0897,
      "reward": 0.77922323346138,
      "reward_std": 0.03731134161353111,
      "reward_std_all": 0.03522837907075882,
      "rewards-max/forensics_sinkhorn_reward": 0.9952621459960938,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9935224056243896,
      "rewards/forensics_sinkhorn_reward": 0.7869980931282043,
      "rewards/forensics_strict_edge_reward": 0.08658739551901817,
      "rewards/forensics_tiou_reward": 0.6007204651832581,
      "step": 700
    },
    {
      "completion_length": 22.796875,
      "epoch": 2.150537634408602,
      "grad_norm": 5.828261375427246,
      "kl": 0.359375,
      "learning_rate": 4.6153846153846156e-07,
      "loss": -0.0161,
      "reward": 0.7858967781066895,
      "reward_std": 0.031106517650187016,
      "reward_std_all": 0.029378613457083702,
      "rewards-max/forensics_sinkhorn_reward": 0.9688081443309784,
      "rewards-max/forensics_strict_edge_reward": 0.48898807168006897,
      "rewards-max/forensics_tiou_reward": 0.933437705039978,
      "rewards/forensics_sinkhorn_reward": 0.7668642997741699,
      "rewards/forensics_strict_edge_reward": 0.05562106519937515,
      "rewards/forensics_tiou_reward": 0.5864724516868591,
      "step": 701
    },
    {
      "completion_length": 28.328125,
      "epoch": 2.153609831029186,
      "grad_norm": 6.110796928405762,
      "kl": 0.4052734375,
      "learning_rate": 4.607692307692307e-07,
      "loss": 0.0903,
      "reward": 0.7452661097049713,
      "reward_std": 0.06533968634903431,
      "reward_std_all": 0.06357379630208015,
      "rewards-max/forensics_sinkhorn_reward": 0.9823749661445618,
      "rewards-max/forensics_strict_edge_reward": 0.42210179567337036,
      "rewards-max/forensics_tiou_reward": 0.9620229601860046,
      "rewards/forensics_sinkhorn_reward": 0.6940712630748749,
      "rewards/forensics_strict_edge_reward": 0.046070738695561886,
      "rewards/forensics_tiou_reward": 0.5593440234661102,
      "step": 702
    },
    {
      "completion_length": 29.625,
      "epoch": 2.1566820276497696,
      "grad_norm": 7.335577964782715,
      "kl": 0.3642578125,
      "learning_rate": 4.6e-07,
      "loss": -0.0072,
      "reward": 0.737880527973175,
      "reward_std": 0.035495450370945036,
      "reward_std_all": 0.0529831238090992,
      "rewards-max/forensics_sinkhorn_reward": 0.9827663600444794,
      "rewards-max/forensics_strict_edge_reward": 0.3002149760723114,
      "rewards-max/forensics_tiou_reward": 0.960147351026535,
      "rewards/forensics_sinkhorn_reward": 0.6812287867069244,
      "rewards/forensics_strict_edge_reward": 0.018693854101002216,
      "rewards/forensics_tiou_reward": 0.4677594006061554,
      "step": 703
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.1597542242703534,
      "grad_norm": 7.170388221740723,
      "kl": 0.3515625,
      "learning_rate": 4.592307692307692e-07,
      "loss": 0.0217,
      "reward": 0.7677284181118011,
      "reward_std": 0.03689729794859886,
      "reward_std_all": 0.03814258798956871,
      "rewards-max/forensics_sinkhorn_reward": 0.9693499207496643,
      "rewards-max/forensics_strict_edge_reward": 0.34965482563711703,
      "rewards-max/forensics_tiou_reward": 0.9094963073730469,
      "rewards/forensics_sinkhorn_reward": 0.7433627247810364,
      "rewards/forensics_strict_edge_reward": 0.025164484919514507,
      "rewards/forensics_tiou_reward": 0.5560831129550934,
      "step": 704
    },
    {
      "completion_length": 24.78125,
      "epoch": 2.162826420890937,
      "grad_norm": 8.499968528747559,
      "kl": 0.5458984375,
      "learning_rate": 4.5846153846153847e-07,
      "loss": 0.0995,
      "reward": 0.7401717603206635,
      "reward_std": 0.0867893174290657,
      "reward_std_all": 0.08048039302229881,
      "rewards-max/forensics_sinkhorn_reward": 0.9836197197437286,
      "rewards-max/forensics_strict_edge_reward": 0.5207925289869308,
      "rewards-max/forensics_tiou_reward": 0.9622168242931366,
      "rewards/forensics_sinkhorn_reward": 0.706816166639328,
      "rewards/forensics_strict_edge_reward": 0.05700070969760418,
      "rewards/forensics_tiou_reward": 0.5292050242424011,
      "step": 705
    },
    {
      "completion_length": 24.140625,
      "epoch": 2.165898617511521,
      "grad_norm": 7.93346643447876,
      "kl": 0.3662109375,
      "learning_rate": 4.5769230769230763e-07,
      "loss": 0.1024,
      "reward": 0.7475211322307587,
      "reward_std": 0.066537294536829,
      "reward_std_all": 0.06391855701804161,
      "rewards-max/forensics_sinkhorn_reward": 0.9812047183513641,
      "rewards-max/forensics_strict_edge_reward": 0.5275479853153229,
      "rewards-max/forensics_tiou_reward": 0.9515701234340668,
      "rewards/forensics_sinkhorn_reward": 0.7269896864891052,
      "rewards/forensics_strict_edge_reward": 0.04596060328185558,
      "rewards/forensics_tiou_reward": 0.5291705429553986,
      "step": 706
    },
    {
      "completion_length": 22.671875,
      "epoch": 2.1689708141321047,
      "grad_norm": 6.899468898773193,
      "kl": 0.5068359375,
      "learning_rate": 4.569230769230769e-07,
      "loss": -0.0056,
      "reward": 0.7821134924888611,
      "reward_std": 0.04076249338686466,
      "reward_std_all": 0.03725290857255459,
      "rewards-max/forensics_sinkhorn_reward": 0.9875491857528687,
      "rewards-max/forensics_strict_edge_reward": 0.48716044053435326,
      "rewards-max/forensics_tiou_reward": 0.9640858471393585,
      "rewards/forensics_sinkhorn_reward": 0.7649939358234406,
      "rewards/forensics_strict_edge_reward": 0.0542127825319767,
      "rewards/forensics_tiou_reward": 0.5513517558574677,
      "step": 707
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.172043010752688,
      "grad_norm": 9.823409080505371,
      "kl": 0.3994140625,
      "learning_rate": 4.561538461538461e-07,
      "loss": 0.048,
      "reward": 0.7613030672073364,
      "reward_std": 0.039287144318223,
      "reward_std_all": 0.0447088498622179,
      "rewards-max/forensics_sinkhorn_reward": 0.9722990393638611,
      "rewards-max/forensics_strict_edge_reward": 0.35242731869220734,
      "rewards-max/forensics_tiou_reward": 0.9320884048938751,
      "rewards/forensics_sinkhorn_reward": 0.7674633860588074,
      "rewards/forensics_strict_edge_reward": 0.035683984868228436,
      "rewards/forensics_tiou_reward": 0.5599998235702515,
      "step": 708
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.175115207373272,
      "grad_norm": 7.54771614074707,
      "kl": 0.447265625,
      "learning_rate": 4.553846153846154e-07,
      "loss": 0.0539,
      "reward": 0.782475620508194,
      "reward_std": 0.025356355123221874,
      "reward_std_all": 0.02605759259313345,
      "rewards-max/forensics_sinkhorn_reward": 0.9476788341999054,
      "rewards-max/forensics_strict_edge_reward": 0.458488866686821,
      "rewards-max/forensics_tiou_reward": 0.8737247288227081,
      "rewards/forensics_sinkhorn_reward": 0.7461371719837189,
      "rewards/forensics_strict_edge_reward": 0.031480466946959496,
      "rewards/forensics_tiou_reward": 0.5091255903244019,
      "step": 709
    },
    {
      "completion_length": 24.015625,
      "epoch": 2.1781874039938556,
      "grad_norm": 5.918900012969971,
      "kl": 0.3369140625,
      "learning_rate": 4.546153846153846e-07,
      "loss": -0.0243,
      "reward": 0.7827095985412598,
      "reward_std": 0.023714527487754822,
      "reward_std_all": 0.025416390039026737,
      "rewards-max/forensics_sinkhorn_reward": 0.9965268969535828,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.99709752202034,
      "rewards/forensics_sinkhorn_reward": 0.7383596003055573,
      "rewards/forensics_strict_edge_reward": 0.06862058117985725,
      "rewards/forensics_tiou_reward": 0.5577957034111023,
      "step": 710
    },
    {
      "completion_length": 23.859375,
      "epoch": 2.1812596006144394,
      "grad_norm": 6.157593250274658,
      "kl": 0.38671875,
      "learning_rate": 4.538461538461538e-07,
      "loss": 0.0011,
      "reward": 0.769619345664978,
      "reward_std": 0.052618082612752914,
      "reward_std_all": 0.048868170008063316,
      "rewards-max/forensics_sinkhorn_reward": 0.987546443939209,
      "rewards-max/forensics_strict_edge_reward": 0.3942641466856003,
      "rewards-max/forensics_tiou_reward": 0.9688098728656769,
      "rewards/forensics_sinkhorn_reward": 0.7767499387264252,
      "rewards/forensics_strict_edge_reward": 0.04251478239893913,
      "rewards/forensics_tiou_reward": 0.6201600432395935,
      "step": 711
    },
    {
      "completion_length": 26.203125,
      "epoch": 2.184331797235023,
      "grad_norm": 5.890629768371582,
      "kl": 0.341796875,
      "learning_rate": 4.5307692307692307e-07,
      "loss": 0.0221,
      "reward": 0.7192094326019287,
      "reward_std": 0.029854143038392067,
      "reward_std_all": 0.05793636292219162,
      "rewards-max/forensics_sinkhorn_reward": 0.9841163158416748,
      "rewards-max/forensics_strict_edge_reward": 0.45237329602241516,
      "rewards-max/forensics_tiou_reward": 0.9641928970813751,
      "rewards/forensics_sinkhorn_reward": 0.7181654572486877,
      "rewards/forensics_strict_edge_reward": 0.07401661528274417,
      "rewards/forensics_tiou_reward": 0.5176456272602081,
      "step": 712
    },
    {
      "completion_length": 21.390625,
      "epoch": 2.187403993855607,
      "grad_norm": 6.132652282714844,
      "kl": 0.4130859375,
      "learning_rate": 4.523076923076923e-07,
      "loss": 0.063,
      "reward": 0.7804799973964691,
      "reward_std": 0.0359455868601799,
      "reward_std_all": 0.03415287099778652,
      "rewards-max/forensics_sinkhorn_reward": 0.9844805002212524,
      "rewards-max/forensics_strict_edge_reward": 0.3107226490974426,
      "rewards-max/forensics_tiou_reward": 0.9656020700931549,
      "rewards/forensics_sinkhorn_reward": 0.7692617774009705,
      "rewards/forensics_strict_edge_reward": 0.053285226225852966,
      "rewards/forensics_tiou_reward": 0.5587331056594849,
      "step": 713
    },
    {
      "completion_length": 24.171875,
      "epoch": 2.1904761904761907,
      "grad_norm": 15.624581336975098,
      "kl": 1.2041015625,
      "learning_rate": 4.5153846153846155e-07,
      "loss": 0.0211,
      "reward": 0.7943039834499359,
      "reward_std": 0.020439331652596593,
      "reward_std_all": 0.01977723161689937,
      "rewards-max/forensics_sinkhorn_reward": 0.9832119047641754,
      "rewards-max/forensics_strict_edge_reward": 0.7514603435993195,
      "rewards-max/forensics_tiou_reward": 0.9668079018592834,
      "rewards/forensics_sinkhorn_reward": 0.7875688672065735,
      "rewards/forensics_strict_edge_reward": 0.07807378843426704,
      "rewards/forensics_tiou_reward": 0.5967632234096527,
      "step": 714
    },
    {
      "completion_length": 23.59375,
      "epoch": 2.193548387096774,
      "grad_norm": 7.1657538414001465,
      "kl": 0.337890625,
      "learning_rate": 4.507692307692307e-07,
      "loss": 0.1064,
      "reward": 0.7682755291461945,
      "reward_std": 0.030368590727448463,
      "reward_std_all": 0.03492813091725111,
      "rewards-max/forensics_sinkhorn_reward": 0.9401063323020935,
      "rewards-max/forensics_strict_edge_reward": 0.18376103788614273,
      "rewards-max/forensics_tiou_reward": 0.8597049117088318,
      "rewards/forensics_sinkhorn_reward": 0.7063979804515839,
      "rewards/forensics_strict_edge_reward": 0.02098175510764122,
      "rewards/forensics_tiou_reward": 0.4945852756500244,
      "step": 715
    },
    {
      "completion_length": 26.59375,
      "epoch": 2.196620583717358,
      "grad_norm": 6.5518670082092285,
      "kl": 0.2958984375,
      "learning_rate": 4.5e-07,
      "loss": -0.0345,
      "reward": 0.7780177891254425,
      "reward_std": 0.04076702520251274,
      "reward_std_all": 0.03864186257123947,
      "rewards-max/forensics_sinkhorn_reward": 0.9626322388648987,
      "rewards-max/forensics_strict_edge_reward": 0.245257206261158,
      "rewards-max/forensics_tiou_reward": 0.9042361974716187,
      "rewards/forensics_sinkhorn_reward": 0.6821832060813904,
      "rewards/forensics_strict_edge_reward": 0.028634427580982447,
      "rewards/forensics_tiou_reward": 0.5216324478387833,
      "step": 716
    },
    {
      "completion_length": 27.65625,
      "epoch": 2.1996927803379416,
      "grad_norm": 6.295001029968262,
      "kl": 0.59765625,
      "learning_rate": 4.492307692307692e-07,
      "loss": 0.072,
      "reward": 0.7695481479167938,
      "reward_std": 0.03877845220267773,
      "reward_std_all": 0.03868933208286762,
      "rewards-max/forensics_sinkhorn_reward": 0.9920960366725922,
      "rewards-max/forensics_strict_edge_reward": 0.5179188400506973,
      "rewards-max/forensics_tiou_reward": 0.9863370358943939,
      "rewards/forensics_sinkhorn_reward": 0.7495100498199463,
      "rewards/forensics_strict_edge_reward": 0.05397367663681507,
      "rewards/forensics_tiou_reward": 0.6419136822223663,
      "step": 717
    },
    {
      "completion_length": 23.625,
      "epoch": 2.2027649769585254,
      "grad_norm": 8.357335090637207,
      "kl": 0.337890625,
      "learning_rate": 4.4846153846153846e-07,
      "loss": -0.0443,
      "reward": 0.7520761489868164,
      "reward_std": 0.025106745772063732,
      "reward_std_all": 0.04276055097579956,
      "rewards-max/forensics_sinkhorn_reward": 0.9863522052764893,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.971580982208252,
      "rewards/forensics_sinkhorn_reward": 0.7402702271938324,
      "rewards/forensics_strict_edge_reward": 0.054453086107969284,
      "rewards/forensics_tiou_reward": 0.4958798587322235,
      "step": 718
    },
    {
      "completion_length": 22.609375,
      "epoch": 2.205837173579109,
      "grad_norm": 7.165841579437256,
      "kl": 0.4375,
      "learning_rate": 4.476923076923076e-07,
      "loss": -0.0615,
      "reward": 0.8032088875770569,
      "reward_std": 0.004739056807011366,
      "reward_std_all": 0.006310093216598034,
      "rewards-max/forensics_sinkhorn_reward": 0.9844098091125488,
      "rewards-max/forensics_strict_edge_reward": 0.5598244965076447,
      "rewards-max/forensics_tiou_reward": 0.9649733901023865,
      "rewards/forensics_sinkhorn_reward": 0.816137969493866,
      "rewards/forensics_strict_edge_reward": 0.042515311390161514,
      "rewards/forensics_tiou_reward": 0.6282524168491364,
      "step": 719
    },
    {
      "completion_length": 25.953125,
      "epoch": 2.208909370199693,
      "grad_norm": 6.795928478240967,
      "kl": 0.5283203125,
      "learning_rate": 4.469230769230769e-07,
      "loss": 0.01,
      "reward": 0.7931855320930481,
      "reward_std": 0.022823249048087746,
      "reward_std_all": 0.021773146698251367,
      "rewards-max/forensics_sinkhorn_reward": 0.9868137836456299,
      "rewards-max/forensics_strict_edge_reward": 0.4668678343296051,
      "rewards-max/forensics_tiou_reward": 0.9690662324428558,
      "rewards/forensics_sinkhorn_reward": 0.8068016171455383,
      "rewards/forensics_strict_edge_reward": 0.03281950764358044,
      "rewards/forensics_tiou_reward": 0.6889516711235046,
      "step": 720
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.2119815668202767,
      "grad_norm": 6.402031898498535,
      "kl": 0.521484375,
      "learning_rate": 4.4615384615384615e-07,
      "loss": 0.0122,
      "reward": 0.7941211760044098,
      "reward_std": 0.02044475544244051,
      "reward_std_all": 0.020034671761095524,
      "rewards-max/forensics_sinkhorn_reward": 0.992080569267273,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9841895699501038,
      "rewards/forensics_sinkhorn_reward": 0.7955025434494019,
      "rewards/forensics_strict_edge_reward": 0.07670468650758266,
      "rewards/forensics_tiou_reward": 0.6354508399963379,
      "step": 721
    },
    {
      "completion_length": 23.921875,
      "epoch": 2.21505376344086,
      "grad_norm": 8.309524536132812,
      "kl": 0.3447265625,
      "learning_rate": 4.4538461538461537e-07,
      "loss": 0.0684,
      "reward": 0.7656365036964417,
      "reward_std": 0.04201675020158291,
      "reward_std_all": 0.04565505124628544,
      "rewards-max/forensics_sinkhorn_reward": 0.9841937124729156,
      "rewards-max/forensics_strict_edge_reward": 0.6398303955793381,
      "rewards-max/forensics_tiou_reward": 0.9657877087593079,
      "rewards/forensics_sinkhorn_reward": 0.7864605188369751,
      "rewards/forensics_strict_edge_reward": 0.08565283939242363,
      "rewards/forensics_tiou_reward": 0.5694224238395691,
      "step": 722
    },
    {
      "completion_length": 24.3125,
      "epoch": 2.218125960061444,
      "grad_norm": 7.938176155090332,
      "kl": 0.3896484375,
      "learning_rate": 4.4461538461538463e-07,
      "loss": 0.0377,
      "reward": 0.7695313096046448,
      "reward_std": 0.03548311721533537,
      "reward_std_all": 0.036242087138816714,
      "rewards-max/forensics_sinkhorn_reward": 0.9892707169055939,
      "rewards-max/forensics_strict_edge_reward": 0.3440793603658676,
      "rewards-max/forensics_tiou_reward": 0.977667510509491,
      "rewards/forensics_sinkhorn_reward": 0.7641326487064362,
      "rewards/forensics_strict_edge_reward": 0.04287166707217693,
      "rewards/forensics_tiou_reward": 0.532218724489212,
      "step": 723
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.2211981566820276,
      "grad_norm": 7.2920823097229,
      "kl": 0.4072265625,
      "learning_rate": 4.438461538461538e-07,
      "loss": 0.0192,
      "reward": 0.7960057854652405,
      "reward_std": 0.01674529304727912,
      "reward_std_all": 0.01690697856247425,
      "rewards-max/forensics_sinkhorn_reward": 0.9919210970401764,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9867921471595764,
      "rewards/forensics_sinkhorn_reward": 0.7847400307655334,
      "rewards/forensics_strict_edge_reward": 0.07160268723964691,
      "rewards/forensics_tiou_reward": 0.5781625509262085,
      "step": 724
    },
    {
      "completion_length": 26.421875,
      "epoch": 2.2242703533026114,
      "grad_norm": 6.544732570648193,
      "kl": 0.4150390625,
      "learning_rate": 4.4307692307692306e-07,
      "loss": 0.0055,
      "reward": 0.7484894096851349,
      "reward_std": 0.02522933017462492,
      "reward_std_all": 0.04100805032067001,
      "rewards-max/forensics_sinkhorn_reward": 0.990036278963089,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9787323474884033,
      "rewards/forensics_sinkhorn_reward": 0.7502906918525696,
      "rewards/forensics_strict_edge_reward": 0.048797509633004665,
      "rewards/forensics_tiou_reward": 0.5718702375888824,
      "step": 725
    },
    {
      "completion_length": 23.65625,
      "epoch": 2.227342549923195,
      "grad_norm": 5.856253147125244,
      "kl": 0.3681640625,
      "learning_rate": 4.423076923076923e-07,
      "loss": -0.0119,
      "reward": 0.8027283549308777,
      "reward_std": 0.007201057160273194,
      "reward_std_all": 0.008017808198928833,
      "rewards-max/forensics_sinkhorn_reward": 0.9932795166969299,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9887500107288361,
      "rewards/forensics_sinkhorn_reward": 0.8512634634971619,
      "rewards/forensics_strict_edge_reward": 0.0970831923186779,
      "rewards/forensics_tiou_reward": 0.6715079545974731,
      "step": 726
    },
    {
      "completion_length": 26.125,
      "epoch": 2.230414746543779,
      "grad_norm": 5.5827555656433105,
      "kl": 0.375,
      "learning_rate": 4.4153846153846154e-07,
      "loss": 0.0192,
      "reward": 0.7630105912685394,
      "reward_std": 0.046416355296969414,
      "reward_std_all": 0.04582674987614155,
      "rewards-max/forensics_sinkhorn_reward": 0.9817119538784027,
      "rewards-max/forensics_strict_edge_reward": 0.5550410598516464,
      "rewards-max/forensics_tiou_reward": 0.9435957074165344,
      "rewards/forensics_sinkhorn_reward": 0.7264200150966644,
      "rewards/forensics_strict_edge_reward": 0.05046136491000652,
      "rewards/forensics_tiou_reward": 0.57496577501297,
      "step": 727
    },
    {
      "completion_length": 24.484375,
      "epoch": 2.2334869431643627,
      "grad_norm": 7.0445709228515625,
      "kl": 0.4453125,
      "learning_rate": 4.407692307692307e-07,
      "loss": 0.0484,
      "reward": 0.731064110994339,
      "reward_std": 0.07046490162611008,
      "reward_std_all": 0.07167273014783859,
      "rewards-max/forensics_sinkhorn_reward": 0.9915422201156616,
      "rewards-max/forensics_strict_edge_reward": 0.31991884112358093,
      "rewards-max/forensics_tiou_reward": 0.9792219698429108,
      "rewards/forensics_sinkhorn_reward": 0.6363194286823273,
      "rewards/forensics_strict_edge_reward": 0.02110006846487522,
      "rewards/forensics_tiou_reward": 0.42551545798778534,
      "step": 728
    },
    {
      "completion_length": 26.546875,
      "epoch": 2.236559139784946,
      "grad_norm": 6.402442932128906,
      "kl": 0.404296875,
      "learning_rate": 4.3999999999999997e-07,
      "loss": 0.1298,
      "reward": 0.7437491714954376,
      "reward_std": 0.06739111244678497,
      "reward_std_all": 0.06755444407463074,
      "rewards-max/forensics_sinkhorn_reward": 0.9853622019290924,
      "rewards-max/forensics_strict_edge_reward": 0.4364918917417526,
      "rewards-max/forensics_tiou_reward": 0.9654324948787689,
      "rewards/forensics_sinkhorn_reward": 0.6599788963794708,
      "rewards/forensics_strict_edge_reward": 0.06224972289055586,
      "rewards/forensics_tiou_reward": 0.4966612905263901,
      "step": 729
    },
    {
      "completion_length": 27.59375,
      "epoch": 2.23963133640553,
      "grad_norm": 5.4537787437438965,
      "kl": 0.3720703125,
      "learning_rate": 4.3923076923076924e-07,
      "loss": -0.0436,
      "reward": 0.8041732013225555,
      "reward_std": 0.0013626486179418862,
      "reward_std_all": 0.003961386624723673,
      "rewards-max/forensics_sinkhorn_reward": 0.9688514173030853,
      "rewards-max/forensics_strict_edge_reward": 0.3489128574728966,
      "rewards-max/forensics_tiou_reward": 0.9247574508190155,
      "rewards/forensics_sinkhorn_reward": 0.771597683429718,
      "rewards/forensics_strict_edge_reward": 0.03444803040474653,
      "rewards/forensics_tiou_reward": 0.5491763055324554,
      "step": 730
    },
    {
      "completion_length": 24.59375,
      "epoch": 2.2427035330261136,
      "grad_norm": 7.594694137573242,
      "kl": 0.34375,
      "learning_rate": 4.3846153846153845e-07,
      "loss": -0.1371,
      "reward": 0.7973219752311707,
      "reward_std": 0.007747816911432892,
      "reward_std_all": 0.010119713842868805,
      "rewards-max/forensics_sinkhorn_reward": 0.9843786060810089,
      "rewards-max/forensics_strict_edge_reward": 0.5083713233470917,
      "rewards-max/forensics_tiou_reward": 0.9683971405029297,
      "rewards/forensics_sinkhorn_reward": 0.7550401389598846,
      "rewards/forensics_strict_edge_reward": 0.05574401840567589,
      "rewards/forensics_tiou_reward": 0.5265204012393951,
      "step": 731
    },
    {
      "completion_length": 23.578125,
      "epoch": 2.2457757296466974,
      "grad_norm": 5.525391578674316,
      "kl": 0.4150390625,
      "learning_rate": 4.376923076923077e-07,
      "loss": -0.0032,
      "reward": 0.7925181984901428,
      "reward_std": 0.016179729194846004,
      "reward_std_all": 0.0172632698668167,
      "rewards-max/forensics_sinkhorn_reward": 0.9645229876041412,
      "rewards-max/forensics_strict_edge_reward": 0.35608183592557907,
      "rewards-max/forensics_tiou_reward": 0.9195369780063629,
      "rewards/forensics_sinkhorn_reward": 0.797184944152832,
      "rewards/forensics_strict_edge_reward": 0.03528723679482937,
      "rewards/forensics_tiou_reward": 0.5758382976055145,
      "step": 732
    },
    {
      "completion_length": 26.953125,
      "epoch": 2.248847926267281,
      "grad_norm": 5.959832668304443,
      "kl": 0.369140625,
      "learning_rate": 4.369230769230769e-07,
      "loss": 0.0749,
      "reward": 0.7672032415866852,
      "reward_std": 0.01789126731455326,
      "reward_std_all": 0.031063946895301342,
      "rewards-max/forensics_sinkhorn_reward": 0.9787514209747314,
      "rewards-max/forensics_strict_edge_reward": 0.29921089112758636,
      "rewards-max/forensics_tiou_reward": 0.9454470276832581,
      "rewards/forensics_sinkhorn_reward": 0.7286712527275085,
      "rewards/forensics_strict_edge_reward": 0.0332035543397069,
      "rewards/forensics_tiou_reward": 0.5177036821842194,
      "step": 733
    },
    {
      "completion_length": 25.8125,
      "epoch": 2.251920122887865,
      "grad_norm": 6.117930889129639,
      "kl": 0.306640625,
      "learning_rate": 4.3615384615384614e-07,
      "loss": 0.0001,
      "reward": 0.7706914842128754,
      "reward_std": 0.03907710500061512,
      "reward_std_all": 0.038507819175720215,
      "rewards-max/forensics_sinkhorn_reward": 0.991704523563385,
      "rewards-max/forensics_strict_edge_reward": 0.6862294971942902,
      "rewards-max/forensics_tiou_reward": 0.9822367131710052,
      "rewards/forensics_sinkhorn_reward": 0.7573533952236176,
      "rewards/forensics_strict_edge_reward": 0.0924866572022438,
      "rewards/forensics_tiou_reward": 0.5949556827545166,
      "step": 734
    },
    {
      "completion_length": 27.078125,
      "epoch": 2.2549923195084487,
      "grad_norm": 7.650981903076172,
      "kl": 0.4736328125,
      "learning_rate": 4.3538461538461536e-07,
      "loss": 0.075,
      "reward": 0.7619753777980804,
      "reward_std": 0.03322834149003029,
      "reward_std_all": 0.03779369406402111,
      "rewards-max/forensics_sinkhorn_reward": 0.9464825391769409,
      "rewards-max/forensics_strict_edge_reward": 0.07206758437678218,
      "rewards-max/forensics_tiou_reward": 0.8652269542217255,
      "rewards/forensics_sinkhorn_reward": 0.7367883026599884,
      "rewards/forensics_strict_edge_reward": 0.009503272827714682,
      "rewards/forensics_tiou_reward": 0.5311365872621536,
      "step": 735
    },
    {
      "completion_length": 29.6875,
      "epoch": 2.258064516129032,
      "grad_norm": 5.5088934898376465,
      "kl": 0.5146484375,
      "learning_rate": 4.346153846153846e-07,
      "loss": -0.0543,
      "reward": 0.7834418714046478,
      "reward_std": 0.03039465844631195,
      "reward_std_all": 0.029503561556339264,
      "rewards-max/forensics_sinkhorn_reward": 0.9612376391887665,
      "rewards-max/forensics_strict_edge_reward": 0.2601064182817936,
      "rewards-max/forensics_tiou_reward": 0.897821843624115,
      "rewards/forensics_sinkhorn_reward": 0.8014760613441467,
      "rewards/forensics_strict_edge_reward": 0.030016101896762848,
      "rewards/forensics_tiou_reward": 0.6320124566555023,
      "step": 736
    },
    {
      "completion_length": 26.390625,
      "epoch": 2.261136712749616,
      "grad_norm": 5.986352920532227,
      "kl": 0.369140625,
      "learning_rate": 4.338461538461538e-07,
      "loss": -0.0361,
      "reward": 0.7767971754074097,
      "reward_std": 0.02423886675387621,
      "reward_std_all": 0.027002627030014992,
      "rewards-max/forensics_sinkhorn_reward": 0.9896179139614105,
      "rewards-max/forensics_strict_edge_reward": 0.6054387986660004,
      "rewards-max/forensics_tiou_reward": 0.9770695567131042,
      "rewards/forensics_sinkhorn_reward": 0.8032339215278625,
      "rewards/forensics_strict_edge_reward": 0.05167268589138985,
      "rewards/forensics_tiou_reward": 0.6241925954818726,
      "step": 737
    },
    {
      "completion_length": 23.28125,
      "epoch": 2.2642089093701996,
      "grad_norm": 6.189913272857666,
      "kl": 0.5166015625,
      "learning_rate": 4.3307692307692305e-07,
      "loss": -0.0052,
      "reward": 0.7666007876396179,
      "reward_std": 0.054817087948322296,
      "reward_std_all": 0.05075406841933727,
      "rewards-max/forensics_sinkhorn_reward": 0.9798405170440674,
      "rewards-max/forensics_strict_edge_reward": 0.46688374876976013,
      "rewards-max/forensics_tiou_reward": 0.952072948217392,
      "rewards/forensics_sinkhorn_reward": 0.7481352686882019,
      "rewards/forensics_strict_edge_reward": 0.06597068905830383,
      "rewards/forensics_tiou_reward": 0.54740309715271,
      "step": 738
    },
    {
      "completion_length": 24.28125,
      "epoch": 2.2672811059907834,
      "grad_norm": 7.403367042541504,
      "kl": 0.46484375,
      "learning_rate": 4.3230769230769227e-07,
      "loss": 0.0682,
      "reward": 0.7549645006656647,
      "reward_std": 0.04069257527589798,
      "reward_std_all": 0.04825534485280514,
      "rewards-max/forensics_sinkhorn_reward": 0.9970746636390686,
      "rewards-max/forensics_strict_edge_reward": 0.8465203046798706,
      "rewards-max/forensics_tiou_reward": 0.9987063407897949,
      "rewards/forensics_sinkhorn_reward": 0.6776012480258942,
      "rewards/forensics_strict_edge_reward": 0.11932516098022461,
      "rewards/forensics_tiou_reward": 0.5056785643100739,
      "step": 739
    },
    {
      "completion_length": 25.5625,
      "epoch": 2.270353302611367,
      "grad_norm": 5.533688068389893,
      "kl": 0.4140625,
      "learning_rate": 4.3153846153846153e-07,
      "loss": -0.0627,
      "reward": 0.7755177915096283,
      "reward_std": 0.04450812190771103,
      "reward_std_all": 0.04187968000769615,
      "rewards-max/forensics_sinkhorn_reward": 0.9796823561191559,
      "rewards-max/forensics_strict_edge_reward": 0.608120322227478,
      "rewards-max/forensics_tiou_reward": 0.9559451639652252,
      "rewards/forensics_sinkhorn_reward": 0.7203781008720398,
      "rewards/forensics_strict_edge_reward": 0.041985297575592995,
      "rewards/forensics_tiou_reward": 0.4554404318332672,
      "step": 740
    },
    {
      "completion_length": 26.109375,
      "epoch": 2.273425499231951,
      "grad_norm": 8.85222053527832,
      "kl": 0.384765625,
      "learning_rate": 4.307692307692308e-07,
      "loss": 0.0687,
      "reward": 0.780571460723877,
      "reward_std": 0.03597056865692139,
      "reward_std_all": 0.034263365902006626,
      "rewards-max/forensics_sinkhorn_reward": 0.9683241248130798,
      "rewards-max/forensics_strict_edge_reward": 0.18642350286245346,
      "rewards-max/forensics_tiou_reward": 0.9138774573802948,
      "rewards/forensics_sinkhorn_reward": 0.7454676628112793,
      "rewards/forensics_strict_edge_reward": 0.019142402336001396,
      "rewards/forensics_tiou_reward": 0.5591548681259155,
      "step": 741
    },
    {
      "completion_length": 22.578125,
      "epoch": 2.2764976958525347,
      "grad_norm": 7.202166557312012,
      "kl": 0.46484375,
      "learning_rate": 4.2999999999999996e-07,
      "loss": 0.0583,
      "reward": 0.7785000801086426,
      "reward_std": 0.029363461770117283,
      "reward_std_all": 0.029702194966375828,
      "rewards-max/forensics_sinkhorn_reward": 0.9893528819084167,
      "rewards-max/forensics_strict_edge_reward": 0.3226740062236786,
      "rewards-max/forensics_tiou_reward": 0.9780097901821136,
      "rewards/forensics_sinkhorn_reward": 0.721848338842392,
      "rewards/forensics_strict_edge_reward": 0.0354336928576231,
      "rewards/forensics_tiou_reward": 0.5217437148094177,
      "step": 742
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.279569892473118,
      "grad_norm": 7.4081573486328125,
      "kl": 0.435546875,
      "learning_rate": 4.292307692307692e-07,
      "loss": 0.0379,
      "reward": 0.7644643187522888,
      "reward_std": 0.03802381455898285,
      "reward_std_all": 0.039500366896390915,
      "rewards-max/forensics_sinkhorn_reward": 0.9823958575725555,
      "rewards-max/forensics_strict_edge_reward": 0.4511003717780113,
      "rewards-max/forensics_tiou_reward": 0.9497747719287872,
      "rewards/forensics_sinkhorn_reward": 0.7260390222072601,
      "rewards/forensics_strict_edge_reward": 0.03551747929304838,
      "rewards/forensics_tiou_reward": 0.518890842795372,
      "step": 743
    },
    {
      "completion_length": 24.609375,
      "epoch": 2.282642089093702,
      "grad_norm": 7.154644012451172,
      "kl": 0.3876953125,
      "learning_rate": 4.2846153846153844e-07,
      "loss": 0.0826,
      "reward": 0.7255823612213135,
      "reward_std": 0.07193965651094913,
      "reward_std_all": 0.07632285729050636,
      "rewards-max/forensics_sinkhorn_reward": 0.9975062608718872,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.6730434000492096,
      "rewards/forensics_strict_edge_reward": 0.07187010534107685,
      "rewards/forensics_tiou_reward": 0.5053895562887192,
      "step": 744
    },
    {
      "completion_length": 26.296875,
      "epoch": 2.2857142857142856,
      "grad_norm": 6.518239974975586,
      "kl": 0.5234375,
      "learning_rate": 4.276923076923077e-07,
      "loss": 0.0798,
      "reward": 0.7486681342124939,
      "reward_std": 0.05559857375919819,
      "reward_std_all": 0.05606766417622566,
      "rewards-max/forensics_sinkhorn_reward": 0.9704758524894714,
      "rewards-max/forensics_strict_edge_reward": 0.27603568881750107,
      "rewards-max/forensics_tiou_reward": 0.9268046021461487,
      "rewards/forensics_sinkhorn_reward": 0.7175424695014954,
      "rewards/forensics_strict_edge_reward": 0.013959669973701239,
      "rewards/forensics_tiou_reward": 0.5501693487167358,
      "step": 745
    },
    {
      "completion_length": 26.265625,
      "epoch": 2.2887864823348694,
      "grad_norm": 6.08986759185791,
      "kl": 0.3486328125,
      "learning_rate": 4.2692307692307687e-07,
      "loss": -0.0723,
      "reward": 0.7468080818653107,
      "reward_std": 0.04284961707890034,
      "reward_std_all": 0.0501249972730875,
      "rewards-max/forensics_sinkhorn_reward": 0.9885404407978058,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9731308817863464,
      "rewards/forensics_sinkhorn_reward": 0.7914640307426453,
      "rewards/forensics_strict_edge_reward": 0.07184238359332085,
      "rewards/forensics_tiou_reward": 0.6715275645256042,
      "step": 746
    },
    {
      "completion_length": 26.515625,
      "epoch": 2.291858678955453,
      "grad_norm": 6.093318939208984,
      "kl": 0.37109375,
      "learning_rate": 4.2615384615384613e-07,
      "loss": 0.0757,
      "reward": 0.763296902179718,
      "reward_std": 0.05090089701116085,
      "reward_std_all": 0.04888797365128994,
      "rewards-max/forensics_sinkhorn_reward": 0.9917387962341309,
      "rewards-max/forensics_strict_edge_reward": 0.5050012171268463,
      "rewards-max/forensics_tiou_reward": 0.9839886724948883,
      "rewards/forensics_sinkhorn_reward": 0.70734041929245,
      "rewards/forensics_strict_edge_reward": 0.053748032078146935,
      "rewards/forensics_tiou_reward": 0.4822233319282532,
      "step": 747
    },
    {
      "completion_length": 23.6875,
      "epoch": 2.294930875576037,
      "grad_norm": 7.478752136230469,
      "kl": 0.3876953125,
      "learning_rate": 4.2538461538461535e-07,
      "loss": 0.0292,
      "reward": 0.7965733408927917,
      "reward_std": 0.014572670217603445,
      "reward_std_all": 0.014820398297160864,
      "rewards-max/forensics_sinkhorn_reward": 0.9544779062271118,
      "rewards-max/forensics_strict_edge_reward": 0.46332086622714996,
      "rewards-max/forensics_tiou_reward": 0.8907224833965302,
      "rewards/forensics_sinkhorn_reward": 0.7638186514377594,
      "rewards/forensics_strict_edge_reward": 0.043592057190835476,
      "rewards/forensics_tiou_reward": 0.5575093924999237,
      "step": 748
    },
    {
      "completion_length": 22.46875,
      "epoch": 2.2980030721966207,
      "grad_norm": 6.764714241027832,
      "kl": 0.658203125,
      "learning_rate": 4.246153846153846e-07,
      "loss": 0.035,
      "reward": 0.7823897302150726,
      "reward_std": 0.026674208231270313,
      "reward_std_all": 0.027239411137998104,
      "rewards-max/forensics_sinkhorn_reward": 0.9652009606361389,
      "rewards-max/forensics_strict_edge_reward": 0.2353852316737175,
      "rewards-max/forensics_tiou_reward": 0.9126200675964355,
      "rewards/forensics_sinkhorn_reward": 0.7894885838031769,
      "rewards/forensics_strict_edge_reward": 0.01326400088146329,
      "rewards/forensics_tiou_reward": 0.5490304827690125,
      "step": 749
    },
    {
      "completion_length": 25.71875,
      "epoch": 2.3010752688172045,
      "grad_norm": 7.897568225860596,
      "kl": 0.484375,
      "learning_rate": 4.2384615384615383e-07,
      "loss": 0.0229,
      "reward": 0.7271607220172882,
      "reward_std": 0.05242263339459896,
      "reward_std_all": 0.0667655523866415,
      "rewards-max/forensics_sinkhorn_reward": 0.9885500073432922,
      "rewards-max/forensics_strict_edge_reward": 0.369280107319355,
      "rewards-max/forensics_tiou_reward": 0.9664406776428223,
      "rewards/forensics_sinkhorn_reward": 0.7109981775283813,
      "rewards/forensics_strict_edge_reward": 0.06395438965409994,
      "rewards/forensics_tiou_reward": 0.5323614776134491,
      "step": 750
    },
    {
      "completion_length": 25.65625,
      "epoch": 2.3041474654377883,
      "grad_norm": 6.969186305999756,
      "kl": 0.38671875,
      "learning_rate": 4.2307692307692304e-07,
      "loss": -0.0199,
      "reward": 0.7607197463512421,
      "reward_std": 0.03578322380781174,
      "reward_std_all": 0.04269145801663399,
      "rewards-max/forensics_sinkhorn_reward": 0.9878802597522736,
      "rewards-max/forensics_strict_edge_reward": 0.3720540702342987,
      "rewards-max/forensics_tiou_reward": 0.9721804261207581,
      "rewards/forensics_sinkhorn_reward": 0.7216518521308899,
      "rewards/forensics_strict_edge_reward": 0.030563117004930973,
      "rewards/forensics_tiou_reward": 0.46674638986587524,
      "step": 751
    },
    {
      "completion_length": 24.578125,
      "epoch": 2.3072196620583716,
      "grad_norm": 7.155886650085449,
      "kl": 0.615234375,
      "learning_rate": 4.223076923076923e-07,
      "loss": 0.1496,
      "reward": 0.7619097828865051,
      "reward_std": 0.04017166141420603,
      "reward_std_all": 0.04473440162837505,
      "rewards-max/forensics_sinkhorn_reward": 0.9811443388462067,
      "rewards-max/forensics_strict_edge_reward": 0.6257185339927673,
      "rewards-max/forensics_tiou_reward": 0.9597567915916443,
      "rewards/forensics_sinkhorn_reward": 0.7139301896095276,
      "rewards/forensics_strict_edge_reward": 0.05604854226112366,
      "rewards/forensics_tiou_reward": 0.5187913477420807,
      "step": 752
    },
    {
      "completion_length": 25.140625,
      "epoch": 2.3102918586789554,
      "grad_norm": 9.612821578979492,
      "kl": 0.62109375,
      "learning_rate": 4.215384615384615e-07,
      "loss": -0.0056,
      "reward": 0.7870070040225983,
      "reward_std": 0.03262443374842405,
      "reward_std_all": 0.03020192962139845,
      "rewards-max/forensics_sinkhorn_reward": 0.9902285635471344,
      "rewards-max/forensics_strict_edge_reward": 0.6271176636219025,
      "rewards-max/forensics_tiou_reward": 0.981764942407608,
      "rewards/forensics_sinkhorn_reward": 0.8041918873786926,
      "rewards/forensics_strict_edge_reward": 0.047071075066924095,
      "rewards/forensics_tiou_reward": 0.6034027636051178,
      "step": 753
    },
    {
      "completion_length": 26.3125,
      "epoch": 2.313364055299539,
      "grad_norm": 6.230573654174805,
      "kl": 0.4609375,
      "learning_rate": 4.207692307692308e-07,
      "loss": -0.0404,
      "reward": 0.7967216372489929,
      "reward_std": 0.01243800693191588,
      "reward_std_all": 0.013417375274002552,
      "rewards-max/forensics_sinkhorn_reward": 0.9898853600025177,
      "rewards-max/forensics_strict_edge_reward": 0.47859422117471695,
      "rewards-max/forensics_tiou_reward": 0.9790570735931396,
      "rewards/forensics_sinkhorn_reward": 0.8120919466018677,
      "rewards/forensics_strict_edge_reward": 0.04431761847808957,
      "rewards/forensics_tiou_reward": 0.6150391697883606,
      "step": 754
    },
    {
      "completion_length": 23.3125,
      "epoch": 2.316436251920123,
      "grad_norm": 9.373584747314453,
      "kl": 0.515625,
      "learning_rate": 4.1999999999999995e-07,
      "loss": 0.0914,
      "reward": 0.774953156709671,
      "reward_std": 0.015191126614809036,
      "reward_std_all": 0.024067006539553404,
      "rewards-max/forensics_sinkhorn_reward": 0.987833708524704,
      "rewards-max/forensics_strict_edge_reward": 0.575577899813652,
      "rewards-max/forensics_tiou_reward": 0.9702381491661072,
      "rewards/forensics_sinkhorn_reward": 0.7194421291351318,
      "rewards/forensics_strict_edge_reward": 0.03530511725693941,
      "rewards/forensics_tiou_reward": 0.49615514278411865,
      "step": 755
    },
    {
      "completion_length": 25.453125,
      "epoch": 2.3195084485407067,
      "grad_norm": 5.979422092437744,
      "kl": 0.3251953125,
      "learning_rate": 4.192307692307692e-07,
      "loss": -0.0429,
      "reward": 0.8028430342674255,
      "reward_std": 0.006828030804172158,
      "reward_std_all": 0.007523742504417896,
      "rewards-max/forensics_sinkhorn_reward": 0.9803070425987244,
      "rewards-max/forensics_strict_edge_reward": 0.5587328374385834,
      "rewards-max/forensics_tiou_reward": 0.9557480216026306,
      "rewards/forensics_sinkhorn_reward": 0.822065144777298,
      "rewards/forensics_strict_edge_reward": 0.09909117221832275,
      "rewards/forensics_tiou_reward": 0.6964108645915985,
      "step": 756
    },
    {
      "completion_length": 25.09375,
      "epoch": 2.3225806451612905,
      "grad_norm": 6.511033058166504,
      "kl": 0.294921875,
      "learning_rate": 4.1846153846153843e-07,
      "loss": 0.0239,
      "reward": 0.7595061957836151,
      "reward_std": 0.02261437918059528,
      "reward_std_all": 0.034556593745946884,
      "rewards-max/forensics_sinkhorn_reward": 0.9662386178970337,
      "rewards-max/forensics_strict_edge_reward": 0.5177594628185034,
      "rewards-max/forensics_tiou_reward": 0.9175514578819275,
      "rewards/forensics_sinkhorn_reward": 0.7222805917263031,
      "rewards/forensics_strict_edge_reward": 0.044219116563908756,
      "rewards/forensics_tiou_reward": 0.4810575693845749,
      "step": 757
    },
    {
      "completion_length": 22.640625,
      "epoch": 2.3256528417818743,
      "grad_norm": 7.601139068603516,
      "kl": 0.48828125,
      "learning_rate": 4.176923076923077e-07,
      "loss": 0.0554,
      "reward": 0.7869070768356323,
      "reward_std": 0.030633988790214062,
      "reward_std_all": 0.028523309156298637,
      "rewards-max/forensics_sinkhorn_reward": 0.9968699812889099,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9980162382125854,
      "rewards/forensics_sinkhorn_reward": 0.7779929637908936,
      "rewards/forensics_strict_edge_reward": 0.10086464881896973,
      "rewards/forensics_tiou_reward": 0.5695469975471497,
      "step": 758
    },
    {
      "completion_length": 25.734375,
      "epoch": 2.3287250384024576,
      "grad_norm": 6.55203914642334,
      "kl": 0.4560546875,
      "learning_rate": 4.1692307692307686e-07,
      "loss": 0.0161,
      "reward": 0.7799972295761108,
      "reward_std": 0.032132877968251705,
      "reward_std_all": 0.03166591376066208,
      "rewards-max/forensics_sinkhorn_reward": 0.9819996356964111,
      "rewards-max/forensics_strict_edge_reward": 0.5660533905029297,
      "rewards-max/forensics_tiou_reward": 0.9601570665836334,
      "rewards/forensics_sinkhorn_reward": 0.7613592147827148,
      "rewards/forensics_strict_edge_reward": 0.06381731107831001,
      "rewards/forensics_tiou_reward": 0.5691463053226471,
      "step": 759
    },
    {
      "completion_length": 24.578125,
      "epoch": 2.3317972350230414,
      "grad_norm": 6.010689735412598,
      "kl": 0.42578125,
      "learning_rate": 4.161538461538461e-07,
      "loss": -0.0947,
      "reward": 0.7528837025165558,
      "reward_std": 0.04563724994659424,
      "reward_std_all": 0.04839093051850796,
      "rewards-max/forensics_sinkhorn_reward": 0.9920201301574707,
      "rewards-max/forensics_strict_edge_reward": 0.6697178184986115,
      "rewards-max/forensics_tiou_reward": 0.9848102927207947,
      "rewards/forensics_sinkhorn_reward": 0.6719415485858917,
      "rewards/forensics_strict_edge_reward": 0.062170229852199554,
      "rewards/forensics_tiou_reward": 0.49163106083869934,
      "step": 760
    },
    {
      "completion_length": 27.265625,
      "epoch": 2.334869431643625,
      "grad_norm": 7.055803298950195,
      "kl": 0.4921875,
      "learning_rate": 4.153846153846154e-07,
      "loss": 0.0975,
      "reward": 0.781157523393631,
      "reward_std": 0.03241080604493618,
      "reward_std_all": 0.031456394121050835,
      "rewards-max/forensics_sinkhorn_reward": 0.9910387992858887,
      "rewards-max/forensics_strict_edge_reward": 0.27135664224624634,
      "rewards-max/forensics_tiou_reward": 0.9803990125656128,
      "rewards/forensics_sinkhorn_reward": 0.7268540263175964,
      "rewards/forensics_strict_edge_reward": 0.02331317774951458,
      "rewards/forensics_tiou_reward": 0.5576348304748535,
      "step": 761
    },
    {
      "completion_length": 27.90625,
      "epoch": 2.337941628264209,
      "grad_norm": 6.836459159851074,
      "kl": 0.33984375,
      "learning_rate": 4.146153846153846e-07,
      "loss": 0.0189,
      "reward": 0.7938038408756256,
      "reward_std": 0.019612678384874016,
      "reward_std_all": 0.019370041554793715,
      "rewards-max/forensics_sinkhorn_reward": 0.9796955287456512,
      "rewards-max/forensics_strict_edge_reward": 0.4288998693227768,
      "rewards-max/forensics_tiou_reward": 0.9551436603069305,
      "rewards/forensics_sinkhorn_reward": 0.7571595907211304,
      "rewards/forensics_strict_edge_reward": 0.04808298498392105,
      "rewards/forensics_tiou_reward": 0.6079494655132294,
      "step": 762
    },
    {
      "completion_length": 26.296875,
      "epoch": 2.3410138248847927,
      "grad_norm": 7.863656520843506,
      "kl": 0.330078125,
      "learning_rate": 4.1384615384615387e-07,
      "loss": -0.0007,
      "reward": 0.7875973880290985,
      "reward_std": 0.02515646070241928,
      "reward_std_all": 0.024505996145308018,
      "rewards-max/forensics_sinkhorn_reward": 0.9925746023654938,
      "rewards-max/forensics_strict_edge_reward": 0.5659050345420837,
      "rewards-max/forensics_tiou_reward": 0.9837562143802643,
      "rewards/forensics_sinkhorn_reward": 0.8263035714626312,
      "rewards/forensics_strict_edge_reward": 0.09060235694050789,
      "rewards/forensics_tiou_reward": 0.6602685451507568,
      "step": 763
    },
    {
      "completion_length": 23.203125,
      "epoch": 2.3440860215053765,
      "grad_norm": 7.651838779449463,
      "kl": 0.41015625,
      "learning_rate": 4.1307692307692303e-07,
      "loss": 0.0038,
      "reward": 0.7800145745277405,
      "reward_std": 0.02561346720904112,
      "reward_std_all": 0.02804807759821415,
      "rewards-max/forensics_sinkhorn_reward": 0.9867787957191467,
      "rewards-max/forensics_strict_edge_reward": 0.2672666907310486,
      "rewards-max/forensics_tiou_reward": 0.9650107324123383,
      "rewards/forensics_sinkhorn_reward": 0.7965494394302368,
      "rewards/forensics_strict_edge_reward": 0.02576659806072712,
      "rewards/forensics_tiou_reward": 0.5879306495189667,
      "step": 764
    },
    {
      "completion_length": 23.234375,
      "epoch": 2.3471582181259603,
      "grad_norm": 6.636072158813477,
      "kl": 0.3388671875,
      "learning_rate": 4.123076923076923e-07,
      "loss": 0.0501,
      "reward": 0.7892454266548157,
      "reward_std": 0.02906073071062565,
      "reward_std_all": 0.027162723243236542,
      "rewards-max/forensics_sinkhorn_reward": 0.988359659910202,
      "rewards-max/forensics_strict_edge_reward": 0.5189115703105927,
      "rewards-max/forensics_tiou_reward": 0.9698596298694611,
      "rewards/forensics_sinkhorn_reward": 0.7673355042934418,
      "rewards/forensics_strict_edge_reward": 0.04131903499364853,
      "rewards/forensics_tiou_reward": 0.6125746369361877,
      "step": 765
    },
    {
      "completion_length": 25.578125,
      "epoch": 2.3502304147465436,
      "grad_norm": 6.961221218109131,
      "kl": 0.3310546875,
      "learning_rate": 4.115384615384615e-07,
      "loss": 0.0059,
      "reward": 0.7880926132202148,
      "reward_std": 0.02970704564359039,
      "reward_std_all": 0.028176101157441735,
      "rewards-max/forensics_sinkhorn_reward": 0.9742896854877472,
      "rewards-max/forensics_strict_edge_reward": 0.4154136925935745,
      "rewards-max/forensics_tiou_reward": 0.9346250891685486,
      "rewards/forensics_sinkhorn_reward": 0.75513556599617,
      "rewards/forensics_strict_edge_reward": 0.03561355173587799,
      "rewards/forensics_tiou_reward": 0.5562860369682312,
      "step": 766
    },
    {
      "completion_length": 24.15625,
      "epoch": 2.3533026113671274,
      "grad_norm": 8.6617431640625,
      "kl": 0.314453125,
      "learning_rate": 4.107692307692308e-07,
      "loss": 0.0433,
      "reward": 0.7560010850429535,
      "reward_std": 0.04045518906787038,
      "reward_std_all": 0.04854974802583456,
      "rewards-max/forensics_sinkhorn_reward": 0.9663171172142029,
      "rewards-max/forensics_strict_edge_reward": 0.5039256066083908,
      "rewards-max/forensics_tiou_reward": 0.9272300601005554,
      "rewards/forensics_sinkhorn_reward": 0.7202900052070618,
      "rewards/forensics_strict_edge_reward": 0.036778160370886326,
      "rewards/forensics_tiou_reward": 0.5138368308544159,
      "step": 767
    },
    {
      "completion_length": 24.765625,
      "epoch": 2.356374807987711,
      "grad_norm": 4.7800774574279785,
      "kl": 0.419921875,
      "learning_rate": 4.0999999999999994e-07,
      "loss": 0.1354,
      "reward": 0.7647939920425415,
      "reward_std": 0.029381978325545788,
      "reward_std_all": 0.03809163719415665,
      "rewards-max/forensics_sinkhorn_reward": 0.9925141632556915,
      "rewards-max/forensics_strict_edge_reward": 0.27579642087221146,
      "rewards-max/forensics_tiou_reward": 0.9854276478290558,
      "rewards/forensics_sinkhorn_reward": 0.6789260804653168,
      "rewards/forensics_strict_edge_reward": 0.026457609608769417,
      "rewards/forensics_tiou_reward": 0.5429934859275818,
      "step": 768
    },
    {
      "completion_length": 27.171875,
      "epoch": 2.359447004608295,
      "grad_norm": 5.31390905380249,
      "kl": 0.40234375,
      "learning_rate": 4.092307692307692e-07,
      "loss": 0.052,
      "reward": 0.7910022437572479,
      "reward_std": 0.019911186769604683,
      "reward_std_all": 0.02007470466196537,
      "rewards-max/forensics_sinkhorn_reward": 0.9895355105400085,
      "rewards-max/forensics_strict_edge_reward": 0.45126278698444366,
      "rewards-max/forensics_tiou_reward": 0.9789640605449677,
      "rewards/forensics_sinkhorn_reward": 0.7412182092666626,
      "rewards/forensics_strict_edge_reward": 0.0639784000813961,
      "rewards/forensics_tiou_reward": 0.5789195895195007,
      "step": 769
    },
    {
      "completion_length": 27.265625,
      "epoch": 2.3625192012288787,
      "grad_norm": 7.493898391723633,
      "kl": 0.34814453125,
      "learning_rate": 4.0846153846153847e-07,
      "loss": 0.0018,
      "reward": 0.72291499376297,
      "reward_std": 0.04439648427069187,
      "reward_std_all": 0.06061429902911186,
      "rewards-max/forensics_sinkhorn_reward": 0.9900447130203247,
      "rewards-max/forensics_strict_edge_reward": 0.5104228854179382,
      "rewards-max/forensics_tiou_reward": 0.9791589081287384,
      "rewards/forensics_sinkhorn_reward": 0.7173957526683807,
      "rewards/forensics_strict_edge_reward": 0.03967478312551975,
      "rewards/forensics_tiou_reward": 0.5677727907896042,
      "step": 770
    },
    {
      "completion_length": 25.796875,
      "epoch": 2.3655913978494625,
      "grad_norm": 9.726734161376953,
      "kl": 0.294921875,
      "learning_rate": 4.076923076923077e-07,
      "loss": 0.0693,
      "reward": 0.7861056625843048,
      "reward_std": 0.02156968740746379,
      "reward_std_all": 0.022484959103167057,
      "rewards-max/forensics_sinkhorn_reward": 0.986830323934555,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9753322303295135,
      "rewards/forensics_sinkhorn_reward": 0.7578750848770142,
      "rewards/forensics_strict_edge_reward": 0.05808017961680889,
      "rewards/forensics_tiou_reward": 0.55599544942379,
      "step": 771
    },
    {
      "completion_length": 23.796875,
      "epoch": 2.3686635944700463,
      "grad_norm": 7.095589637756348,
      "kl": 0.353515625,
      "learning_rate": 4.069230769230769e-07,
      "loss": -0.0624,
      "reward": 0.7867237329483032,
      "reward_std": 0.011073692934587598,
      "reward_std_all": 0.01568853366188705,
      "rewards-max/forensics_sinkhorn_reward": 0.9836576282978058,
      "rewards-max/forensics_strict_edge_reward": 0.5361751317977905,
      "rewards-max/forensics_tiou_reward": 0.9554786384105682,
      "rewards/forensics_sinkhorn_reward": 0.774566650390625,
      "rewards/forensics_strict_edge_reward": 0.06591927260160446,
      "rewards/forensics_tiou_reward": 0.6117797791957855,
      "step": 772
    },
    {
      "completion_length": 22.609375,
      "epoch": 2.3717357910906296,
      "grad_norm": 9.189444541931152,
      "kl": 0.2705078125,
      "learning_rate": 4.061538461538461e-07,
      "loss": 0.0799,
      "reward": 0.7580145597457886,
      "reward_std": 0.042283281683921814,
      "reward_std_all": 0.04414815828204155,
      "rewards-max/forensics_sinkhorn_reward": 0.9849857985973358,
      "rewards-max/forensics_strict_edge_reward": 0.45270346850156784,
      "rewards-max/forensics_tiou_reward": 0.967386782169342,
      "rewards/forensics_sinkhorn_reward": 0.7276213765144348,
      "rewards/forensics_strict_edge_reward": 0.028806833550333977,
      "rewards/forensics_tiou_reward": 0.4604261517524719,
      "step": 773
    },
    {
      "completion_length": 25.125,
      "epoch": 2.3748079877112134,
      "grad_norm": 7.869714260101318,
      "kl": 0.4228515625,
      "learning_rate": 4.053846153846154e-07,
      "loss": 0.0378,
      "reward": 0.7954366505146027,
      "reward_std": 0.016393079422414303,
      "reward_std_all": 0.01660594344139099,
      "rewards-max/forensics_sinkhorn_reward": 0.9713252186775208,
      "rewards-max/forensics_strict_edge_reward": 0.3243762031197548,
      "rewards-max/forensics_tiou_reward": 0.9359497129917145,
      "rewards/forensics_sinkhorn_reward": 0.7356827259063721,
      "rewards/forensics_strict_edge_reward": 0.024485383182764053,
      "rewards/forensics_tiou_reward": 0.5180778205394745,
      "step": 774
    },
    {
      "completion_length": 26.390625,
      "epoch": 2.377880184331797,
      "grad_norm": 7.781935691833496,
      "kl": 0.4482421875,
      "learning_rate": 4.046153846153846e-07,
      "loss": -0.0165,
      "reward": 0.7857457995414734,
      "reward_std": 0.020514145959168673,
      "reward_std_all": 0.021778141148388386,
      "rewards-max/forensics_sinkhorn_reward": 0.982392281293869,
      "rewards-max/forensics_strict_edge_reward": 0.4548458904027939,
      "rewards-max/forensics_tiou_reward": 0.9563650786876678,
      "rewards/forensics_sinkhorn_reward": 0.7612766921520233,
      "rewards/forensics_strict_edge_reward": 0.04659823141992092,
      "rewards/forensics_tiou_reward": 0.5789870321750641,
      "step": 775
    },
    {
      "completion_length": 23.59375,
      "epoch": 2.380952380952381,
      "grad_norm": 7.385324954986572,
      "kl": 0.4111328125,
      "learning_rate": 4.0384615384615386e-07,
      "loss": 0.1081,
      "reward": 0.755078136920929,
      "reward_std": 0.07577282469719648,
      "reward_std_all": 0.06910340022295713,
      "rewards-max/forensics_sinkhorn_reward": 0.9707460403442383,
      "rewards-max/forensics_strict_edge_reward": 0.38126203045248985,
      "rewards-max/forensics_tiou_reward": 0.9222934246063232,
      "rewards/forensics_sinkhorn_reward": 0.7053577601909637,
      "rewards/forensics_strict_edge_reward": 0.03673593420535326,
      "rewards/forensics_tiou_reward": 0.4763841778039932,
      "step": 776
    },
    {
      "completion_length": 25.078125,
      "epoch": 2.3840245775729647,
      "grad_norm": 4.967263698577881,
      "kl": 0.4833984375,
      "learning_rate": 4.03076923076923e-07,
      "loss": -0.0166,
      "reward": 0.8012271821498871,
      "reward_std": 0.008736429270356894,
      "reward_std_all": 0.009718886110931635,
      "rewards-max/forensics_sinkhorn_reward": 0.991836667060852,
      "rewards-max/forensics_strict_edge_reward": 0.638425350189209,
      "rewards-max/forensics_tiou_reward": 0.9834539890289307,
      "rewards/forensics_sinkhorn_reward": 0.8048698008060455,
      "rewards/forensics_strict_edge_reward": 0.05862163193523884,
      "rewards/forensics_tiou_reward": 0.7131851315498352,
      "step": 777
    },
    {
      "completion_length": 26.0625,
      "epoch": 2.3870967741935485,
      "grad_norm": 5.631503105163574,
      "kl": 0.453125,
      "learning_rate": 4.023076923076923e-07,
      "loss": -0.0798,
      "reward": 0.7891322672367096,
      "reward_std": 0.01516020717099309,
      "reward_std_all": 0.017279574647545815,
      "rewards-max/forensics_sinkhorn_reward": 0.9740473628044128,
      "rewards-max/forensics_strict_edge_reward": 0.43441142141819,
      "rewards-max/forensics_tiou_reward": 0.9363828301429749,
      "rewards/forensics_sinkhorn_reward": 0.7947063446044922,
      "rewards/forensics_strict_edge_reward": 0.046160243451595306,
      "rewards/forensics_tiou_reward": 0.6380482614040375,
      "step": 778
    },
    {
      "completion_length": 28.25,
      "epoch": 2.3901689708141323,
      "grad_norm": 6.351677417755127,
      "kl": 0.3671875,
      "learning_rate": 4.015384615384615e-07,
      "loss": 0.038,
      "reward": 0.7969973683357239,
      "reward_std": 0.015588000416755676,
      "reward_std_all": 0.015429961029440165,
      "rewards-max/forensics_sinkhorn_reward": 0.9817287027835846,
      "rewards-max/forensics_strict_edge_reward": 0.34612782299518585,
      "rewards-max/forensics_tiou_reward": 0.956371545791626,
      "rewards/forensics_sinkhorn_reward": 0.7877970337867737,
      "rewards/forensics_strict_edge_reward": 0.047054581344127655,
      "rewards/forensics_tiou_reward": 0.593553900718689,
      "step": 779
    },
    {
      "completion_length": 24.75,
      "epoch": 2.3932411674347156,
      "grad_norm": 6.221599102020264,
      "kl": 0.396484375,
      "learning_rate": 4.0076923076923077e-07,
      "loss": 0.0791,
      "reward": 0.7826876938343048,
      "reward_std": 0.03887627832591534,
      "reward_std_all": 0.035552436485886574,
      "rewards-max/forensics_sinkhorn_reward": 0.9837498366832733,
      "rewards-max/forensics_strict_edge_reward": 0.5261107087135315,
      "rewards-max/forensics_tiou_reward": 0.9612324833869934,
      "rewards/forensics_sinkhorn_reward": 0.7807163000106812,
      "rewards/forensics_strict_edge_reward": 0.04071083292365074,
      "rewards/forensics_tiou_reward": 0.5834091007709503,
      "step": 780
    },
    {
      "completion_length": 24.0,
      "epoch": 2.3963133640552994,
      "grad_norm": 5.618096828460693,
      "kl": 0.470703125,
      "learning_rate": 4e-07,
      "loss": -0.0151,
      "reward": 0.8001621961593628,
      "reward_std": 0.00983495032414794,
      "reward_std_all": 0.010762246791273355,
      "rewards-max/forensics_sinkhorn_reward": 0.970746636390686,
      "rewards-max/forensics_strict_edge_reward": 0.3560166209936142,
      "rewards-max/forensics_tiou_reward": 0.9297449886798859,
      "rewards/forensics_sinkhorn_reward": 0.8038541376590729,
      "rewards/forensics_strict_edge_reward": 0.03895123861730099,
      "rewards/forensics_tiou_reward": 0.619262158870697,
      "step": 781
    },
    {
      "completion_length": 24.171875,
      "epoch": 2.399385560675883,
      "grad_norm": 8.146854400634766,
      "kl": 0.3642578125,
      "learning_rate": 3.992307692307692e-07,
      "loss": 0.1293,
      "reward": 0.7458517551422119,
      "reward_std": 0.03784189186990261,
      "reward_std_all": 0.048915889114141464,
      "rewards-max/forensics_sinkhorn_reward": 0.9797111749649048,
      "rewards-max/forensics_strict_edge_reward": 0.5249289721250534,
      "rewards-max/forensics_tiou_reward": 0.9453970789909363,
      "rewards/forensics_sinkhorn_reward": 0.7373707592487335,
      "rewards/forensics_strict_edge_reward": 0.053359173238277435,
      "rewards/forensics_tiou_reward": 0.5170495510101318,
      "step": 782
    },
    {
      "completion_length": 27.78125,
      "epoch": 2.402457757296467,
      "grad_norm": 8.891806602478027,
      "kl": 0.380859375,
      "learning_rate": 3.9846153846153846e-07,
      "loss": 0.0977,
      "reward": 0.7849623560905457,
      "reward_std": 0.019701754674315453,
      "reward_std_all": 0.021792233921587467,
      "rewards-max/forensics_sinkhorn_reward": 0.9785626530647278,
      "rewards-max/forensics_strict_edge_reward": 0.5074994415044785,
      "rewards-max/forensics_tiou_reward": 0.9527324736118317,
      "rewards/forensics_sinkhorn_reward": 0.7626773715019226,
      "rewards/forensics_strict_edge_reward": 0.037679336965084076,
      "rewards/forensics_tiou_reward": 0.5588997602462769,
      "step": 783
    },
    {
      "completion_length": 26.03125,
      "epoch": 2.4055299539170507,
      "grad_norm": 6.148967742919922,
      "kl": 0.51953125,
      "learning_rate": 3.976923076923077e-07,
      "loss": 0.0619,
      "reward": 0.7888803780078888,
      "reward_std": 0.02104832883924246,
      "reward_std_all": 0.021460206247866154,
      "rewards-max/forensics_sinkhorn_reward": 0.9931147694587708,
      "rewards-max/forensics_strict_edge_reward": 0.7032848298549652,
      "rewards-max/forensics_tiou_reward": 0.9873029887676239,
      "rewards/forensics_sinkhorn_reward": 0.7946884334087372,
      "rewards/forensics_strict_edge_reward": 0.054249074310064316,
      "rewards/forensics_tiou_reward": 0.6066336631774902,
      "step": 784
    },
    {
      "completion_length": 24.703125,
      "epoch": 2.4086021505376345,
      "grad_norm": 7.0396575927734375,
      "kl": 0.4501953125,
      "learning_rate": 3.969230769230769e-07,
      "loss": -0.0029,
      "reward": 0.7742178738117218,
      "reward_std": 0.03340902831405401,
      "reward_std_all": 0.03393068164587021,
      "rewards-max/forensics_sinkhorn_reward": 0.9918508231639862,
      "rewards-max/forensics_strict_edge_reward": 0.6291489899158478,
      "rewards-max/forensics_tiou_reward": 0.9833638668060303,
      "rewards/forensics_sinkhorn_reward": 0.7797924876213074,
      "rewards/forensics_strict_edge_reward": 0.0867069885134697,
      "rewards/forensics_tiou_reward": 0.5765040516853333,
      "step": 785
    },
    {
      "completion_length": 24.1875,
      "epoch": 2.4116743471582183,
      "grad_norm": 6.597545146942139,
      "kl": 0.3759765625,
      "learning_rate": 3.961538461538461e-07,
      "loss": 0.0414,
      "reward": 0.7790134251117706,
      "reward_std": 0.03277663764311001,
      "reward_std_all": 0.03342448268085718,
      "rewards-max/forensics_sinkhorn_reward": 0.9735713601112366,
      "rewards-max/forensics_strict_edge_reward": 0.4514618217945099,
      "rewards-max/forensics_tiou_reward": 0.9448096454143524,
      "rewards/forensics_sinkhorn_reward": 0.7484623491764069,
      "rewards/forensics_strict_edge_reward": 0.030991942156106234,
      "rewards/forensics_tiou_reward": 0.5437470972537994,
      "step": 786
    },
    {
      "completion_length": 21.1875,
      "epoch": 2.4147465437788016,
      "grad_norm": 7.129003524780273,
      "kl": 0.3876953125,
      "learning_rate": 3.9538461538461537e-07,
      "loss": -0.0644,
      "reward": 0.7846085429191589,
      "reward_std": 0.022427070827689022,
      "reward_std_all": 0.023465041420422494,
      "rewards-max/forensics_sinkhorn_reward": 0.9749515652656555,
      "rewards-max/forensics_strict_edge_reward": 0.36461610347032547,
      "rewards-max/forensics_tiou_reward": 0.9308158457279205,
      "rewards/forensics_sinkhorn_reward": 0.8178835213184357,
      "rewards/forensics_strict_edge_reward": 0.032599362544715405,
      "rewards/forensics_tiou_reward": 0.6353903114795685,
      "step": 787
    },
    {
      "completion_length": 24.46875,
      "epoch": 2.4178187403993854,
      "grad_norm": 7.212502479553223,
      "kl": 0.3349609375,
      "learning_rate": 3.946153846153846e-07,
      "loss": -0.1095,
      "reward": 0.7888605892658234,
      "reward_std": 0.017044979031197727,
      "reward_std_all": 0.018462103558704257,
      "rewards-max/forensics_sinkhorn_reward": 0.9919645488262177,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9869333207607269,
      "rewards/forensics_sinkhorn_reward": 0.8278629183769226,
      "rewards/forensics_strict_edge_reward": 0.10581773892045021,
      "rewards/forensics_tiou_reward": 0.68962162733078,
      "step": 788
    },
    {
      "completion_length": 23.109375,
      "epoch": 2.420890937019969,
      "grad_norm": 6.841639041900635,
      "kl": 0.4287109375,
      "learning_rate": 3.9384615384615385e-07,
      "loss": 0.0584,
      "reward": 0.7039613127708435,
      "reward_std": 0.07239957712590694,
      "reward_std_all": 0.09220970049500465,
      "rewards-max/forensics_sinkhorn_reward": 0.9939706325531006,
      "rewards-max/forensics_strict_edge_reward": 0.6951080709695816,
      "rewards-max/forensics_tiou_reward": 0.9895485043525696,
      "rewards/forensics_sinkhorn_reward": 0.6795498430728912,
      "rewards/forensics_strict_edge_reward": 0.05308349244296551,
      "rewards/forensics_tiou_reward": 0.5090837776660919,
      "step": 789
    },
    {
      "completion_length": 23.546875,
      "epoch": 2.423963133640553,
      "grad_norm": 6.342113494873047,
      "kl": 0.3681640625,
      "learning_rate": 3.9307692307692306e-07,
      "loss": -0.0044,
      "reward": 0.7865955531597137,
      "reward_std": 0.024980424903333187,
      "reward_std_all": 0.02482197992503643,
      "rewards-max/forensics_sinkhorn_reward": 0.9837155938148499,
      "rewards-max/forensics_strict_edge_reward": 0.4257221072912216,
      "rewards-max/forensics_tiou_reward": 0.9655499458312988,
      "rewards/forensics_sinkhorn_reward": 0.7976897060871124,
      "rewards/forensics_strict_edge_reward": 0.045038068667054176,
      "rewards/forensics_tiou_reward": 0.5991405546665192,
      "step": 790
    },
    {
      "completion_length": 24.8125,
      "epoch": 2.4270353302611367,
      "grad_norm": 6.254292011260986,
      "kl": 0.421875,
      "learning_rate": 3.923076923076923e-07,
      "loss": 0.0162,
      "reward": 0.7664771676063538,
      "reward_std": 0.06022595800459385,
      "reward_std_all": 0.055140480399131775,
      "rewards-max/forensics_sinkhorn_reward": 0.9906259179115295,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9802729785442352,
      "rewards/forensics_sinkhorn_reward": 0.7560818791389465,
      "rewards/forensics_strict_edge_reward": 0.08279415965080261,
      "rewards/forensics_tiou_reward": 0.5573713183403015,
      "step": 791
    },
    {
      "completion_length": 26.40625,
      "epoch": 2.4301075268817205,
      "grad_norm": 6.3040900230407715,
      "kl": 0.5068359375,
      "learning_rate": 3.9153846153846154e-07,
      "loss": -0.0093,
      "reward": 0.773961752653122,
      "reward_std": 0.04200987331569195,
      "reward_std_all": 0.04022950679063797,
      "rewards-max/forensics_sinkhorn_reward": 0.9882082343101501,
      "rewards-max/forensics_strict_edge_reward": 0.6069065183401108,
      "rewards-max/forensics_tiou_reward": 0.9783362746238708,
      "rewards/forensics_sinkhorn_reward": 0.7606194615364075,
      "rewards/forensics_strict_edge_reward": 0.06652095494791865,
      "rewards/forensics_tiou_reward": 0.5836021602153778,
      "step": 792
    },
    {
      "completion_length": 22.375,
      "epoch": 2.4331797235023043,
      "grad_norm": 7.307013034820557,
      "kl": 0.3818359375,
      "learning_rate": 3.9076923076923076e-07,
      "loss": -0.0873,
      "reward": 0.7878096699714661,
      "reward_std": 0.01850252551957965,
      "reward_std_all": 0.02003486710600555,
      "rewards-max/forensics_sinkhorn_reward": 0.9908718168735504,
      "rewards-max/forensics_strict_edge_reward": 0.4475281238555908,
      "rewards-max/forensics_tiou_reward": 0.9812487363815308,
      "rewards/forensics_sinkhorn_reward": 0.7702983021736145,
      "rewards/forensics_strict_edge_reward": 0.037756931968033314,
      "rewards/forensics_tiou_reward": 0.5715952068567276,
      "step": 793
    },
    {
      "completion_length": 26.296875,
      "epoch": 2.436251920122888,
      "grad_norm": 7.003600597381592,
      "kl": 0.41943359375,
      "learning_rate": 3.8999999999999997e-07,
      "loss": -0.0167,
      "reward": 0.769720733165741,
      "reward_std": 0.033728585578501225,
      "reward_std_all": 0.03729533217847347,
      "rewards-max/forensics_sinkhorn_reward": 0.9911089241504669,
      "rewards-max/forensics_strict_edge_reward": 0.23518037796020508,
      "rewards-max/forensics_tiou_reward": 0.9802626371383667,
      "rewards/forensics_sinkhorn_reward": 0.733365535736084,
      "rewards/forensics_strict_edge_reward": 0.02852315455675125,
      "rewards/forensics_tiou_reward": 0.5811481773853302,
      "step": 794
    },
    {
      "completion_length": 31.703125,
      "epoch": 2.439324116743472,
      "grad_norm": 6.327180862426758,
      "kl": 0.458984375,
      "learning_rate": 3.892307692307692e-07,
      "loss": 0.0559,
      "reward": 0.7369509637355804,
      "reward_std": 0.07257961854338646,
      "reward_std_all": 0.07095700316131115,
      "rewards-max/forensics_sinkhorn_reward": 0.9724356234073639,
      "rewards-max/forensics_strict_edge_reward": 0.19672978296875954,
      "rewards-max/forensics_tiou_reward": 0.9378496408462524,
      "rewards/forensics_sinkhorn_reward": 0.6563230752944946,
      "rewards/forensics_strict_edge_reward": 0.025145783787593246,
      "rewards/forensics_tiou_reward": 0.5057654827833176,
      "step": 795
    },
    {
      "completion_length": 26.515625,
      "epoch": 2.442396313364055,
      "grad_norm": 9.119251251220703,
      "kl": 0.443359375,
      "learning_rate": 3.8846153846153845e-07,
      "loss": -0.0302,
      "reward": 0.7705588936805725,
      "reward_std": 0.026973950676620007,
      "reward_std_all": 0.031539215706288815,
      "rewards-max/forensics_sinkhorn_reward": 0.9948359131813049,
      "rewards-max/forensics_strict_edge_reward": 0.5112913623452187,
      "rewards-max/forensics_tiou_reward": 0.9903986752033234,
      "rewards/forensics_sinkhorn_reward": 0.7689341902732849,
      "rewards/forensics_strict_edge_reward": 0.039111591409891844,
      "rewards/forensics_tiou_reward": 0.5723758041858673,
      "step": 796
    },
    {
      "completion_length": 22.734375,
      "epoch": 2.445468509984639,
      "grad_norm": 9.875432014465332,
      "kl": 0.3779296875,
      "learning_rate": 3.8769230769230767e-07,
      "loss": 0.0422,
      "reward": 0.7797949016094208,
      "reward_std": 0.02894569654017687,
      "reward_std_all": 0.02952106110751629,
      "rewards-max/forensics_sinkhorn_reward": 0.9950907826423645,
      "rewards-max/forensics_strict_edge_reward": 0.6851765215396881,
      "rewards-max/forensics_tiou_reward": 0.9919711351394653,
      "rewards/forensics_sinkhorn_reward": 0.7068031430244446,
      "rewards/forensics_strict_edge_reward": 0.05355863273143768,
      "rewards/forensics_tiou_reward": 0.5771176218986511,
      "step": 797
    },
    {
      "completion_length": 24.484375,
      "epoch": 2.4485407066052227,
      "grad_norm": 6.804983615875244,
      "kl": 0.5341796875,
      "learning_rate": 3.8692307692307693e-07,
      "loss": 0.0595,
      "reward": 0.7520293891429901,
      "reward_std": 0.03495929250493646,
      "reward_std_all": 0.045515756122767925,
      "rewards-max/forensics_sinkhorn_reward": 0.9829851686954498,
      "rewards-max/forensics_strict_edge_reward": 0.48453274369239807,
      "rewards-max/forensics_tiou_reward": 0.9572062492370605,
      "rewards/forensics_sinkhorn_reward": 0.7242242693901062,
      "rewards/forensics_strict_edge_reward": 0.031632788479328156,
      "rewards/forensics_tiou_reward": 0.5409336388111115,
      "step": 798
    },
    {
      "completion_length": 26.25,
      "epoch": 2.4516129032258065,
      "grad_norm": 7.9171061515808105,
      "kl": 0.3916015625,
      "learning_rate": 3.8615384615384615e-07,
      "loss": -0.0489,
      "reward": 0.7906030118465424,
      "reward_std": 0.014822309836745262,
      "reward_std_all": 0.01686606789007783,
      "rewards-max/forensics_sinkhorn_reward": 0.9680677354335785,
      "rewards-max/forensics_strict_edge_reward": 0.2783918157219887,
      "rewards-max/forensics_tiou_reward": 0.9163536131381989,
      "rewards/forensics_sinkhorn_reward": 0.7677752673625946,
      "rewards/forensics_strict_edge_reward": 0.025215138914063573,
      "rewards/forensics_tiou_reward": 0.5851896405220032,
      "step": 799
    },
    {
      "completion_length": 23.875,
      "epoch": 2.4546850998463903,
      "grad_norm": 8.031664848327637,
      "kl": 0.4111328125,
      "learning_rate": 3.8538461538461536e-07,
      "loss": 0.068,
      "reward": 0.782302588224411,
      "reward_std": 0.040298039093613625,
      "reward_std_all": 0.03699174802750349,
      "rewards-max/forensics_sinkhorn_reward": 0.9846743047237396,
      "rewards-max/forensics_strict_edge_reward": 0.49984145164489746,
      "rewards-max/forensics_tiou_reward": 0.967255562543869,
      "rewards/forensics_sinkhorn_reward": 0.764811784029007,
      "rewards/forensics_strict_edge_reward": 0.04730812460184097,
      "rewards/forensics_tiou_reward": 0.5656608045101166,
      "step": 800
    },
    {
      "completion_length": 25.875,
      "epoch": 2.457757296466974,
      "grad_norm": 7.073458194732666,
      "kl": 0.4658203125,
      "learning_rate": 3.8461538461538463e-07,
      "loss": 0.0173,
      "reward": 0.7687886655330658,
      "reward_std": 0.03735128045082092,
      "reward_std_all": 0.03815420623868704,
      "rewards-max/forensics_sinkhorn_reward": 0.9918071031570435,
      "rewards-max/forensics_strict_edge_reward": 0.6018703132867813,
      "rewards-max/forensics_tiou_reward": 0.9856353402137756,
      "rewards/forensics_sinkhorn_reward": 0.7282200157642365,
      "rewards/forensics_strict_edge_reward": 0.07557754777371883,
      "rewards/forensics_tiou_reward": 0.5568550229072571,
      "step": 801
    },
    {
      "completion_length": 23.875,
      "epoch": 2.460829493087558,
      "grad_norm": 7.932344436645508,
      "kl": 0.40625,
      "learning_rate": 3.8384615384615384e-07,
      "loss": 0.0817,
      "reward": 0.7353830635547638,
      "reward_std": 0.038430606946349144,
      "reward_std_all": 0.05800306983292103,
      "rewards-max/forensics_sinkhorn_reward": 0.9849450290203094,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9696964621543884,
      "rewards/forensics_sinkhorn_reward": 0.6941980123519897,
      "rewards/forensics_strict_edge_reward": 0.04718996770679951,
      "rewards/forensics_tiou_reward": 0.4810390770435333,
      "step": 802
    },
    {
      "completion_length": 26.5625,
      "epoch": 2.463901689708141,
      "grad_norm": 7.509507656097412,
      "kl": 0.529296875,
      "learning_rate": 3.8307692307692305e-07,
      "loss": 0.0431,
      "reward": 0.7824069559574127,
      "reward_std": 0.026117917150259018,
      "reward_std_all": 0.026463113725185394,
      "rewards-max/forensics_sinkhorn_reward": 0.9915041625499725,
      "rewards-max/forensics_strict_edge_reward": 0.6213910579681396,
      "rewards-max/forensics_tiou_reward": 0.9829157292842865,
      "rewards/forensics_sinkhorn_reward": 0.7717793881893158,
      "rewards/forensics_strict_edge_reward": 0.05640403740108013,
      "rewards/forensics_tiou_reward": 0.5658184587955475,
      "step": 803
    },
    {
      "completion_length": 23.328125,
      "epoch": 2.466973886328725,
      "grad_norm": 7.560812473297119,
      "kl": 0.419921875,
      "learning_rate": 3.8230769230769227e-07,
      "loss": -0.0104,
      "reward": 0.7431517243385315,
      "reward_std": 0.040111277252435684,
      "reward_std_all": 0.05453115329146385,
      "rewards-max/forensics_sinkhorn_reward": 0.9751966893672943,
      "rewards-max/forensics_strict_edge_reward": 0.5499051213264465,
      "rewards-max/forensics_tiou_reward": 0.9354477524757385,
      "rewards/forensics_sinkhorn_reward": 0.6947483718395233,
      "rewards/forensics_strict_edge_reward": 0.045062776654958725,
      "rewards/forensics_tiou_reward": 0.5117048919200897,
      "step": 804
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.4700460829493087,
      "grad_norm": 8.443219184875488,
      "kl": 0.4541015625,
      "learning_rate": 3.8153846153846153e-07,
      "loss": 0.0686,
      "reward": 0.7822081446647644,
      "reward_std": 0.02667667344212532,
      "reward_std_all": 0.02700053621083498,
      "rewards-max/forensics_sinkhorn_reward": 0.9905146658420563,
      "rewards-max/forensics_strict_edge_reward": 0.2929607331752777,
      "rewards-max/forensics_tiou_reward": 0.9811969697475433,
      "rewards/forensics_sinkhorn_reward": 0.7341307997703552,
      "rewards/forensics_strict_edge_reward": 0.02757628634572029,
      "rewards/forensics_tiou_reward": 0.5221691727638245,
      "step": 805
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.4731182795698925,
      "grad_norm": 7.612294673919678,
      "kl": 0.484375,
      "learning_rate": 3.8076923076923075e-07,
      "loss": -0.0199,
      "reward": 0.795587956905365,
      "reward_std": 0.009823299362324178,
      "reward_std_all": 0.012053830898366868,
      "rewards-max/forensics_sinkhorn_reward": 0.980061411857605,
      "rewards-max/forensics_strict_edge_reward": 0.33878660947084427,
      "rewards-max/forensics_tiou_reward": 0.9559473097324371,
      "rewards/forensics_sinkhorn_reward": 0.7582188844680786,
      "rewards/forensics_strict_edge_reward": 0.028042211662977934,
      "rewards/forensics_tiou_reward": 0.5433502197265625,
      "step": 806
    },
    {
      "completion_length": 24.875,
      "epoch": 2.4761904761904763,
      "grad_norm": 8.10153579711914,
      "kl": 0.3623046875,
      "learning_rate": 3.7999999999999996e-07,
      "loss": -0.0718,
      "reward": 0.7877340018749237,
      "reward_std": 0.008481496712192893,
      "reward_std_all": 0.015482795890420675,
      "rewards-max/forensics_sinkhorn_reward": 0.9703752994537354,
      "rewards-max/forensics_strict_edge_reward": 0.22578205168247223,
      "rewards-max/forensics_tiou_reward": 0.9319236576557159,
      "rewards/forensics_sinkhorn_reward": 0.7181919813156128,
      "rewards/forensics_strict_edge_reward": 0.01607827004045248,
      "rewards/forensics_tiou_reward": 0.46109166741371155,
      "step": 807
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.47926267281106,
      "grad_norm": 7.901859283447266,
      "kl": 0.3798828125,
      "learning_rate": 3.792307692307692e-07,
      "loss": -0.0008,
      "reward": 0.7862077951431274,
      "reward_std": 0.02892138622701168,
      "reward_std_all": 0.027932438999414444,
      "rewards-max/forensics_sinkhorn_reward": 0.9915095269680023,
      "rewards-max/forensics_strict_edge_reward": 0.5887030363082886,
      "rewards-max/forensics_tiou_reward": 0.9838305711746216,
      "rewards/forensics_sinkhorn_reward": 0.7857484519481659,
      "rewards/forensics_strict_edge_reward": 0.05965644400566816,
      "rewards/forensics_tiou_reward": 0.6086008548736572,
      "step": 808
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.482334869431644,
      "grad_norm": 7.51352071762085,
      "kl": 0.4326171875,
      "learning_rate": 3.7846153846153844e-07,
      "loss": 0.0674,
      "reward": 0.7832672297954559,
      "reward_std": 0.024340469390153885,
      "reward_std_all": 0.025019115768373013,
      "rewards-max/forensics_sinkhorn_reward": 0.960403710603714,
      "rewards-max/forensics_strict_edge_reward": 0.298451105132699,
      "rewards-max/forensics_tiou_reward": 0.9097061455249786,
      "rewards/forensics_sinkhorn_reward": 0.726885974407196,
      "rewards/forensics_strict_edge_reward": 0.028433787636458874,
      "rewards/forensics_tiou_reward": 0.5633091479539871,
      "step": 809
    },
    {
      "completion_length": 28.125,
      "epoch": 2.485407066052227,
      "grad_norm": 8.261204719543457,
      "kl": 0.400390625,
      "learning_rate": 3.776923076923077e-07,
      "loss": 0.0131,
      "reward": 0.7464011311531067,
      "reward_std": 0.05163401924073696,
      "reward_std_all": 0.05388274043798447,
      "rewards-max/forensics_sinkhorn_reward": 0.9794059097766876,
      "rewards-max/forensics_strict_edge_reward": 0.4258756637573242,
      "rewards-max/forensics_tiou_reward": 0.9436348080635071,
      "rewards/forensics_sinkhorn_reward": 0.7249434590339661,
      "rewards/forensics_strict_edge_reward": 0.034457020461559296,
      "rewards/forensics_tiou_reward": 0.514562577009201,
      "step": 810
    },
    {
      "completion_length": 25.609375,
      "epoch": 2.488479262672811,
      "grad_norm": 8.197881698608398,
      "kl": 0.466796875,
      "learning_rate": 3.769230769230769e-07,
      "loss": 0.0817,
      "reward": 0.7859300673007965,
      "reward_std": 0.02156653441488743,
      "reward_std_all": 0.022505829110741615,
      "rewards-max/forensics_sinkhorn_reward": 0.9913068115711212,
      "rewards-max/forensics_strict_edge_reward": 0.6736744344234467,
      "rewards-max/forensics_tiou_reward": 0.9824237525463104,
      "rewards/forensics_sinkhorn_reward": 0.7447963953018188,
      "rewards/forensics_strict_edge_reward": 0.04326254315674305,
      "rewards/forensics_tiou_reward": 0.5578848719596863,
      "step": 811
    },
    {
      "completion_length": 25.296875,
      "epoch": 2.4915514592933947,
      "grad_norm": 8.437742233276367,
      "kl": 0.3955078125,
      "learning_rate": 3.7615384615384614e-07,
      "loss": 0.078,
      "reward": 0.7129054367542267,
      "reward_std": 0.059894856065511703,
      "reward_std_all": 0.07558180578052998,
      "rewards-max/forensics_sinkhorn_reward": 0.9792073369026184,
      "rewards-max/forensics_strict_edge_reward": 0.16654029116034508,
      "rewards-max/forensics_tiou_reward": 0.9469620287418365,
      "rewards/forensics_sinkhorn_reward": 0.631809800863266,
      "rewards/forensics_strict_edge_reward": 0.017149158287793398,
      "rewards/forensics_tiou_reward": 0.4253724068403244,
      "step": 812
    },
    {
      "completion_length": 26.734375,
      "epoch": 2.4946236559139785,
      "grad_norm": 6.9723663330078125,
      "kl": 0.4052734375,
      "learning_rate": 3.7538461538461535e-07,
      "loss": -0.064,
      "reward": 0.8050651252269745,
      "reward_std": 0.0024514906108379364,
      "reward_std_all": 0.004202952375635505,
      "rewards-max/forensics_sinkhorn_reward": 0.9972748160362244,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9991350471973419,
      "rewards/forensics_sinkhorn_reward": 0.8262074887752533,
      "rewards/forensics_strict_edge_reward": 0.09463372081518173,
      "rewards/forensics_tiou_reward": 0.6705223917961121,
      "step": 813
    },
    {
      "completion_length": 30.078125,
      "epoch": 2.4976958525345623,
      "grad_norm": 6.349152565002441,
      "kl": 0.466796875,
      "learning_rate": 3.746153846153846e-07,
      "loss": -0.087,
      "reward": 0.803477942943573,
      "reward_std": 0.0049694705521687865,
      "reward_std_all": 0.006466463208198547,
      "rewards-max/forensics_sinkhorn_reward": 0.9951187372207642,
      "rewards-max/forensics_strict_edge_reward": 0.6838940382003784,
      "rewards-max/forensics_tiou_reward": 0.9915244579315186,
      "rewards/forensics_sinkhorn_reward": 0.8299105167388916,
      "rewards/forensics_strict_edge_reward": 0.08327123522758484,
      "rewards/forensics_tiou_reward": 0.6658865213394165,
      "step": 814
    },
    {
      "completion_length": 24.140625,
      "epoch": 2.500768049155146,
      "grad_norm": 8.70930004119873,
      "kl": 0.396484375,
      "learning_rate": 3.7384615384615383e-07,
      "loss": -0.0652,
      "reward": 0.7698506116867065,
      "reward_std": 0.045466140494681895,
      "reward_std_all": 0.04351931530982256,
      "rewards-max/forensics_sinkhorn_reward": 0.9925187230110168,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9870639145374298,
      "rewards/forensics_sinkhorn_reward": 0.7446149289608002,
      "rewards/forensics_strict_edge_reward": 0.0801120437681675,
      "rewards/forensics_tiou_reward": 0.5699781477451324,
      "step": 815
    },
    {
      "completion_length": 23.796875,
      "epoch": 2.50384024577573,
      "grad_norm": 5.505352020263672,
      "kl": 0.4033203125,
      "learning_rate": 3.7307692307692304e-07,
      "loss": -0.0141,
      "reward": 0.7938914000988007,
      "reward_std": 0.02158671838697046,
      "reward_std_all": 0.0206473603611812,
      "rewards-max/forensics_sinkhorn_reward": 0.9836336672306061,
      "rewards-max/forensics_strict_edge_reward": 0.5497493892908096,
      "rewards-max/forensics_tiou_reward": 0.9555497169494629,
      "rewards/forensics_sinkhorn_reward": 0.8457545042037964,
      "rewards/forensics_strict_edge_reward": 0.06254669278860092,
      "rewards/forensics_tiou_reward": 0.6626186966896057,
      "step": 816
    },
    {
      "completion_length": 23.96875,
      "epoch": 2.506912442396313,
      "grad_norm": 7.089393138885498,
      "kl": 0.451171875,
      "learning_rate": 3.7230769230769226e-07,
      "loss": 0.0093,
      "reward": 0.7750764191150665,
      "reward_std": 0.041460826992988586,
      "reward_std_all": 0.03976248949766159,
      "rewards-max/forensics_sinkhorn_reward": 0.9741358458995819,
      "rewards-max/forensics_strict_edge_reward": 0.4968930333852768,
      "rewards-max/forensics_tiou_reward": 0.9401527345180511,
      "rewards/forensics_sinkhorn_reward": 0.7336195409297943,
      "rewards/forensics_strict_edge_reward": 0.03286350518465042,
      "rewards/forensics_tiou_reward": 0.4984688460826874,
      "step": 817
    },
    {
      "completion_length": 25.359375,
      "epoch": 2.509984639016897,
      "grad_norm": 5.027641296386719,
      "kl": 0.359375,
      "learning_rate": 3.715384615384615e-07,
      "loss": -0.0159,
      "reward": 0.7643784582614899,
      "reward_std": 0.03258526185527444,
      "reward_std_all": 0.036878906190395355,
      "rewards-max/forensics_sinkhorn_reward": 0.9848129749298096,
      "rewards-max/forensics_strict_edge_reward": 0.29302991926670074,
      "rewards-max/forensics_tiou_reward": 0.9627552628517151,
      "rewards/forensics_sinkhorn_reward": 0.7446340918540955,
      "rewards/forensics_strict_edge_reward": 0.0347448056563735,
      "rewards/forensics_tiou_reward": 0.5374898463487625,
      "step": 818
    },
    {
      "completion_length": 28.1875,
      "epoch": 2.5130568356374807,
      "grad_norm": 6.0638322830200195,
      "kl": 0.416015625,
      "learning_rate": 3.707692307692308e-07,
      "loss": 0.0323,
      "reward": 0.7966210544109344,
      "reward_std": 0.017386562190949917,
      "reward_std_all": 0.01681164000183344,
      "rewards-max/forensics_sinkhorn_reward": 0.9859111607074738,
      "rewards-max/forensics_strict_edge_reward": 0.384507954120636,
      "rewards-max/forensics_tiou_reward": 0.9677169919013977,
      "rewards/forensics_sinkhorn_reward": 0.8318474590778351,
      "rewards/forensics_strict_edge_reward": 0.04702863097190857,
      "rewards/forensics_tiou_reward": 0.6822109818458557,
      "step": 819
    },
    {
      "completion_length": 28.5625,
      "epoch": 2.5161290322580645,
      "grad_norm": 6.41172456741333,
      "kl": 0.451171875,
      "learning_rate": 3.7e-07,
      "loss": 0.092,
      "reward": 0.7566149830818176,
      "reward_std": 0.04945036955177784,
      "reward_std_all": 0.05296717770397663,
      "rewards-max/forensics_sinkhorn_reward": 0.9844441711902618,
      "rewards-max/forensics_strict_edge_reward": 0.412410169839859,
      "rewards-max/forensics_tiou_reward": 0.965514063835144,
      "rewards/forensics_sinkhorn_reward": 0.7016825079917908,
      "rewards/forensics_strict_edge_reward": 0.027576440945267677,
      "rewards/forensics_tiou_reward": 0.5589068233966827,
      "step": 820
    },
    {
      "completion_length": 27.171875,
      "epoch": 2.5192012288786483,
      "grad_norm": 5.353311061859131,
      "kl": 0.3544921875,
      "learning_rate": 3.692307692307692e-07,
      "loss": 0.1039,
      "reward": 0.7656204998493195,
      "reward_std": 0.04439181834459305,
      "reward_std_all": 0.04617496021091938,
      "rewards-max/forensics_sinkhorn_reward": 0.9900098145008087,
      "rewards-max/forensics_strict_edge_reward": 0.6150864660739899,
      "rewards-max/forensics_tiou_reward": 0.9788773655891418,
      "rewards/forensics_sinkhorn_reward": 0.7403101325035095,
      "rewards/forensics_strict_edge_reward": 0.08890871331095695,
      "rewards/forensics_tiou_reward": 0.6099173426628113,
      "step": 821
    },
    {
      "completion_length": 25.578125,
      "epoch": 2.522273425499232,
      "grad_norm": 5.79367208480835,
      "kl": 0.48828125,
      "learning_rate": 3.6846153846153843e-07,
      "loss": 0.0524,
      "reward": 0.772015780210495,
      "reward_std": 0.030045046471059322,
      "reward_std_all": 0.03199971001595259,
      "rewards-max/forensics_sinkhorn_reward": 0.9812765121459961,
      "rewards-max/forensics_strict_edge_reward": 0.32197293639183044,
      "rewards-max/forensics_tiou_reward": 0.9540406763553619,
      "rewards/forensics_sinkhorn_reward": 0.754792183637619,
      "rewards/forensics_strict_edge_reward": 0.04937927611172199,
      "rewards/forensics_tiou_reward": 0.5965515375137329,
      "step": 822
    },
    {
      "completion_length": 26.203125,
      "epoch": 2.525345622119816,
      "grad_norm": 6.214488983154297,
      "kl": 0.40625,
      "learning_rate": 3.676923076923077e-07,
      "loss": -0.0278,
      "reward": 0.8015609085559845,
      "reward_std": 0.007136810454539955,
      "reward_std_all": 0.008371041156351566,
      "rewards-max/forensics_sinkhorn_reward": 0.9702653586864471,
      "rewards-max/forensics_strict_edge_reward": 0.5479269623756409,
      "rewards-max/forensics_tiou_reward": 0.9317159056663513,
      "rewards/forensics_sinkhorn_reward": 0.7702450156211853,
      "rewards/forensics_strict_edge_reward": 0.06832255423069,
      "rewards/forensics_tiou_reward": 0.5613172650337219,
      "step": 823
    },
    {
      "completion_length": 28.671875,
      "epoch": 2.528417818740399,
      "grad_norm": 7.610814094543457,
      "kl": 0.6328125,
      "learning_rate": 3.669230769230769e-07,
      "loss": 0.019,
      "reward": 0.7588022947311401,
      "reward_std": 0.037888236343860626,
      "reward_std_all": 0.041777122765779495,
      "rewards-max/forensics_sinkhorn_reward": 0.9869123101234436,
      "rewards-max/forensics_strict_edge_reward": 0.6340298652648926,
      "rewards-max/forensics_tiou_reward": 0.9716879725456238,
      "rewards/forensics_sinkhorn_reward": 0.7100359499454498,
      "rewards/forensics_strict_edge_reward": 0.04111691005527973,
      "rewards/forensics_tiou_reward": 0.5287037491798401,
      "step": 824
    },
    {
      "completion_length": 26.328125,
      "epoch": 2.531490015360983,
      "grad_norm": 6.241874694824219,
      "kl": 0.54296875,
      "learning_rate": 3.661538461538461e-07,
      "loss": 0.0897,
      "reward": 0.750735729932785,
      "reward_std": 0.03248929884284735,
      "reward_std_all": 0.04669440537691116,
      "rewards-max/forensics_sinkhorn_reward": 0.9872514009475708,
      "rewards-max/forensics_strict_edge_reward": 0.5050012171268463,
      "rewards-max/forensics_tiou_reward": 0.9694048464298248,
      "rewards/forensics_sinkhorn_reward": 0.7388203144073486,
      "rewards/forensics_strict_edge_reward": 0.04155874904245138,
      "rewards/forensics_tiou_reward": 0.5608122646808624,
      "step": 825
    },
    {
      "completion_length": 25.296875,
      "epoch": 2.5345622119815667,
      "grad_norm": 8.086163520812988,
      "kl": 0.373046875,
      "learning_rate": 3.6538461538461534e-07,
      "loss": -0.0753,
      "reward": 0.7622698843479156,
      "reward_std": 0.0403684756020084,
      "reward_std_all": 0.042071937350556254,
      "rewards-max/forensics_sinkhorn_reward": 0.9921801090240479,
      "rewards-max/forensics_strict_edge_reward": 0.7201169431209564,
      "rewards-max/forensics_tiou_reward": 0.9858238101005554,
      "rewards/forensics_sinkhorn_reward": 0.7561697661876678,
      "rewards/forensics_strict_edge_reward": 0.04488661326467991,
      "rewards/forensics_tiou_reward": 0.5473917126655579,
      "step": 826
    },
    {
      "completion_length": 23.453125,
      "epoch": 2.5376344086021505,
      "grad_norm": 7.250401020050049,
      "kl": 0.60546875,
      "learning_rate": 3.646153846153846e-07,
      "loss": 0.0022,
      "reward": 0.7942897379398346,
      "reward_std": 0.01783185126259923,
      "reward_std_all": 0.017690502107143402,
      "rewards-max/forensics_sinkhorn_reward": 0.991058737039566,
      "rewards-max/forensics_strict_edge_reward": 0.6174952685832977,
      "rewards-max/forensics_tiou_reward": 0.9841850697994232,
      "rewards/forensics_sinkhorn_reward": 0.7877391278743744,
      "rewards/forensics_strict_edge_reward": 0.07813605293631554,
      "rewards/forensics_tiou_reward": 0.5890226364135742,
      "step": 827
    },
    {
      "completion_length": 23.84375,
      "epoch": 2.5407066052227343,
      "grad_norm": 7.225547790527344,
      "kl": 0.3349609375,
      "learning_rate": 3.638461538461538e-07,
      "loss": 0.0782,
      "reward": 0.7651346623897552,
      "reward_std": 0.05530589260160923,
      "reward_std_all": 0.05167974345386028,
      "rewards-max/forensics_sinkhorn_reward": 0.9719429016113281,
      "rewards-max/forensics_strict_edge_reward": 0.2688199281692505,
      "rewards-max/forensics_tiou_reward": 0.9303423464298248,
      "rewards/forensics_sinkhorn_reward": 0.7298985421657562,
      "rewards/forensics_strict_edge_reward": 0.027827457524836063,
      "rewards/forensics_tiou_reward": 0.4836718440055847,
      "step": 828
    },
    {
      "completion_length": 23.296875,
      "epoch": 2.543778801843318,
      "grad_norm": 6.569465160369873,
      "kl": 0.380859375,
      "learning_rate": 3.6307692307692303e-07,
      "loss": -0.031,
      "reward": 0.762230783700943,
      "reward_std": 0.025739886797964573,
      "reward_std_all": 0.03578813560307026,
      "rewards-max/forensics_sinkhorn_reward": 0.9795988202095032,
      "rewards-max/forensics_strict_edge_reward": 0.48137304186820984,
      "rewards-max/forensics_tiou_reward": 0.9544539451599121,
      "rewards/forensics_sinkhorn_reward": 0.7655376493930817,
      "rewards/forensics_strict_edge_reward": 0.05643431283533573,
      "rewards/forensics_tiou_reward": 0.5725785493850708,
      "step": 829
    },
    {
      "completion_length": 27.765625,
      "epoch": 2.546850998463902,
      "grad_norm": 7.307592868804932,
      "kl": 0.509765625,
      "learning_rate": 3.623076923076923e-07,
      "loss": -0.0153,
      "reward": 0.7850960791110992,
      "reward_std": 0.03626834345050156,
      "reward_std_all": 0.03324979776516557,
      "rewards-max/forensics_sinkhorn_reward": 0.9869597852230072,
      "rewards-max/forensics_strict_edge_reward": 0.5583210289478302,
      "rewards-max/forensics_tiou_reward": 0.9732242822647095,
      "rewards/forensics_sinkhorn_reward": 0.7317997515201569,
      "rewards/forensics_strict_edge_reward": 0.09313244745135307,
      "rewards/forensics_tiou_reward": 0.6027312278747559,
      "step": 830
    },
    {
      "completion_length": 27.671875,
      "epoch": 2.549923195084485,
      "grad_norm": 6.25775671005249,
      "kl": 0.30859375,
      "learning_rate": 3.615384615384615e-07,
      "loss": -0.0096,
      "reward": 0.734590470790863,
      "reward_std": 0.0754730086773634,
      "reward_std_all": 0.07410230673849583,
      "rewards-max/forensics_sinkhorn_reward": 0.9786783754825592,
      "rewards-max/forensics_strict_edge_reward": 0.4939252734184265,
      "rewards-max/forensics_tiou_reward": 0.9539969265460968,
      "rewards/forensics_sinkhorn_reward": 0.6646734476089478,
      "rewards/forensics_strict_edge_reward": 0.046324027702212334,
      "rewards/forensics_tiou_reward": 0.4937084913253784,
      "step": 831
    },
    {
      "completion_length": 25.265625,
      "epoch": 2.5529953917050694,
      "grad_norm": 6.580816745758057,
      "kl": 0.4189453125,
      "learning_rate": 3.607692307692308e-07,
      "loss": 0.0014,
      "reward": 0.7949571013450623,
      "reward_std": 0.019489123427774757,
      "reward_std_all": 0.01899560051970184,
      "rewards-max/forensics_sinkhorn_reward": 0.9865654110908508,
      "rewards-max/forensics_strict_edge_reward": 0.32268916070461273,
      "rewards-max/forensics_tiou_reward": 0.9692380428314209,
      "rewards/forensics_sinkhorn_reward": 0.8380211293697357,
      "rewards/forensics_strict_edge_reward": 0.036087736021727324,
      "rewards/forensics_tiou_reward": 0.666308581829071,
      "step": 832
    },
    {
      "completion_length": 25.1875,
      "epoch": 2.5560675883256527,
      "grad_norm": 6.844775676727295,
      "kl": 0.35546875,
      "learning_rate": 3.6e-07,
      "loss": -0.0735,
      "reward": 0.7809818387031555,
      "reward_std": 0.02089624386280775,
      "reward_std_all": 0.023645970970392227,
      "rewards-max/forensics_sinkhorn_reward": 0.9918777048587799,
      "rewards-max/forensics_strict_edge_reward": 0.5361751317977905,
      "rewards-max/forensics_tiou_reward": 0.9829923212528229,
      "rewards/forensics_sinkhorn_reward": 0.7587034404277802,
      "rewards/forensics_strict_edge_reward": 0.05158737301826477,
      "rewards/forensics_tiou_reward": 0.557808518409729,
      "step": 833
    },
    {
      "completion_length": 22.3125,
      "epoch": 2.5591397849462365,
      "grad_norm": 7.172451019287109,
      "kl": 0.3515625,
      "learning_rate": 3.592307692307692e-07,
      "loss": 0.0819,
      "reward": 0.7662995159626007,
      "reward_std": 0.03718791902065277,
      "reward_std_all": 0.038797298446297646,
      "rewards-max/forensics_sinkhorn_reward": 0.9885372221469879,
      "rewards-max/forensics_strict_edge_reward": 0.47831887006759644,
      "rewards-max/forensics_tiou_reward": 0.971775472164154,
      "rewards/forensics_sinkhorn_reward": 0.7600722014904022,
      "rewards/forensics_strict_edge_reward": 0.06385607644915581,
      "rewards/forensics_tiou_reward": 0.5419555604457855,
      "step": 834
    },
    {
      "completion_length": 24.953125,
      "epoch": 2.5622119815668203,
      "grad_norm": 7.892332553863525,
      "kl": 0.4716796875,
      "learning_rate": 3.584615384615384e-07,
      "loss": 0.0568,
      "reward": 0.7726082503795624,
      "reward_std": 0.02503300877287984,
      "reward_std_all": 0.03147492650896311,
      "rewards-max/forensics_sinkhorn_reward": 0.9667366147041321,
      "rewards-max/forensics_strict_edge_reward": 0.3800390809774399,
      "rewards-max/forensics_tiou_reward": 0.9197893738746643,
      "rewards/forensics_sinkhorn_reward": 0.695838063955307,
      "rewards/forensics_strict_edge_reward": 0.01869608135893941,
      "rewards/forensics_tiou_reward": 0.4667137861251831,
      "step": 835
    },
    {
      "completion_length": 25.109375,
      "epoch": 2.565284178187404,
      "grad_norm": 6.32806921005249,
      "kl": 0.3935546875,
      "learning_rate": 3.576923076923077e-07,
      "loss": 0.1546,
      "reward": 0.7175504863262177,
      "reward_std": 0.06157814525067806,
      "reward_std_all": 0.07283543422818184,
      "rewards-max/forensics_sinkhorn_reward": 0.9691857993602753,
      "rewards-max/forensics_strict_edge_reward": 0.25445423275232315,
      "rewards-max/forensics_tiou_reward": 0.9269618690013885,
      "rewards/forensics_sinkhorn_reward": 0.6244775056838989,
      "rewards/forensics_strict_edge_reward": 0.011688035912811756,
      "rewards/forensics_tiou_reward": 0.3611207604408264,
      "step": 836
    },
    {
      "completion_length": 22.171875,
      "epoch": 2.568356374807988,
      "grad_norm": 5.721482753753662,
      "kl": 0.31640625,
      "learning_rate": 3.569230769230769e-07,
      "loss": 0.166,
      "reward": 0.750569611787796,
      "reward_std": 0.04449997469782829,
      "reward_std_all": 0.051321689039468765,
      "rewards-max/forensics_sinkhorn_reward": 0.9896491169929504,
      "rewards-max/forensics_strict_edge_reward": 0.4208383075892925,
      "rewards-max/forensics_tiou_reward": 0.9796993136405945,
      "rewards/forensics_sinkhorn_reward": 0.7278697192668915,
      "rewards/forensics_strict_edge_reward": 0.060037825955078006,
      "rewards/forensics_tiou_reward": 0.5539526790380478,
      "step": 837
    },
    {
      "completion_length": 23.59375,
      "epoch": 2.571428571428571,
      "grad_norm": 8.459214210510254,
      "kl": 0.3828125,
      "learning_rate": 3.561538461538461e-07,
      "loss": -0.0614,
      "reward": 0.7840503454208374,
      "reward_std": 0.02576564345508814,
      "reward_std_all": 0.026007363572716713,
      "rewards-max/forensics_sinkhorn_reward": 0.9762679040431976,
      "rewards-max/forensics_strict_edge_reward": 0.405864953994751,
      "rewards-max/forensics_tiou_reward": 0.9404314160346985,
      "rewards/forensics_sinkhorn_reward": 0.8150739669799805,
      "rewards/forensics_strict_edge_reward": 0.04541327618062496,
      "rewards/forensics_tiou_reward": 0.6227365732192993,
      "step": 838
    },
    {
      "completion_length": 30.515625,
      "epoch": 2.5745007680491554,
      "grad_norm": 6.85586404800415,
      "kl": 0.40234375,
      "learning_rate": 3.553846153846154e-07,
      "loss": -0.0223,
      "reward": 0.7348417043685913,
      "reward_std": 0.07442381232976913,
      "reward_std_all": 0.07364431023597717,
      "rewards-max/forensics_sinkhorn_reward": 0.9777177274227142,
      "rewards-max/forensics_strict_edge_reward": 0.5398962646722794,
      "rewards-max/forensics_tiou_reward": 0.9471963047981262,
      "rewards/forensics_sinkhorn_reward": 0.6462371647357941,
      "rewards/forensics_strict_edge_reward": 0.04807007871568203,
      "rewards/forensics_tiou_reward": 0.4892764389514923,
      "step": 839
    },
    {
      "completion_length": 22.671875,
      "epoch": 2.5775729646697387,
      "grad_norm": 7.504136562347412,
      "kl": 0.41015625,
      "learning_rate": 3.546153846153846e-07,
      "loss": -0.0416,
      "reward": 0.7871611714363098,
      "reward_std": 0.021207128651440144,
      "reward_std_all": 0.021864954382181168,
      "rewards-max/forensics_sinkhorn_reward": 0.990415096282959,
      "rewards-max/forensics_strict_edge_reward": 0.6839536428451538,
      "rewards-max/forensics_tiou_reward": 0.9781396687030792,
      "rewards/forensics_sinkhorn_reward": 0.7965507209300995,
      "rewards/forensics_strict_edge_reward": 0.06246918998658657,
      "rewards/forensics_tiou_reward": 0.595302402973175,
      "step": 840
    },
    {
      "completion_length": 23.65625,
      "epoch": 2.5806451612903225,
      "grad_norm": 6.316244125366211,
      "kl": 0.5859375,
      "learning_rate": 3.5384615384615386e-07,
      "loss": 0.0269,
      "reward": 0.7936840951442719,
      "reward_std": 0.018049754202365875,
      "reward_std_all": 0.018352946965023875,
      "rewards-max/forensics_sinkhorn_reward": 0.9748079776763916,
      "rewards-max/forensics_strict_edge_reward": 0.32254111021757126,
      "rewards-max/forensics_tiou_reward": 0.9376550018787384,
      "rewards/forensics_sinkhorn_reward": 0.7543468773365021,
      "rewards/forensics_strict_edge_reward": 0.03577573876827955,
      "rewards/forensics_tiou_reward": 0.5200793445110321,
      "step": 841
    },
    {
      "completion_length": 21.90625,
      "epoch": 2.5837173579109063,
      "grad_norm": 7.382846355438232,
      "kl": 0.4384765625,
      "learning_rate": 3.530769230769231e-07,
      "loss": -0.0182,
      "reward": 0.7662633955478668,
      "reward_std": 0.056038280948996544,
      "reward_std_all": 0.05220794305205345,
      "rewards-max/forensics_sinkhorn_reward": 0.9706597328186035,
      "rewards-max/forensics_strict_edge_reward": 0.4569893330335617,
      "rewards-max/forensics_tiou_reward": 0.927458643913269,
      "rewards/forensics_sinkhorn_reward": 0.7451773285865784,
      "rewards/forensics_strict_edge_reward": 0.03602268360555172,
      "rewards/forensics_tiou_reward": 0.5485211610794067,
      "step": 842
    },
    {
      "completion_length": 27.0,
      "epoch": 2.58678955453149,
      "grad_norm": 6.341114044189453,
      "kl": 0.3798828125,
      "learning_rate": 3.523076923076923e-07,
      "loss": 0.0338,
      "reward": 0.7857369780540466,
      "reward_std": 0.02597358077764511,
      "reward_std_all": 0.02623731829226017,
      "rewards-max/forensics_sinkhorn_reward": 0.9413813054561615,
      "rewards-max/forensics_strict_edge_reward": 0.3695305436849594,
      "rewards-max/forensics_tiou_reward": 0.8692122399806976,
      "rewards/forensics_sinkhorn_reward": 0.7171582579612732,
      "rewards/forensics_strict_edge_reward": 0.024239065125584602,
      "rewards/forensics_tiou_reward": 0.5259580016136169,
      "step": 843
    },
    {
      "completion_length": 27.453125,
      "epoch": 2.589861751152074,
      "grad_norm": 6.115594387054443,
      "kl": 0.3369140625,
      "learning_rate": 3.515384615384615e-07,
      "loss": -0.0058,
      "reward": 0.7662867903709412,
      "reward_std": 0.05103098787367344,
      "reward_std_all": 0.048324509523808956,
      "rewards-max/forensics_sinkhorn_reward": 0.99039426445961,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.979633092880249,
      "rewards/forensics_sinkhorn_reward": 0.79718217253685,
      "rewards/forensics_strict_edge_reward": 0.075715446844697,
      "rewards/forensics_tiou_reward": 0.6420197486877441,
      "step": 844
    },
    {
      "completion_length": 24.671875,
      "epoch": 2.592933947772657,
      "grad_norm": 8.493576049804688,
      "kl": 0.3740234375,
      "learning_rate": 3.5076923076923077e-07,
      "loss": -0.0093,
      "reward": 0.729075014591217,
      "reward_std": 0.0515296570956707,
      "reward_std_all": 0.06438424065709114,
      "rewards-max/forensics_sinkhorn_reward": 0.9854762852191925,
      "rewards-max/forensics_strict_edge_reward": 0.6516327261924744,
      "rewards-max/forensics_tiou_reward": 0.9655067026615143,
      "rewards/forensics_sinkhorn_reward": 0.6846733391284943,
      "rewards/forensics_strict_edge_reward": 0.052289869636297226,
      "rewards/forensics_tiou_reward": 0.44734832644462585,
      "step": 845
    },
    {
      "completion_length": 27.796875,
      "epoch": 2.5960061443932414,
      "grad_norm": 6.450737476348877,
      "kl": 0.4287109375,
      "learning_rate": 3.5e-07,
      "loss": -0.0294,
      "reward": 0.7984493672847748,
      "reward_std": 0.011455360101535916,
      "reward_std_all": 0.012185930623672903,
      "rewards-max/forensics_sinkhorn_reward": 0.988378494977951,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.9758926033973694,
      "rewards/forensics_sinkhorn_reward": 0.8330399096012115,
      "rewards/forensics_strict_edge_reward": 0.05761471018195152,
      "rewards/forensics_tiou_reward": 0.665941596031189,
      "step": 846
    },
    {
      "completion_length": 25.828125,
      "epoch": 2.5990783410138247,
      "grad_norm": 10.960427284240723,
      "kl": 0.3994140625,
      "learning_rate": 3.492307692307692e-07,
      "loss": 0.007,
      "reward": 0.7942247986793518,
      "reward_std": 0.02037693839520216,
      "reward_std_all": 0.01965111494064331,
      "rewards-max/forensics_sinkhorn_reward": 0.9619282484054565,
      "rewards-max/forensics_strict_edge_reward": 0.4608110338449478,
      "rewards-max/forensics_tiou_reward": 0.9086866676807404,
      "rewards/forensics_sinkhorn_reward": 0.7767699956893921,
      "rewards/forensics_strict_edge_reward": 0.07138463295996189,
      "rewards/forensics_tiou_reward": 0.6298488080501556,
      "step": 847
    },
    {
      "completion_length": 26.578125,
      "epoch": 2.6021505376344085,
      "grad_norm": 7.688470363616943,
      "kl": 0.4072265625,
      "learning_rate": 3.4846153846153846e-07,
      "loss": 0.0614,
      "reward": 0.7660135626792908,
      "reward_std": 0.027672667056322098,
      "reward_std_all": 0.036387307569384575,
      "rewards-max/forensics_sinkhorn_reward": 0.9803090989589691,
      "rewards-max/forensics_strict_edge_reward": 0.49192555248737335,
      "rewards-max/forensics_tiou_reward": 0.9510591328144073,
      "rewards/forensics_sinkhorn_reward": 0.7374382317066193,
      "rewards/forensics_strict_edge_reward": 0.04531365446746349,
      "rewards/forensics_tiou_reward": 0.5169061124324799,
      "step": 848
    },
    {
      "completion_length": 23.734375,
      "epoch": 2.6052227342549923,
      "grad_norm": 6.9669365882873535,
      "kl": 0.439453125,
      "learning_rate": 3.476923076923077e-07,
      "loss": 0.0373,
      "reward": 0.80173459649086,
      "reward_std": 0.007561220088973641,
      "reward_std_all": 0.008490458130836487,
      "rewards-max/forensics_sinkhorn_reward": 0.9936559498310089,
      "rewards-max/forensics_strict_edge_reward": 0.32746124267578125,
      "rewards-max/forensics_tiou_reward": 0.9885217845439911,
      "rewards/forensics_sinkhorn_reward": 0.7881985604763031,
      "rewards/forensics_strict_edge_reward": 0.03579157404601574,
      "rewards/forensics_tiou_reward": 0.6457666158676147,
      "step": 849
    },
    {
      "completion_length": 20.484375,
      "epoch": 2.608294930875576,
      "grad_norm": 6.474708080291748,
      "kl": 0.4052734375,
      "learning_rate": 3.4692307692307694e-07,
      "loss": -0.0669,
      "reward": 0.8050264418125153,
      "reward_std": 0.002458340022712946,
      "reward_std_all": 0.004079216392710805,
      "rewards-max/forensics_sinkhorn_reward": 0.9909146726131439,
      "rewards-max/forensics_strict_edge_reward": 0.5702091455459595,
      "rewards-max/forensics_tiou_reward": 0.9760412573814392,
      "rewards/forensics_sinkhorn_reward": 0.8366568386554718,
      "rewards/forensics_strict_edge_reward": 0.07276105880737305,
      "rewards/forensics_tiou_reward": 0.6716743111610413,
      "step": 850
    },
    {
      "completion_length": 22.5,
      "epoch": 2.61136712749616,
      "grad_norm": 7.896765232086182,
      "kl": 0.416015625,
      "learning_rate": 3.461538461538461e-07,
      "loss": 0.0166,
      "reward": 0.790987491607666,
      "reward_std": 0.02177418931387365,
      "reward_std_all": 0.021745556965470314,
      "rewards-max/forensics_sinkhorn_reward": 0.9852606058120728,
      "rewards-max/forensics_strict_edge_reward": 0.545007586479187,
      "rewards-max/forensics_tiou_reward": 0.9678672850131989,
      "rewards/forensics_sinkhorn_reward": 0.7637763619422913,
      "rewards/forensics_strict_edge_reward": 0.04382513649761677,
      "rewards/forensics_tiou_reward": 0.5779678821563721,
      "step": 851
    },
    {
      "completion_length": 28.3125,
      "epoch": 2.6144393241167436,
      "grad_norm": 7.161405086517334,
      "kl": 0.3271484375,
      "learning_rate": 3.4538461538461537e-07,
      "loss": 0.0043,
      "reward": 0.7763296961784363,
      "reward_std": 0.03120073862373829,
      "reward_std_all": 0.03200276382267475,
      "rewards-max/forensics_sinkhorn_reward": 0.9915313124656677,
      "rewards-max/forensics_strict_edge_reward": 0.30658135563135147,
      "rewards-max/forensics_tiou_reward": 0.9800249338150024,
      "rewards/forensics_sinkhorn_reward": 0.718596488237381,
      "rewards/forensics_strict_edge_reward": 0.0226894561201334,
      "rewards/forensics_tiou_reward": 0.5385503768920898,
      "step": 852
    },
    {
      "completion_length": 25.625,
      "epoch": 2.6175115207373274,
      "grad_norm": 7.428017616271973,
      "kl": 0.3642578125,
      "learning_rate": 3.446153846153846e-07,
      "loss": 0.1083,
      "reward": 0.7646661400794983,
      "reward_std": 0.04358463454991579,
      "reward_std_all": 0.04313822370022535,
      "rewards-max/forensics_sinkhorn_reward": 0.9923006296157837,
      "rewards-max/forensics_strict_edge_reward": 0.6420974880456924,
      "rewards-max/forensics_tiou_reward": 0.9859323799610138,
      "rewards/forensics_sinkhorn_reward": 0.7402931749820709,
      "rewards/forensics_strict_edge_reward": 0.06162188947200775,
      "rewards/forensics_tiou_reward": 0.5842627286911011,
      "step": 853
    },
    {
      "completion_length": 26.21875,
      "epoch": 2.6205837173579107,
      "grad_norm": 6.7206501960754395,
      "kl": 0.52734375,
      "learning_rate": 3.4384615384615385e-07,
      "loss": -0.0144,
      "reward": 0.7770251929759979,
      "reward_std": 0.023713608272373676,
      "reward_std_all": 0.027526870369911194,
      "rewards-max/forensics_sinkhorn_reward": 0.9845795035362244,
      "rewards-max/forensics_strict_edge_reward": 0.49555233120918274,
      "rewards-max/forensics_tiou_reward": 0.9660999178886414,
      "rewards/forensics_sinkhorn_reward": 0.7998298704624176,
      "rewards/forensics_strict_edge_reward": 0.061523113399744034,
      "rewards/forensics_tiou_reward": 0.6047899723052979,
      "step": 854
    },
    {
      "completion_length": 24.09375,
      "epoch": 2.6236559139784945,
      "grad_norm": 7.405760288238525,
      "kl": 0.380859375,
      "learning_rate": 3.4307692307692307e-07,
      "loss": 0.0236,
      "reward": 0.7800958454608917,
      "reward_std": 0.016177434474229813,
      "reward_std_all": 0.021737375296652317,
      "rewards-max/forensics_sinkhorn_reward": 0.9891221523284912,
      "rewards-max/forensics_strict_edge_reward": 0.469832107424736,
      "rewards-max/forensics_tiou_reward": 0.9713389277458191,
      "rewards/forensics_sinkhorn_reward": 0.7475212514400482,
      "rewards/forensics_strict_edge_reward": 0.033335781656205654,
      "rewards/forensics_tiou_reward": 0.5490395873785019,
      "step": 855
    },
    {
      "completion_length": 25.703125,
      "epoch": 2.6267281105990783,
      "grad_norm": 11.250617980957031,
      "kl": 0.7783203125,
      "learning_rate": 3.423076923076923e-07,
      "loss": -0.0333,
      "reward": 0.79289111495018,
      "reward_std": 0.011175166349858046,
      "reward_std_all": 0.013726762495934963,
      "rewards-max/forensics_sinkhorn_reward": 0.9843449890613556,
      "rewards-max/forensics_strict_edge_reward": 0.5539559721946716,
      "rewards-max/forensics_tiou_reward": 0.9637299180030823,
      "rewards/forensics_sinkhorn_reward": 0.7682692110538483,
      "rewards/forensics_strict_edge_reward": 0.06185546703636646,
      "rewards/forensics_tiou_reward": 0.55146524310112,
      "step": 856
    },
    {
      "completion_length": 24.484375,
      "epoch": 2.629800307219662,
      "grad_norm": 6.755513668060303,
      "kl": 0.42578125,
      "learning_rate": 3.415384615384615e-07,
      "loss": 0.0457,
      "reward": 0.7763256430625916,
      "reward_std": 0.03874434903264046,
      "reward_std_all": 0.03693270683288574,
      "rewards-max/forensics_sinkhorn_reward": 0.9928907454013824,
      "rewards-max/forensics_strict_edge_reward": 0.8260191679000854,
      "rewards-max/forensics_tiou_reward": 0.988551139831543,
      "rewards/forensics_sinkhorn_reward": 0.7764897048473358,
      "rewards/forensics_strict_edge_reward": 0.10319133847951889,
      "rewards/forensics_tiou_reward": 0.584279477596283,
      "step": 857
    },
    {
      "completion_length": 25.34375,
      "epoch": 2.632872503840246,
      "grad_norm": 8.072522163391113,
      "kl": 0.630859375,
      "learning_rate": 3.4076923076923076e-07,
      "loss": 0.1319,
      "reward": 0.7228322625160217,
      "reward_std": 0.09017454832792282,
      "reward_std_all": 0.08904189616441727,
      "rewards-max/forensics_sinkhorn_reward": 0.9788363575935364,
      "rewards-max/forensics_strict_edge_reward": 0.5950735807418823,
      "rewards-max/forensics_tiou_reward": 0.9547199606895447,
      "rewards/forensics_sinkhorn_reward": 0.7171736359596252,
      "rewards/forensics_strict_edge_reward": 0.08328768238425255,
      "rewards/forensics_tiou_reward": 0.5256931185722351,
      "step": 858
    },
    {
      "completion_length": 24.546875,
      "epoch": 2.6359447004608296,
      "grad_norm": 7.134343147277832,
      "kl": 0.4287109375,
      "learning_rate": 3.4000000000000003e-07,
      "loss": -0.0163,
      "reward": 0.7728487849235535,
      "reward_std": 0.04011937603354454,
      "reward_std_all": 0.038956575095653534,
      "rewards-max/forensics_sinkhorn_reward": 0.9871099293231964,
      "rewards-max/forensics_strict_edge_reward": 0.40130549669265747,
      "rewards-max/forensics_tiou_reward": 0.9685830175876617,
      "rewards/forensics_sinkhorn_reward": 0.7813018560409546,
      "rewards/forensics_strict_edge_reward": 0.03150959499180317,
      "rewards/forensics_tiou_reward": 0.587823748588562,
      "step": 859
    },
    {
      "completion_length": 27.515625,
      "epoch": 2.6390168970814134,
      "grad_norm": 9.191202163696289,
      "kl": 0.564453125,
      "learning_rate": 3.392307692307692e-07,
      "loss": 0.0511,
      "reward": 0.749152809381485,
      "reward_std": 0.05702375003602356,
      "reward_std_all": 0.05681250337511301,
      "rewards-max/forensics_sinkhorn_reward": 0.9887593984603882,
      "rewards-max/forensics_strict_edge_reward": 0.5677289068698883,
      "rewards-max/forensics_tiou_reward": 0.9777261018753052,
      "rewards/forensics_sinkhorn_reward": 0.6914909780025482,
      "rewards/forensics_strict_edge_reward": 0.0382709763944149,
      "rewards/forensics_tiou_reward": 0.5293237268924713,
      "step": 860
    },
    {
      "completion_length": 25.4375,
      "epoch": 2.6420890937019967,
      "grad_norm": 7.160676956176758,
      "kl": 0.392578125,
      "learning_rate": 3.3846153846153845e-07,
      "loss": 0.0209,
      "reward": 0.7760152220726013,
      "reward_std": 0.04112605657428503,
      "reward_std_all": 0.03880580607801676,
      "rewards-max/forensics_sinkhorn_reward": 0.9886974990367889,
      "rewards-max/forensics_strict_edge_reward": 0.44627827405929565,
      "rewards-max/forensics_tiou_reward": 0.9759306907653809,
      "rewards/forensics_sinkhorn_reward": 0.7861670255661011,
      "rewards/forensics_strict_edge_reward": 0.07120381295681,
      "rewards/forensics_tiou_reward": 0.6394486129283905,
      "step": 861
    },
    {
      "completion_length": 24.421875,
      "epoch": 2.6451612903225805,
      "grad_norm": 11.507755279541016,
      "kl": 0.453125,
      "learning_rate": 3.3769230769230767e-07,
      "loss": 0.0768,
      "reward": 0.7241202294826508,
      "reward_std": 0.04863972030580044,
      "reward_std_all": 0.06223298795521259,
      "rewards-max/forensics_sinkhorn_reward": 0.9934123456478119,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9877482354640961,
      "rewards/forensics_sinkhorn_reward": 0.6589848697185516,
      "rewards/forensics_strict_edge_reward": 0.083401869982481,
      "rewards/forensics_tiou_reward": 0.5070415437221527,
      "step": 862
    },
    {
      "completion_length": 23.25,
      "epoch": 2.6482334869431643,
      "grad_norm": 8.150142669677734,
      "kl": 0.4306640625,
      "learning_rate": 3.3692307692307693e-07,
      "loss": 0.0265,
      "reward": 0.7904033958911896,
      "reward_std": 0.0217261528596282,
      "reward_std_all": 0.02176107047125697,
      "rewards-max/forensics_sinkhorn_reward": 0.986903578042984,
      "rewards-max/forensics_strict_edge_reward": 0.302240751683712,
      "rewards-max/forensics_tiou_reward": 0.9715959131717682,
      "rewards/forensics_sinkhorn_reward": 0.8021446168422699,
      "rewards/forensics_strict_edge_reward": 0.024226035457104445,
      "rewards/forensics_tiou_reward": 0.628168523311615,
      "step": 863
    },
    {
      "completion_length": 26.640625,
      "epoch": 2.651305683563748,
      "grad_norm": 7.149962425231934,
      "kl": 0.4794921875,
      "learning_rate": 3.361538461538461e-07,
      "loss": -0.011,
      "reward": 0.8039959073066711,
      "reward_std": 0.004214603686705232,
      "reward_std_all": 0.005585879785940051,
      "rewards-max/forensics_sinkhorn_reward": 0.9794449508190155,
      "rewards-max/forensics_strict_edge_reward": 0.5258367955684662,
      "rewards-max/forensics_tiou_reward": 0.9538272321224213,
      "rewards/forensics_sinkhorn_reward": 0.8393520414829254,
      "rewards/forensics_strict_edge_reward": 0.05345795303583145,
      "rewards/forensics_tiou_reward": 0.6757857203483582,
      "step": 864
    },
    {
      "completion_length": 22.625,
      "epoch": 2.654377880184332,
      "grad_norm": 10.207578659057617,
      "kl": 0.4091796875,
      "learning_rate": 3.3538461538461536e-07,
      "loss": -0.053,
      "reward": 0.7781343758106232,
      "reward_std": 0.03086286596953869,
      "reward_std_all": 0.03154214471578598,
      "rewards-max/forensics_sinkhorn_reward": 0.9926165044307709,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.987002819776535,
      "rewards/forensics_sinkhorn_reward": 0.7423073649406433,
      "rewards/forensics_strict_edge_reward": 0.04915943741798401,
      "rewards/forensics_tiou_reward": 0.526321291923523,
      "step": 865
    },
    {
      "completion_length": 26.0625,
      "epoch": 2.6574500768049156,
      "grad_norm": 5.971826076507568,
      "kl": 0.4521484375,
      "learning_rate": 3.346153846153846e-07,
      "loss": 0.0078,
      "reward": 0.7924434542655945,
      "reward_std": 0.017425938305677846,
      "reward_std_all": 0.018085711868479848,
      "rewards-max/forensics_sinkhorn_reward": 0.9649275541305542,
      "rewards-max/forensics_strict_edge_reward": 0.302240751683712,
      "rewards-max/forensics_tiou_reward": 0.9165400564670563,
      "rewards/forensics_sinkhorn_reward": 0.8048286736011505,
      "rewards/forensics_strict_edge_reward": 0.04013329092413187,
      "rewards/forensics_tiou_reward": 0.6373547911643982,
      "step": 866
    },
    {
      "completion_length": 25.359375,
      "epoch": 2.6605222734254994,
      "grad_norm": 7.203794002532959,
      "kl": 0.3408203125,
      "learning_rate": 3.3384615384615384e-07,
      "loss": -0.1005,
      "reward": 0.7679040431976318,
      "reward_std": 0.03561128908768296,
      "reward_std_all": 0.03711467236280441,
      "rewards-max/forensics_sinkhorn_reward": 0.9828933477401733,
      "rewards-max/forensics_strict_edge_reward": 0.7336080372333527,
      "rewards-max/forensics_tiou_reward": 0.9612695872783661,
      "rewards/forensics_sinkhorn_reward": 0.7174535095691681,
      "rewards/forensics_strict_edge_reward": 0.08136480860412121,
      "rewards/forensics_tiou_reward": 0.5177294611930847,
      "step": 867
    },
    {
      "completion_length": 26.75,
      "epoch": 2.6635944700460827,
      "grad_norm": 7.1645073890686035,
      "kl": 0.4853515625,
      "learning_rate": 3.330769230769231e-07,
      "loss": -0.0282,
      "reward": 0.7964654862880707,
      "reward_std": 0.014141349121928215,
      "reward_std_all": 0.014648659154772758,
      "rewards-max/forensics_sinkhorn_reward": 0.9826688170433044,
      "rewards-max/forensics_strict_edge_reward": 0.5003864914178848,
      "rewards-max/forensics_tiou_reward": 0.9546830952167511,
      "rewards/forensics_sinkhorn_reward": 0.7661002278327942,
      "rewards/forensics_strict_edge_reward": 0.0502562690526247,
      "rewards/forensics_tiou_reward": 0.5622111558914185,
      "step": 868
    },
    {
      "completion_length": 25.34375,
      "epoch": 2.6666666666666665,
      "grad_norm": 6.1418375968933105,
      "kl": 0.3466796875,
      "learning_rate": 3.3230769230769227e-07,
      "loss": -0.1615,
      "reward": 0.795463502407074,
      "reward_std": 0.011606496758759022,
      "reward_std_all": 0.013053934439085424,
      "rewards-max/forensics_sinkhorn_reward": 0.9909048974514008,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9819626212120056,
      "rewards/forensics_sinkhorn_reward": 0.8436160087585449,
      "rewards/forensics_strict_edge_reward": 0.08435605466365814,
      "rewards/forensics_tiou_reward": 0.6801272630691528,
      "step": 869
    },
    {
      "completion_length": 23.296875,
      "epoch": 2.6697388632872503,
      "grad_norm": 7.907392501831055,
      "kl": 0.66796875,
      "learning_rate": 3.3153846153846154e-07,
      "loss": -0.0192,
      "reward": 0.7946873903274536,
      "reward_std": 0.019857436418533325,
      "reward_std_all": 0.019297017133794725,
      "rewards-max/forensics_sinkhorn_reward": 0.9923078119754791,
      "rewards-max/forensics_strict_edge_reward": 0.6666710078716278,
      "rewards-max/forensics_tiou_reward": 0.9853260517120361,
      "rewards/forensics_sinkhorn_reward": 0.8214833438396454,
      "rewards/forensics_strict_edge_reward": 0.06297503784298897,
      "rewards/forensics_tiou_reward": 0.6171519160270691,
      "step": 870
    },
    {
      "completion_length": 28.484375,
      "epoch": 2.672811059907834,
      "grad_norm": 8.897217750549316,
      "kl": 0.4951171875,
      "learning_rate": 3.3076923076923075e-07,
      "loss": -0.0184,
      "reward": 0.7368243932723999,
      "reward_std": 0.06918036937713623,
      "reward_std_all": 0.0688573345541954,
      "rewards-max/forensics_sinkhorn_reward": 0.9742861688137054,
      "rewards-max/forensics_strict_edge_reward": 0.3471262529492378,
      "rewards-max/forensics_tiou_reward": 0.9390884339809418,
      "rewards/forensics_sinkhorn_reward": 0.6487956345081329,
      "rewards/forensics_strict_edge_reward": 0.036932699382305145,
      "rewards/forensics_tiou_reward": 0.519263967871666,
      "step": 871
    },
    {
      "completion_length": 22.125,
      "epoch": 2.675883256528418,
      "grad_norm": 6.781499862670898,
      "kl": 0.318359375,
      "learning_rate": 3.3e-07,
      "loss": -0.023,
      "reward": 0.747080385684967,
      "reward_std": 0.0182045305846259,
      "reward_std_all": 0.03849713783711195,
      "rewards-max/forensics_sinkhorn_reward": 0.9868039786815643,
      "rewards-max/forensics_strict_edge_reward": 0.6627446413040161,
      "rewards-max/forensics_tiou_reward": 0.9686426520347595,
      "rewards/forensics_sinkhorn_reward": 0.745336502790451,
      "rewards/forensics_strict_edge_reward": 0.04584223311394453,
      "rewards/forensics_tiou_reward": 0.5273503214120865,
      "step": 872
    },
    {
      "completion_length": 24.8125,
      "epoch": 2.6789554531490016,
      "grad_norm": 6.238702297210693,
      "kl": 0.3486328125,
      "learning_rate": 3.292307692307692e-07,
      "loss": -0.0231,
      "reward": 0.8004781007766724,
      "reward_std": 0.008973251504357904,
      "reward_std_all": 0.010158141842111945,
      "rewards-max/forensics_sinkhorn_reward": 0.9904341399669647,
      "rewards-max/forensics_strict_edge_reward": 0.41391946375370026,
      "rewards-max/forensics_tiou_reward": 0.9828528761863708,
      "rewards/forensics_sinkhorn_reward": 0.7856674194335938,
      "rewards/forensics_strict_edge_reward": 0.05300196632742882,
      "rewards/forensics_tiou_reward": 0.6829551160335541,
      "step": 873
    },
    {
      "completion_length": 25.6875,
      "epoch": 2.6820276497695854,
      "grad_norm": 8.625774383544922,
      "kl": 0.349609375,
      "learning_rate": 3.2846153846153844e-07,
      "loss": 0.0048,
      "reward": 0.7515120804309845,
      "reward_std": 0.0502915708348155,
      "reward_std_all": 0.05213022977113724,
      "rewards-max/forensics_sinkhorn_reward": 0.9794743955135345,
      "rewards-max/forensics_strict_edge_reward": 0.3576568216085434,
      "rewards-max/forensics_tiou_reward": 0.9534538090229034,
      "rewards/forensics_sinkhorn_reward": 0.7235515713691711,
      "rewards/forensics_strict_edge_reward": 0.02508952422067523,
      "rewards/forensics_tiou_reward": 0.6038747131824493,
      "step": 874
    },
    {
      "completion_length": 26.078125,
      "epoch": 2.6850998463901687,
      "grad_norm": 7.052289962768555,
      "kl": 0.35546875,
      "learning_rate": 3.2769230769230766e-07,
      "loss": 0.0114,
      "reward": 0.789503425359726,
      "reward_std": 0.015445558819919825,
      "reward_std_all": 0.017500512301921844,
      "rewards-max/forensics_sinkhorn_reward": 0.976707398891449,
      "rewards-max/forensics_strict_edge_reward": 0.2539179176092148,
      "rewards-max/forensics_tiou_reward": 0.9421499967575073,
      "rewards/forensics_sinkhorn_reward": 0.7485498189926147,
      "rewards/forensics_strict_edge_reward": 0.015651671215891838,
      "rewards/forensics_tiou_reward": 0.5528879165649414,
      "step": 875
    },
    {
      "completion_length": 28.109375,
      "epoch": 2.688172043010753,
      "grad_norm": 6.113202095031738,
      "kl": 0.431640625,
      "learning_rate": 3.269230769230769e-07,
      "loss": -0.1286,
      "reward": 0.7657591998577118,
      "reward_std": 0.05356023460626602,
      "reward_std_all": 0.050267551094293594,
      "rewards-max/forensics_sinkhorn_reward": 0.991198718547821,
      "rewards-max/forensics_strict_edge_reward": 0.5133360177278519,
      "rewards-max/forensics_tiou_reward": 0.9803189039230347,
      "rewards/forensics_sinkhorn_reward": 0.7545269727706909,
      "rewards/forensics_strict_edge_reward": 0.05674806609749794,
      "rewards/forensics_tiou_reward": 0.6137415766716003,
      "step": 876
    },
    {
      "completion_length": 22.359375,
      "epoch": 2.6912442396313363,
      "grad_norm": 5.9308085441589355,
      "kl": 0.4541015625,
      "learning_rate": 3.2615384615384614e-07,
      "loss": -0.0207,
      "reward": 0.7834952473640442,
      "reward_std": 0.023381050676107407,
      "reward_std_all": 0.024182844907045364,
      "rewards-max/forensics_sinkhorn_reward": 0.9893244206905365,
      "rewards-max/forensics_strict_edge_reward": 0.5382090210914612,
      "rewards-max/forensics_tiou_reward": 0.9753795564174652,
      "rewards/forensics_sinkhorn_reward": 0.8067809343338013,
      "rewards/forensics_strict_edge_reward": 0.06550371274352074,
      "rewards/forensics_tiou_reward": 0.6384442448616028,
      "step": 877
    },
    {
      "completion_length": 23.4375,
      "epoch": 2.69431643625192,
      "grad_norm": 5.981542110443115,
      "kl": 0.3642578125,
      "learning_rate": 3.2538461538461535e-07,
      "loss": 0.12,
      "reward": 0.7860228717327118,
      "reward_std": 0.028087503276765347,
      "reward_std_all": 0.027151585556566715,
      "rewards-max/forensics_sinkhorn_reward": 0.9867967367172241,
      "rewards-max/forensics_strict_edge_reward": 0.34721747040748596,
      "rewards-max/forensics_tiou_reward": 0.9707077741622925,
      "rewards/forensics_sinkhorn_reward": 0.776016503572464,
      "rewards/forensics_strict_edge_reward": 0.03929794020950794,
      "rewards/forensics_tiou_reward": 0.5964114367961884,
      "step": 878
    },
    {
      "completion_length": 25.046875,
      "epoch": 2.697388632872504,
      "grad_norm": 8.051307678222656,
      "kl": 0.6064453125,
      "learning_rate": 3.246153846153846e-07,
      "loss": 0.0368,
      "reward": 0.7904303073883057,
      "reward_std": 0.028060161508619785,
      "reward_std_all": 0.025884700939059258,
      "rewards-max/forensics_sinkhorn_reward": 0.9883333444595337,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.9751244485378265,
      "rewards/forensics_sinkhorn_reward": 0.7786221206188202,
      "rewards/forensics_strict_edge_reward": 0.07814368978142738,
      "rewards/forensics_tiou_reward": 0.6326114535331726,
      "step": 879
    },
    {
      "completion_length": 26.25,
      "epoch": 2.7004608294930876,
      "grad_norm": 8.377829551696777,
      "kl": 0.4443359375,
      "learning_rate": 3.2384615384615383e-07,
      "loss": -0.0218,
      "reward": 0.7840330898761749,
      "reward_std": 0.025131722912192345,
      "reward_std_all": 0.025712359696626663,
      "rewards-max/forensics_sinkhorn_reward": 0.9669730961322784,
      "rewards-max/forensics_strict_edge_reward": 0.32978805899620056,
      "rewards-max/forensics_tiou_reward": 0.9196880757808685,
      "rewards/forensics_sinkhorn_reward": 0.7943158149719238,
      "rewards/forensics_strict_edge_reward": 0.029915230348706245,
      "rewards/forensics_tiou_reward": 0.6048378348350525,
      "step": 880
    },
    {
      "completion_length": 23.140625,
      "epoch": 2.7035330261136714,
      "grad_norm": 6.331047058105469,
      "kl": 0.3583984375,
      "learning_rate": 3.230769230769231e-07,
      "loss": -0.0685,
      "reward": 0.7937566637992859,
      "reward_std": 0.021233326522633433,
      "reward_std_all": 0.020467570051550865,
      "rewards-max/forensics_sinkhorn_reward": 0.976506769657135,
      "rewards-max/forensics_strict_edge_reward": 0.5895819067955017,
      "rewards-max/forensics_tiou_reward": 0.9502542316913605,
      "rewards/forensics_sinkhorn_reward": 0.8112251162528992,
      "rewards/forensics_strict_edge_reward": 0.05501183308660984,
      "rewards/forensics_tiou_reward": 0.6015108227729797,
      "step": 881
    },
    {
      "completion_length": 26.984375,
      "epoch": 2.7066052227342547,
      "grad_norm": 5.453010082244873,
      "kl": 0.3466796875,
      "learning_rate": 3.2230769230769226e-07,
      "loss": 0.0512,
      "reward": 0.7395807206630707,
      "reward_std": 0.04301777482032776,
      "reward_std_all": 0.05704457871615887,
      "rewards-max/forensics_sinkhorn_reward": 0.9930550158023834,
      "rewards-max/forensics_strict_edge_reward": 0.6266469359397888,
      "rewards-max/forensics_tiou_reward": 0.9882421791553497,
      "rewards/forensics_sinkhorn_reward": 0.7768179476261139,
      "rewards/forensics_strict_edge_reward": 0.07581830769777298,
      "rewards/forensics_tiou_reward": 0.6369176805019379,
      "step": 882
    },
    {
      "completion_length": 26.109375,
      "epoch": 2.709677419354839,
      "grad_norm": 6.035955905914307,
      "kl": 0.4853515625,
      "learning_rate": 3.215384615384615e-07,
      "loss": -0.0479,
      "reward": 0.773467093706131,
      "reward_std": 0.029519627336412668,
      "reward_std_all": 0.0313763408921659,
      "rewards-max/forensics_sinkhorn_reward": 0.9867283701896667,
      "rewards-max/forensics_strict_edge_reward": 0.5538647174835205,
      "rewards-max/forensics_tiou_reward": 0.9723107814788818,
      "rewards/forensics_sinkhorn_reward": 0.7708736658096313,
      "rewards/forensics_strict_edge_reward": 0.06192343309521675,
      "rewards/forensics_tiou_reward": 0.6278471648693085,
      "step": 883
    },
    {
      "completion_length": 28.09375,
      "epoch": 2.7127496159754223,
      "grad_norm": 6.196976184844971,
      "kl": 0.3681640625,
      "learning_rate": 3.2076923076923074e-07,
      "loss": 0.0016,
      "reward": 0.7729741334915161,
      "reward_std": 0.03032353799790144,
      "reward_std_all": 0.0330170001834631,
      "rewards-max/forensics_sinkhorn_reward": 0.9777450859546661,
      "rewards-max/forensics_strict_edge_reward": 0.46008437871932983,
      "rewards-max/forensics_tiou_reward": 0.9488265812397003,
      "rewards/forensics_sinkhorn_reward": 0.7555570602416992,
      "rewards/forensics_strict_edge_reward": 0.0381249887868762,
      "rewards/forensics_tiou_reward": 0.5756028592586517,
      "step": 884
    },
    {
      "completion_length": 27.453125,
      "epoch": 2.715821812596006,
      "grad_norm": 6.558163166046143,
      "kl": 0.427734375,
      "learning_rate": 3.2e-07,
      "loss": 0.0408,
      "reward": 0.7720976769924164,
      "reward_std": 0.055674800649285316,
      "reward_std_all": 0.0508145559579134,
      "rewards-max/forensics_sinkhorn_reward": 0.9829977452754974,
      "rewards-max/forensics_strict_edge_reward": 0.2518748939037323,
      "rewards-max/forensics_tiou_reward": 0.9608168601989746,
      "rewards/forensics_sinkhorn_reward": 0.7304612398147583,
      "rewards/forensics_strict_edge_reward": 0.03522852435708046,
      "rewards/forensics_tiou_reward": 0.5684128403663635,
      "step": 885
    },
    {
      "completion_length": 25.546875,
      "epoch": 2.71889400921659,
      "grad_norm": 6.717520236968994,
      "kl": 0.375,
      "learning_rate": 3.1923076923076917e-07,
      "loss": 0.0068,
      "reward": 0.7944483160972595,
      "reward_std": 0.017478263471275568,
      "reward_std_all": 0.017611573915928602,
      "rewards-max/forensics_sinkhorn_reward": 0.9915390014648438,
      "rewards-max/forensics_strict_edge_reward": 0.7032848298549652,
      "rewards-max/forensics_tiou_reward": 0.9811055064201355,
      "rewards/forensics_sinkhorn_reward": 0.7542906105518341,
      "rewards/forensics_strict_edge_reward": 0.04802286624908447,
      "rewards/forensics_tiou_reward": 0.5921644866466522,
      "step": 886
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.7219662058371736,
      "grad_norm": 6.745694637298584,
      "kl": 0.4521484375,
      "learning_rate": 3.1846153846153843e-07,
      "loss": 0.0509,
      "reward": 0.7693423628807068,
      "reward_std": 0.038584306836128235,
      "reward_std_all": 0.03857608139514923,
      "rewards-max/forensics_sinkhorn_reward": 0.9800081253051758,
      "rewards-max/forensics_strict_edge_reward": 0.35175131633877754,
      "rewards-max/forensics_tiou_reward": 0.9518658816814423,
      "rewards/forensics_sinkhorn_reward": 0.7271981835365295,
      "rewards/forensics_strict_edge_reward": 0.033262575045228004,
      "rewards/forensics_tiou_reward": 0.49713656306266785,
      "step": 887
    },
    {
      "completion_length": 21.921875,
      "epoch": 2.7250384024577574,
      "grad_norm": 6.465214729309082,
      "kl": 0.4619140625,
      "learning_rate": 3.176923076923077e-07,
      "loss": 0.0028,
      "reward": 0.777488499879837,
      "reward_std": 0.02941431850194931,
      "reward_std_all": 0.030106527730822563,
      "rewards-max/forensics_sinkhorn_reward": 0.9769309163093567,
      "rewards-max/forensics_strict_edge_reward": 0.2725442200899124,
      "rewards-max/forensics_tiou_reward": 0.9475794434547424,
      "rewards/forensics_sinkhorn_reward": 0.7880045771598816,
      "rewards/forensics_strict_edge_reward": 0.05049903504550457,
      "rewards/forensics_tiou_reward": 0.5990827083587646,
      "step": 888
    },
    {
      "completion_length": 26.734375,
      "epoch": 2.7281105990783407,
      "grad_norm": 7.201443672180176,
      "kl": 0.3203125,
      "learning_rate": 3.169230769230769e-07,
      "loss": 0.0441,
      "reward": 0.7805119156837463,
      "reward_std": 0.0379656832665205,
      "reward_std_all": 0.035512606613337994,
      "rewards-max/forensics_sinkhorn_reward": 0.9927189946174622,
      "rewards-max/forensics_strict_edge_reward": 0.5233204662799835,
      "rewards-max/forensics_tiou_reward": 0.9849116206169128,
      "rewards/forensics_sinkhorn_reward": 0.729945719242096,
      "rewards/forensics_strict_edge_reward": 0.051653435453772545,
      "rewards/forensics_tiou_reward": 0.5727866590023041,
      "step": 889
    },
    {
      "completion_length": 25.640625,
      "epoch": 2.731182795698925,
      "grad_norm": 7.180627822875977,
      "kl": 0.3798828125,
      "learning_rate": 3.161538461538462e-07,
      "loss": 0.0503,
      "reward": 0.7805483341217041,
      "reward_std": 0.03323350753635168,
      "reward_std_all": 0.032050744630396366,
      "rewards-max/forensics_sinkhorn_reward": 0.9916606843471527,
      "rewards-max/forensics_strict_edge_reward": 0.7255018949508667,
      "rewards-max/forensics_tiou_reward": 0.982403039932251,
      "rewards/forensics_sinkhorn_reward": 0.7683763206005096,
      "rewards/forensics_strict_edge_reward": 0.10053730756044388,
      "rewards/forensics_tiou_reward": 0.5571185201406479,
      "step": 890
    },
    {
      "completion_length": 26.34375,
      "epoch": 2.7342549923195083,
      "grad_norm": 7.362051963806152,
      "kl": 0.37890625,
      "learning_rate": 3.1538461538461534e-07,
      "loss": 0.0322,
      "reward": 0.7587323486804962,
      "reward_std": 0.05020732805132866,
      "reward_std_all": 0.05038585886359215,
      "rewards-max/forensics_sinkhorn_reward": 0.9810680449008942,
      "rewards-max/forensics_strict_edge_reward": 0.47232839092612267,
      "rewards-max/forensics_tiou_reward": 0.9560579657554626,
      "rewards/forensics_sinkhorn_reward": 0.7719915211200714,
      "rewards/forensics_strict_edge_reward": 0.07466662116348743,
      "rewards/forensics_tiou_reward": 0.5944794714450836,
      "step": 891
    },
    {
      "completion_length": 24.984375,
      "epoch": 2.737327188940092,
      "grad_norm": 8.560494422912598,
      "kl": 0.4814453125,
      "learning_rate": 3.146153846153846e-07,
      "loss": 0.0272,
      "reward": 0.7463036477565765,
      "reward_std": 0.03735646978020668,
      "reward_std_all": 0.04984035901725292,
      "rewards-max/forensics_sinkhorn_reward": 0.9701461791992188,
      "rewards-max/forensics_strict_edge_reward": 0.539009615778923,
      "rewards-max/forensics_tiou_reward": 0.9289914071559906,
      "rewards/forensics_sinkhorn_reward": 0.7343985736370087,
      "rewards/forensics_strict_edge_reward": 0.07488384377211332,
      "rewards/forensics_tiou_reward": 0.5686286389827728,
      "step": 892
    },
    {
      "completion_length": 22.671875,
      "epoch": 2.740399385560676,
      "grad_norm": 6.809384346008301,
      "kl": 0.41015625,
      "learning_rate": 3.138461538461538e-07,
      "loss": 0.0351,
      "reward": 0.7771878242492676,
      "reward_std": 0.029780052602291107,
      "reward_std_all": 0.03066825494170189,
      "rewards-max/forensics_sinkhorn_reward": 0.9904544353485107,
      "rewards-max/forensics_strict_edge_reward": 0.6698661744594574,
      "rewards-max/forensics_tiou_reward": 0.9790585339069366,
      "rewards/forensics_sinkhorn_reward": 0.7655188739299774,
      "rewards/forensics_strict_edge_reward": 0.0415144395083189,
      "rewards/forensics_tiou_reward": 0.5880814492702484,
      "step": 893
    },
    {
      "completion_length": 27.375,
      "epoch": 2.7434715821812596,
      "grad_norm": 69.94015502929688,
      "kl": 0.86328125,
      "learning_rate": 3.130769230769231e-07,
      "loss": 0.0475,
      "reward": 0.7660289704799652,
      "reward_std": 0.0405180249363184,
      "reward_std_all": 0.04354542400687933,
      "rewards-max/forensics_sinkhorn_reward": 0.9752341210842133,
      "rewards-max/forensics_strict_edge_reward": 0.46885301172733307,
      "rewards-max/forensics_tiou_reward": 0.9445244371891022,
      "rewards/forensics_sinkhorn_reward": 0.7634795308113098,
      "rewards/forensics_strict_edge_reward": 0.04739512503147125,
      "rewards/forensics_tiou_reward": 0.5580366849899292,
      "step": 894
    },
    {
      "completion_length": 26.6875,
      "epoch": 2.7465437788018434,
      "grad_norm": 7.41630220413208,
      "kl": 0.7890625,
      "learning_rate": 3.1230769230769225e-07,
      "loss": 0.004,
      "reward": 0.7782987654209137,
      "reward_std": 0.023430767469108105,
      "reward_std_all": 0.029324336908757687,
      "rewards-max/forensics_sinkhorn_reward": 0.9910906851291656,
      "rewards-max/forensics_strict_edge_reward": 0.49086248874664307,
      "rewards-max/forensics_tiou_reward": 0.9815119802951813,
      "rewards/forensics_sinkhorn_reward": 0.7511171698570251,
      "rewards/forensics_strict_edge_reward": 0.06671086698770523,
      "rewards/forensics_tiou_reward": 0.5616590976715088,
      "step": 895
    },
    {
      "completion_length": 26.125,
      "epoch": 2.749615975422427,
      "grad_norm": 6.4375128746032715,
      "kl": 0.40625,
      "learning_rate": 3.115384615384615e-07,
      "loss": -0.0721,
      "reward": 0.8032996952533722,
      "reward_std": 0.005411093006841838,
      "reward_std_all": 0.006607401184737682,
      "rewards-max/forensics_sinkhorn_reward": 0.9897387623786926,
      "rewards-max/forensics_strict_edge_reward": 0.6660739481449127,
      "rewards-max/forensics_tiou_reward": 0.9789333343505859,
      "rewards/forensics_sinkhorn_reward": 0.7988943755626678,
      "rewards/forensics_strict_edge_reward": 0.0829339437186718,
      "rewards/forensics_tiou_reward": 0.6567265391349792,
      "step": 896
    },
    {
      "completion_length": 25.046875,
      "epoch": 2.752688172043011,
      "grad_norm": 8.87699031829834,
      "kl": 0.5927734375,
      "learning_rate": 3.1076923076923073e-07,
      "loss": 0.0354,
      "reward": 0.7866417765617371,
      "reward_std": 0.031260351883247495,
      "reward_std_all": 0.02928079874254763,
      "rewards-max/forensics_sinkhorn_reward": 0.9911565184593201,
      "rewards-max/forensics_strict_edge_reward": 0.6187687814235687,
      "rewards-max/forensics_tiou_reward": 0.9827758371829987,
      "rewards/forensics_sinkhorn_reward": 0.8083444833755493,
      "rewards/forensics_strict_edge_reward": 0.07947049103677273,
      "rewards/forensics_tiou_reward": 0.661795437335968,
      "step": 897
    },
    {
      "completion_length": 24.25,
      "epoch": 2.7557603686635943,
      "grad_norm": 6.854848861694336,
      "kl": 0.4296875,
      "learning_rate": 3.1e-07,
      "loss": 0.0851,
      "reward": 0.7648019194602966,
      "reward_std": 0.04568787105381489,
      "reward_std_all": 0.04480586387217045,
      "rewards-max/forensics_sinkhorn_reward": 0.9876150488853455,
      "rewards-max/forensics_strict_edge_reward": 0.5900678336620331,
      "rewards-max/forensics_tiou_reward": 0.9722054600715637,
      "rewards/forensics_sinkhorn_reward": 0.7664331793785095,
      "rewards/forensics_strict_edge_reward": 0.058132341131567955,
      "rewards/forensics_tiou_reward": 0.5870113968849182,
      "step": 898
    },
    {
      "completion_length": 25.296875,
      "epoch": 2.758832565284178,
      "grad_norm": 6.782342910766602,
      "kl": 0.470703125,
      "learning_rate": 3.0923076923076926e-07,
      "loss": 0.0667,
      "reward": 0.7498252987861633,
      "reward_std": 0.032141128554940224,
      "reward_std_all": 0.04806767217814922,
      "rewards-max/forensics_sinkhorn_reward": 0.9859011471271515,
      "rewards-max/forensics_strict_edge_reward": 0.4533289223909378,
      "rewards-max/forensics_tiou_reward": 0.9637520611286163,
      "rewards/forensics_sinkhorn_reward": 0.7083384692668915,
      "rewards/forensics_strict_edge_reward": 0.03798675537109375,
      "rewards/forensics_tiou_reward": 0.545295238494873,
      "step": 899
    },
    {
      "completion_length": 25.296875,
      "epoch": 2.761904761904762,
      "grad_norm": 7.701238632202148,
      "kl": 0.537109375,
      "learning_rate": 3.084615384615384e-07,
      "loss": 0.0307,
      "reward": 0.7785468101501465,
      "reward_std": 0.02168181072920561,
      "reward_std_all": 0.026470842771232128,
      "rewards-max/forensics_sinkhorn_reward": 0.9767727255821228,
      "rewards-max/forensics_strict_edge_reward": 0.5338451862335205,
      "rewards-max/forensics_tiou_reward": 0.9475212395191193,
      "rewards/forensics_sinkhorn_reward": 0.764971524477005,
      "rewards/forensics_strict_edge_reward": 0.05900139734148979,
      "rewards/forensics_tiou_reward": 0.5794037878513336,
      "step": 900
    },
    {
      "completion_length": 26.578125,
      "epoch": 2.7649769585253456,
      "grad_norm": 6.752664566040039,
      "kl": 0.4091796875,
      "learning_rate": 3.076923076923077e-07,
      "loss": -0.0144,
      "reward": 0.7686228156089783,
      "reward_std": 0.04747213050723076,
      "reward_std_all": 0.04541301168501377,
      "rewards-max/forensics_sinkhorn_reward": 0.9915279150009155,
      "rewards-max/forensics_strict_edge_reward": 0.7995419204235077,
      "rewards-max/forensics_tiou_reward": 0.9848530888557434,
      "rewards/forensics_sinkhorn_reward": 0.7846150100231171,
      "rewards/forensics_strict_edge_reward": 0.10547704622149467,
      "rewards/forensics_tiou_reward": 0.6075381636619568,
      "step": 901
    },
    {
      "completion_length": 25.09375,
      "epoch": 2.7680491551459294,
      "grad_norm": 7.042634963989258,
      "kl": 0.41015625,
      "learning_rate": 3.069230769230769e-07,
      "loss": -0.1096,
      "reward": 0.7687377035617828,
      "reward_std": 0.038439396768808365,
      "reward_std_all": 0.03878004662692547,
      "rewards-max/forensics_sinkhorn_reward": 0.9823583662509918,
      "rewards-max/forensics_strict_edge_reward": 0.5205644369125366,
      "rewards-max/forensics_tiou_reward": 0.9566388428211212,
      "rewards/forensics_sinkhorn_reward": 0.7593831419944763,
      "rewards/forensics_strict_edge_reward": 0.06154472753405571,
      "rewards/forensics_tiou_reward": 0.5634974837303162,
      "step": 902
    },
    {
      "completion_length": 27.921875,
      "epoch": 2.771121351766513,
      "grad_norm": 6.465124130249023,
      "kl": 0.4560546875,
      "learning_rate": 3.0615384615384617e-07,
      "loss": -0.0373,
      "reward": 0.7802339196205139,
      "reward_std": 0.04526614770293236,
      "reward_std_all": 0.040964625775814056,
      "rewards-max/forensics_sinkhorn_reward": 0.990854799747467,
      "rewards-max/forensics_strict_edge_reward": 0.5213330686092377,
      "rewards-max/forensics_tiou_reward": 0.9795635044574738,
      "rewards/forensics_sinkhorn_reward": 0.7855241298675537,
      "rewards/forensics_strict_edge_reward": 0.05885007232427597,
      "rewards/forensics_tiou_reward": 0.6519058346748352,
      "step": 903
    },
    {
      "completion_length": 29.09375,
      "epoch": 2.774193548387097,
      "grad_norm": 7.247808456420898,
      "kl": 0.4912109375,
      "learning_rate": 3.0538461538461533e-07,
      "loss": 0.1423,
      "reward": 0.7783367037773132,
      "reward_std": 0.0385558120906353,
      "reward_std_all": 0.03648202493786812,
      "rewards-max/forensics_sinkhorn_reward": 0.9864718914031982,
      "rewards-max/forensics_strict_edge_reward": 0.7172991037368774,
      "rewards-max/forensics_tiou_reward": 0.9709744453430176,
      "rewards/forensics_sinkhorn_reward": 0.7542280554771423,
      "rewards/forensics_strict_edge_reward": 0.08142232149839401,
      "rewards/forensics_tiou_reward": 0.570000171661377,
      "step": 904
    },
    {
      "completion_length": 25.390625,
      "epoch": 2.7772657450076803,
      "grad_norm": 8.040797233581543,
      "kl": 0.4013671875,
      "learning_rate": 3.046153846153846e-07,
      "loss": -0.0108,
      "reward": 0.7849804759025574,
      "reward_std": 0.011486715637147427,
      "reward_std_all": 0.018221409991383553,
      "rewards-max/forensics_sinkhorn_reward": 0.9752059578895569,
      "rewards-max/forensics_strict_edge_reward": 0.38075632601976395,
      "rewards-max/forensics_tiou_reward": 0.9373834729194641,
      "rewards/forensics_sinkhorn_reward": 0.7116073071956635,
      "rewards/forensics_strict_edge_reward": 0.029109668917953968,
      "rewards/forensics_tiou_reward": 0.5249670743942261,
      "step": 905
    },
    {
      "completion_length": 22.15625,
      "epoch": 2.780337941628264,
      "grad_norm": 10.676332473754883,
      "kl": 0.4892578125,
      "learning_rate": 3.038461538461538e-07,
      "loss": -0.0011,
      "reward": 0.8001686930656433,
      "reward_std": 0.00713448622263968,
      "reward_std_all": 0.008855321444571018,
      "rewards-max/forensics_sinkhorn_reward": 0.9854944050312042,
      "rewards-max/forensics_strict_edge_reward": 0.42036955058574677,
      "rewards-max/forensics_tiou_reward": 0.964509516954422,
      "rewards/forensics_sinkhorn_reward": 0.8096527457237244,
      "rewards/forensics_strict_edge_reward": 0.038835735991597176,
      "rewards/forensics_tiou_reward": 0.5855958461761475,
      "step": 906
    },
    {
      "completion_length": 24.25,
      "epoch": 2.783410138248848,
      "grad_norm": 8.217864990234375,
      "kl": 0.416015625,
      "learning_rate": 3.030769230769231e-07,
      "loss": 0.0707,
      "reward": 0.7900663316249847,
      "reward_std": 0.02555087930522859,
      "reward_std_all": 0.024488972267135978,
      "rewards-max/forensics_sinkhorn_reward": 0.9925036132335663,
      "rewards-max/forensics_strict_edge_reward": 0.667184591293335,
      "rewards-max/forensics_tiou_reward": 0.9851827025413513,
      "rewards/forensics_sinkhorn_reward": 0.7683935761451721,
      "rewards/forensics_strict_edge_reward": 0.08824802190065384,
      "rewards/forensics_tiou_reward": 0.6045838594436646,
      "step": 907
    },
    {
      "completion_length": 23.828125,
      "epoch": 2.7864823348694316,
      "grad_norm": 9.2918119430542,
      "kl": 0.4873046875,
      "learning_rate": 3.0230769230769235e-07,
      "loss": -0.0627,
      "reward": 0.7758753895759583,
      "reward_std": 0.03375947009772062,
      "reward_std_all": 0.03384675830602646,
      "rewards-max/forensics_sinkhorn_reward": 0.9947541058063507,
      "rewards-max/forensics_strict_edge_reward": 0.6292484402656555,
      "rewards-max/forensics_tiou_reward": 0.991685688495636,
      "rewards/forensics_sinkhorn_reward": 0.7983540296554565,
      "rewards/forensics_strict_edge_reward": 0.04631470609456301,
      "rewards/forensics_tiou_reward": 0.6210220754146576,
      "step": 908
    },
    {
      "completion_length": 28.59375,
      "epoch": 2.7895545314900154,
      "grad_norm": 6.547627925872803,
      "kl": 0.3583984375,
      "learning_rate": 3.015384615384615e-07,
      "loss": -0.0257,
      "reward": 0.7520737946033478,
      "reward_std": 0.054163575172424316,
      "reward_std_all": 0.053965309634804726,
      "rewards-max/forensics_sinkhorn_reward": 0.990045577287674,
      "rewards-max/forensics_strict_edge_reward": 0.6891378909349442,
      "rewards-max/forensics_tiou_reward": 0.9794509708881378,
      "rewards/forensics_sinkhorn_reward": 0.7714040875434875,
      "rewards/forensics_strict_edge_reward": 0.13178826868534088,
      "rewards/forensics_tiou_reward": 0.6660880446434021,
      "step": 909
    },
    {
      "completion_length": 24.90625,
      "epoch": 2.792626728110599,
      "grad_norm": 6.832962989807129,
      "kl": 0.576171875,
      "learning_rate": 3.0076923076923077e-07,
      "loss": -0.0013,
      "reward": 0.7670847773551941,
      "reward_std": 0.04840308800339699,
      "reward_std_all": 0.04626905731856823,
      "rewards-max/forensics_sinkhorn_reward": 0.9931645095348358,
      "rewards-max/forensics_strict_edge_reward": 0.41355738043785095,
      "rewards-max/forensics_tiou_reward": 0.9835247099399567,
      "rewards/forensics_sinkhorn_reward": 0.7645005285739899,
      "rewards/forensics_strict_edge_reward": 0.05676097422838211,
      "rewards/forensics_tiou_reward": 0.5856820642948151,
      "step": 910
    },
    {
      "completion_length": 26.734375,
      "epoch": 2.795698924731183,
      "grad_norm": 7.675106048583984,
      "kl": 0.435546875,
      "learning_rate": 3e-07,
      "loss": 0.1069,
      "reward": 0.775401771068573,
      "reward_std": 0.04180098790675402,
      "reward_std_all": 0.03921883460134268,
      "rewards-max/forensics_sinkhorn_reward": 0.9970852732658386,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9983541071414948,
      "rewards/forensics_sinkhorn_reward": 0.7978350222110748,
      "rewards/forensics_strict_edge_reward": 0.09011563286185265,
      "rewards/forensics_tiou_reward": 0.6266875267028809,
      "step": 911
    },
    {
      "completion_length": 25.125,
      "epoch": 2.7987711213517663,
      "grad_norm": 6.482424736022949,
      "kl": 0.4033203125,
      "learning_rate": 2.9923076923076925e-07,
      "loss": -0.0091,
      "reward": 0.7470490336418152,
      "reward_std": 0.022827632958069444,
      "reward_std_all": 0.04022679105401039,
      "rewards-max/forensics_sinkhorn_reward": 0.9858291447162628,
      "rewards-max/forensics_strict_edge_reward": 0.6500435769557953,
      "rewards-max/forensics_tiou_reward": 0.969620406627655,
      "rewards/forensics_sinkhorn_reward": 0.7446429133415222,
      "rewards/forensics_strict_edge_reward": 0.06386488303542137,
      "rewards/forensics_tiou_reward": 0.5529327690601349,
      "step": 912
    },
    {
      "completion_length": 27.40625,
      "epoch": 2.80184331797235,
      "grad_norm": 6.751163005828857,
      "kl": 0.44140625,
      "learning_rate": 2.984615384615384e-07,
      "loss": 0.0449,
      "reward": 0.7697280049324036,
      "reward_std": 0.02161207521567121,
      "reward_std_all": 0.029181926511228085,
      "rewards-max/forensics_sinkhorn_reward": 0.9851233959197998,
      "rewards-max/forensics_strict_edge_reward": 0.6941707730293274,
      "rewards-max/forensics_tiou_reward": 0.9695813655853271,
      "rewards/forensics_sinkhorn_reward": 0.7575969994068146,
      "rewards/forensics_strict_edge_reward": 0.06852269172668457,
      "rewards/forensics_tiou_reward": 0.5661586970090866,
      "step": 913
    },
    {
      "completion_length": 23.109375,
      "epoch": 2.804915514592934,
      "grad_norm": 8.705304145812988,
      "kl": 0.4453125,
      "learning_rate": 2.976923076923077e-07,
      "loss": -0.1038,
      "reward": 0.8043690919876099,
      "reward_std": 0.001562979246955365,
      "reward_std_all": 0.003965875715948641,
      "rewards-max/forensics_sinkhorn_reward": 0.9674431383609772,
      "rewards-max/forensics_strict_edge_reward": 0.5706980228424072,
      "rewards-max/forensics_tiou_reward": 0.9114024043083191,
      "rewards/forensics_sinkhorn_reward": 0.7969626784324646,
      "rewards/forensics_strict_edge_reward": 0.039875488728284836,
      "rewards/forensics_tiou_reward": 0.5754261910915375,
      "step": 914
    },
    {
      "completion_length": 23.578125,
      "epoch": 2.8079877112135176,
      "grad_norm": 7.002984523773193,
      "kl": 0.3779296875,
      "learning_rate": 2.969230769230769e-07,
      "loss": -0.0331,
      "reward": 0.7437092959880829,
      "reward_std": 0.03185788355767727,
      "reward_std_all": 0.045523831620812416,
      "rewards-max/forensics_sinkhorn_reward": 0.9861666560173035,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.971212774515152,
      "rewards/forensics_sinkhorn_reward": 0.6975494027137756,
      "rewards/forensics_strict_edge_reward": 0.05611528269946575,
      "rewards/forensics_tiou_reward": 0.5245310068130493,
      "step": 915
    },
    {
      "completion_length": 24.515625,
      "epoch": 2.8110599078341014,
      "grad_norm": 6.7724809646606445,
      "kl": 0.525390625,
      "learning_rate": 2.9615384615384616e-07,
      "loss": -0.0572,
      "reward": 0.7408053278923035,
      "reward_std": 0.04068576171994209,
      "reward_std_all": 0.053273867815732956,
      "rewards-max/forensics_sinkhorn_reward": 0.9872049987316132,
      "rewards-max/forensics_strict_edge_reward": 0.44464913103729486,
      "rewards-max/forensics_tiou_reward": 0.9721822738647461,
      "rewards/forensics_sinkhorn_reward": 0.7472817301750183,
      "rewards/forensics_strict_edge_reward": 0.07877467921935022,
      "rewards/forensics_tiou_reward": 0.5468648374080658,
      "step": 916
    },
    {
      "completion_length": 30.296875,
      "epoch": 2.814132104454685,
      "grad_norm": 6.938330173492432,
      "kl": 0.515625,
      "learning_rate": 2.953846153846154e-07,
      "loss": 0.1214,
      "reward": 0.7523263096809387,
      "reward_std": 0.05330822989344597,
      "reward_std_all": 0.05369814857840538,
      "rewards-max/forensics_sinkhorn_reward": 0.9802164435386658,
      "rewards-max/forensics_strict_edge_reward": 0.3072589263319969,
      "rewards-max/forensics_tiou_reward": 0.9515245258808136,
      "rewards/forensics_sinkhorn_reward": 0.7277724146842957,
      "rewards/forensics_strict_edge_reward": 0.043985177762806416,
      "rewards/forensics_tiou_reward": 0.5659001469612122,
      "step": 917
    },
    {
      "completion_length": 22.265625,
      "epoch": 2.817204301075269,
      "grad_norm": 7.430271625518799,
      "kl": 0.423828125,
      "learning_rate": 2.946153846153846e-07,
      "loss": -0.0003,
      "reward": 0.791936844587326,
      "reward_std": 0.024003521946724504,
      "reward_std_all": 0.02299216517712921,
      "rewards-max/forensics_sinkhorn_reward": 0.9834480285644531,
      "rewards-max/forensics_strict_edge_reward": 0.2715209499001503,
      "rewards-max/forensics_tiou_reward": 0.9670516848564148,
      "rewards/forensics_sinkhorn_reward": 0.8194043934345245,
      "rewards/forensics_strict_edge_reward": 0.03339411085471511,
      "rewards/forensics_tiou_reward": 0.6343660950660706,
      "step": 918
    },
    {
      "completion_length": 24.390625,
      "epoch": 2.8202764976958523,
      "grad_norm": 7.503511905670166,
      "kl": 0.3955078125,
      "learning_rate": 2.9384615384615385e-07,
      "loss": 0.0205,
      "reward": 0.763988196849823,
      "reward_std": 0.03817950002849102,
      "reward_std_all": 0.039772287011146545,
      "rewards-max/forensics_sinkhorn_reward": 0.9831717312335968,
      "rewards-max/forensics_strict_edge_reward": 0.43841664493083954,
      "rewards-max/forensics_tiou_reward": 0.9611066281795502,
      "rewards/forensics_sinkhorn_reward": 0.7787635922431946,
      "rewards/forensics_strict_edge_reward": 0.04170634178444743,
      "rewards/forensics_tiou_reward": 0.5764296650886536,
      "step": 919
    },
    {
      "completion_length": 25.546875,
      "epoch": 2.823348694316436,
      "grad_norm": 6.489324569702148,
      "kl": 0.4169921875,
      "learning_rate": 2.9307692307692307e-07,
      "loss": -0.0032,
      "reward": 0.7896105945110321,
      "reward_std": 0.02927778009325266,
      "reward_std_all": 0.027325598523020744,
      "rewards-max/forensics_sinkhorn_reward": 0.9875050187110901,
      "rewards-max/forensics_strict_edge_reward": 0.5866461992263794,
      "rewards-max/forensics_tiou_reward": 0.9703608751296997,
      "rewards/forensics_sinkhorn_reward": 0.8270061910152435,
      "rewards/forensics_strict_edge_reward": 0.05934719182550907,
      "rewards/forensics_tiou_reward": 0.6607172787189484,
      "step": 920
    },
    {
      "completion_length": 30.421875,
      "epoch": 2.82642089093702,
      "grad_norm": 6.006433963775635,
      "kl": 0.4736328125,
      "learning_rate": 2.9230769230769234e-07,
      "loss": -0.0894,
      "reward": 0.777879536151886,
      "reward_std": 0.01667588809505105,
      "reward_std_all": 0.022020198171958327,
      "rewards-max/forensics_sinkhorn_reward": 0.9920873045921326,
      "rewards-max/forensics_strict_edge_reward": 0.3194632828235626,
      "rewards-max/forensics_tiou_reward": 0.984673023223877,
      "rewards/forensics_sinkhorn_reward": 0.7865723669528961,
      "rewards/forensics_strict_edge_reward": 0.04620855860412121,
      "rewards/forensics_tiou_reward": 0.6385621428489685,
      "step": 921
    },
    {
      "completion_length": 24.34375,
      "epoch": 2.8294930875576036,
      "grad_norm": 7.778463840484619,
      "kl": 0.673828125,
      "learning_rate": 2.915384615384615e-07,
      "loss": 0.0399,
      "reward": 0.7745282351970673,
      "reward_std": 0.03218661155551672,
      "reward_std_all": 0.033237713389098644,
      "rewards-max/forensics_sinkhorn_reward": 0.9903733134269714,
      "rewards-max/forensics_strict_edge_reward": 0.3754010945558548,
      "rewards-max/forensics_tiou_reward": 0.9782288670539856,
      "rewards/forensics_sinkhorn_reward": 0.7671362459659576,
      "rewards/forensics_strict_edge_reward": 0.04435279592871666,
      "rewards/forensics_tiou_reward": 0.591307520866394,
      "step": 922
    },
    {
      "completion_length": 26.9375,
      "epoch": 2.8325652841781874,
      "grad_norm": 6.124637603759766,
      "kl": 0.451171875,
      "learning_rate": 2.9076923076923076e-07,
      "loss": 0.0171,
      "reward": 0.7430132031440735,
      "reward_std": 0.04827421344816685,
      "reward_std_all": 0.06025219336152077,
      "rewards-max/forensics_sinkhorn_reward": 0.9838407039642334,
      "rewards-max/forensics_strict_edge_reward": 0.5677289068698883,
      "rewards-max/forensics_tiou_reward": 0.9637381136417389,
      "rewards/forensics_sinkhorn_reward": 0.7225812077522278,
      "rewards/forensics_strict_edge_reward": 0.07328265719115734,
      "rewards/forensics_tiou_reward": 0.593485563993454,
      "step": 923
    },
    {
      "completion_length": 27.578125,
      "epoch": 2.835637480798771,
      "grad_norm": 7.254532337188721,
      "kl": 0.45703125,
      "learning_rate": 2.9e-07,
      "loss": 0.0167,
      "reward": 0.7680425047874451,
      "reward_std": 0.05053405091166496,
      "reward_std_all": 0.04775804467499256,
      "rewards-max/forensics_sinkhorn_reward": 0.9798007309436798,
      "rewards-max/forensics_strict_edge_reward": 0.3907725661993027,
      "rewards-max/forensics_tiou_reward": 0.952607274055481,
      "rewards/forensics_sinkhorn_reward": 0.7307131588459015,
      "rewards/forensics_strict_edge_reward": 0.04735777527093887,
      "rewards/forensics_tiou_reward": 0.5883434414863586,
      "step": 924
    },
    {
      "completion_length": 22.75,
      "epoch": 2.838709677419355,
      "grad_norm": 6.787035942077637,
      "kl": 0.380859375,
      "learning_rate": 2.8923076923076924e-07,
      "loss": -0.0634,
      "reward": 0.7940072417259216,
      "reward_std": 0.012017414206638932,
      "reward_std_all": 0.013662148034200072,
      "rewards-max/forensics_sinkhorn_reward": 0.9866615533828735,
      "rewards-max/forensics_strict_edge_reward": 0.5336397588253021,
      "rewards-max/forensics_tiou_reward": 0.9698834419250488,
      "rewards/forensics_sinkhorn_reward": 0.7999978065490723,
      "rewards/forensics_strict_edge_reward": 0.05948355980217457,
      "rewards/forensics_tiou_reward": 0.6041010022163391,
      "step": 925
    },
    {
      "completion_length": 25.0625,
      "epoch": 2.8417818740399383,
      "grad_norm": 9.735454559326172,
      "kl": 0.412109375,
      "learning_rate": 2.884615384615384e-07,
      "loss": -0.0344,
      "reward": 0.7827306687831879,
      "reward_std": 0.02741706185042858,
      "reward_std_all": 0.02747183758765459,
      "rewards-max/forensics_sinkhorn_reward": 0.9690017402172089,
      "rewards-max/forensics_strict_edge_reward": 0.3460000604391098,
      "rewards-max/forensics_tiou_reward": 0.9179352521896362,
      "rewards/forensics_sinkhorn_reward": 0.7688192427158356,
      "rewards/forensics_strict_edge_reward": 0.04262619465589523,
      "rewards/forensics_tiou_reward": 0.5565126240253448,
      "step": 926
    },
    {
      "completion_length": 28.734375,
      "epoch": 2.8448540706605225,
      "grad_norm": 6.743410110473633,
      "kl": 0.4453125,
      "learning_rate": 2.8769230769230767e-07,
      "loss": 0.1078,
      "reward": 0.7813642919063568,
      "reward_std": 0.019897670950740576,
      "reward_std_all": 0.023769432213157415,
      "rewards-max/forensics_sinkhorn_reward": 0.9918520748615265,
      "rewards-max/forensics_strict_edge_reward": 0.5272036492824554,
      "rewards-max/forensics_tiou_reward": 0.9845899939537048,
      "rewards/forensics_sinkhorn_reward": 0.6908762454986572,
      "rewards/forensics_strict_edge_reward": 0.06274938210844994,
      "rewards/forensics_tiou_reward": 0.5959039926528931,
      "step": 927
    },
    {
      "completion_length": 22.546875,
      "epoch": 2.847926267281106,
      "grad_norm": 6.907643795013428,
      "kl": 0.326171875,
      "learning_rate": 2.8692307692307694e-07,
      "loss": -0.0527,
      "reward": 0.7740796208381653,
      "reward_std": 0.007823750842362642,
      "reward_std_all": 0.021853120531886816,
      "rewards-max/forensics_sinkhorn_reward": 0.9697376191616058,
      "rewards-max/forensics_strict_edge_reward": 0.31592947617173195,
      "rewards-max/forensics_tiou_reward": 0.9261585474014282,
      "rewards/forensics_sinkhorn_reward": 0.7953813672065735,
      "rewards/forensics_strict_edge_reward": 0.05329345539212227,
      "rewards/forensics_tiou_reward": 0.5895416885614395,
      "step": 928
    },
    {
      "completion_length": 25.1875,
      "epoch": 2.8509984639016897,
      "grad_norm": 7.78753137588501,
      "kl": 0.3564453125,
      "learning_rate": 2.8615384615384615e-07,
      "loss": 0.0247,
      "reward": 0.7855700850486755,
      "reward_std": 0.01963481865823269,
      "reward_std_all": 0.021534686908125877,
      "rewards-max/forensics_sinkhorn_reward": 0.9774500131607056,
      "rewards-max/forensics_strict_edge_reward": 0.4686945676803589,
      "rewards-max/forensics_tiou_reward": 0.9479703307151794,
      "rewards/forensics_sinkhorn_reward": 0.7874151766300201,
      "rewards/forensics_strict_edge_reward": 0.04693545587360859,
      "rewards/forensics_tiou_reward": 0.6044517457485199,
      "step": 929
    },
    {
      "completion_length": 25.4375,
      "epoch": 2.8540706605222734,
      "grad_norm": 6.120380878448486,
      "kl": 0.3779296875,
      "learning_rate": 2.8538461538461536e-07,
      "loss": 0.1243,
      "reward": 0.7478415369987488,
      "reward_std": 0.07262203469872475,
      "reward_std_all": 0.06816888600587845,
      "rewards-max/forensics_sinkhorn_reward": 0.9955582618713379,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9939626455307007,
      "rewards/forensics_sinkhorn_reward": 0.7572857737541199,
      "rewards/forensics_strict_edge_reward": 0.08258629217743874,
      "rewards/forensics_tiou_reward": 0.6050849556922913,
      "step": 930
    },
    {
      "completion_length": 23.0,
      "epoch": 2.857142857142857,
      "grad_norm": 7.135372638702393,
      "kl": 0.4189453125,
      "learning_rate": 2.846153846153846e-07,
      "loss": -0.1021,
      "reward": 0.7986738979816437,
      "reward_std": 0.007573445443995297,
      "reward_std_all": 0.009548238245770335,
      "rewards-max/forensics_sinkhorn_reward": 0.9944595396518707,
      "rewards-max/forensics_strict_edge_reward": 0.6400359272956848,
      "rewards-max/forensics_tiou_reward": 0.9917403161525726,
      "rewards/forensics_sinkhorn_reward": 0.808476448059082,
      "rewards/forensics_strict_edge_reward": 0.0751267522573471,
      "rewards/forensics_tiou_reward": 0.5720599293708801,
      "step": 931
    },
    {
      "completion_length": 27.078125,
      "epoch": 2.860215053763441,
      "grad_norm": 7.292209625244141,
      "kl": 0.3876953125,
      "learning_rate": 2.8384615384615384e-07,
      "loss": 0.064,
      "reward": 0.7899309992790222,
      "reward_std": 0.02119907271116972,
      "reward_std_all": 0.02125740610063076,
      "rewards-max/forensics_sinkhorn_reward": 0.9492267370223999,
      "rewards-max/forensics_strict_edge_reward": 0.3161965496838093,
      "rewards-max/forensics_tiou_reward": 0.8909493982791901,
      "rewards/forensics_sinkhorn_reward": 0.709895133972168,
      "rewards/forensics_strict_edge_reward": 0.03259760420769453,
      "rewards/forensics_tiou_reward": 0.5252390801906586,
      "step": 932
    },
    {
      "completion_length": 24.21875,
      "epoch": 2.8632872503840243,
      "grad_norm": 7.784215927124023,
      "kl": 0.380859375,
      "learning_rate": 2.8307692307692306e-07,
      "loss": 0.0022,
      "reward": 0.7910719513893127,
      "reward_std": 0.02484310488216579,
      "reward_std_all": 0.023698794655501842,
      "rewards-max/forensics_sinkhorn_reward": 0.9781085252761841,
      "rewards-max/forensics_strict_edge_reward": 0.4866228401660919,
      "rewards-max/forensics_tiou_reward": 0.9513039886951447,
      "rewards/forensics_sinkhorn_reward": 0.7570351958274841,
      "rewards/forensics_strict_edge_reward": 0.052182141691446304,
      "rewards/forensics_tiou_reward": 0.5811933875083923,
      "step": 933
    },
    {
      "completion_length": 22.125,
      "epoch": 2.8663594470046085,
      "grad_norm": 7.881099700927734,
      "kl": 0.466796875,
      "learning_rate": 2.823076923076923e-07,
      "loss": -0.0507,
      "reward": 0.8033088743686676,
      "reward_std": 0.005640567513182759,
      "reward_std_all": 0.006691449787467718,
      "rewards-max/forensics_sinkhorn_reward": 0.9854763448238373,
      "rewards-max/forensics_strict_edge_reward": 0.5139672756195068,
      "rewards-max/forensics_tiou_reward": 0.9676525890827179,
      "rewards/forensics_sinkhorn_reward": 0.8496031165122986,
      "rewards/forensics_strict_edge_reward": 0.04814107343554497,
      "rewards/forensics_tiou_reward": 0.6627320051193237,
      "step": 934
    },
    {
      "completion_length": 26.609375,
      "epoch": 2.869431643625192,
      "grad_norm": 7.341990947723389,
      "kl": 0.34375,
      "learning_rate": 2.815384615384615e-07,
      "loss": -0.1019,
      "reward": 0.781179279088974,
      "reward_std": 0.020178716687951237,
      "reward_std_all": 0.023059326224029064,
      "rewards-max/forensics_sinkhorn_reward": 0.9957590997219086,
      "rewards-max/forensics_strict_edge_reward": 0.3796778470277786,
      "rewards-max/forensics_tiou_reward": 0.9925812780857086,
      "rewards/forensics_sinkhorn_reward": 0.7808010578155518,
      "rewards/forensics_strict_edge_reward": 0.026467199437320232,
      "rewards/forensics_tiou_reward": 0.5827010869979858,
      "step": 935
    },
    {
      "completion_length": 27.640625,
      "epoch": 2.8725038402457757,
      "grad_norm": 9.331137657165527,
      "kl": 0.5107421875,
      "learning_rate": 2.8076923076923075e-07,
      "loss": 0.116,
      "reward": 0.7457557022571564,
      "reward_std": 0.07412547990679741,
      "reward_std_all": 0.06965073850005865,
      "rewards-max/forensics_sinkhorn_reward": 0.9423065781593323,
      "rewards-max/forensics_strict_edge_reward": 0.3968253582715988,
      "rewards-max/forensics_tiou_reward": 0.8634766042232513,
      "rewards/forensics_sinkhorn_reward": 0.7085253596305847,
      "rewards/forensics_strict_edge_reward": 0.04400235041975975,
      "rewards/forensics_tiou_reward": 0.5326545387506485,
      "step": 936
    },
    {
      "completion_length": 25.125,
      "epoch": 2.8755760368663594,
      "grad_norm": 8.177803039550781,
      "kl": 0.6005859375,
      "learning_rate": 2.8e-07,
      "loss": -0.0491,
      "reward": 0.7665502429008484,
      "reward_std": 0.04833015613257885,
      "reward_std_all": 0.046367873437702656,
      "rewards-max/forensics_sinkhorn_reward": 0.9954009652137756,
      "rewards-max/forensics_strict_edge_reward": 0.5830906629562378,
      "rewards-max/forensics_tiou_reward": 0.9926756918430328,
      "rewards/forensics_sinkhorn_reward": 0.7850104570388794,
      "rewards/forensics_strict_edge_reward": 0.0514763742685318,
      "rewards/forensics_tiou_reward": 0.6192410290241241,
      "step": 937
    },
    {
      "completion_length": 22.28125,
      "epoch": 2.878648233486943,
      "grad_norm": 8.13086223602295,
      "kl": 0.390625,
      "learning_rate": 2.7923076923076923e-07,
      "loss": 0.1277,
      "reward": 0.7837696969509125,
      "reward_std": 0.037605900317430496,
      "reward_std_all": 0.03467123582959175,
      "rewards-max/forensics_sinkhorn_reward": 0.9628253281116486,
      "rewards-max/forensics_strict_edge_reward": 0.3444538861513138,
      "rewards-max/forensics_tiou_reward": 0.9176586866378784,
      "rewards/forensics_sinkhorn_reward": 0.7749249339103699,
      "rewards/forensics_strict_edge_reward": 0.05910301208496094,
      "rewards/forensics_tiou_reward": 0.570174902677536,
      "step": 938
    },
    {
      "completion_length": 25.359375,
      "epoch": 2.881720430107527,
      "grad_norm": 6.067758560180664,
      "kl": 0.36328125,
      "learning_rate": 2.7846153846153845e-07,
      "loss": -0.0959,
      "reward": 0.7927417755126953,
      "reward_std": 0.02300450298935175,
      "reward_std_all": 0.0220155268907547,
      "rewards-max/forensics_sinkhorn_reward": 0.9891313016414642,
      "rewards-max/forensics_strict_edge_reward": 0.5179875791072845,
      "rewards-max/forensics_tiou_reward": 0.977619081735611,
      "rewards/forensics_sinkhorn_reward": 0.8203259408473969,
      "rewards/forensics_strict_edge_reward": 0.059118978679180145,
      "rewards/forensics_tiou_reward": 0.6244910955429077,
      "step": 939
    },
    {
      "completion_length": 27.546875,
      "epoch": 2.8847926267281108,
      "grad_norm": 6.217824459075928,
      "kl": 0.3994140625,
      "learning_rate": 2.7769230769230766e-07,
      "loss": -0.0129,
      "reward": 0.7879936099052429,
      "reward_std": 0.030432160012423992,
      "reward_std_all": 0.028459020890295506,
      "rewards-max/forensics_sinkhorn_reward": 0.9834473729133606,
      "rewards-max/forensics_strict_edge_reward": 0.5500316619873047,
      "rewards-max/forensics_tiou_reward": 0.9613581597805023,
      "rewards/forensics_sinkhorn_reward": 0.7849539518356323,
      "rewards/forensics_strict_edge_reward": 0.0426163375377655,
      "rewards/forensics_tiou_reward": 0.5920946002006531,
      "step": 940
    },
    {
      "completion_length": 22.5,
      "epoch": 2.8878648233486945,
      "grad_norm": 8.686580657958984,
      "kl": 0.3955078125,
      "learning_rate": 2.7692307692307693e-07,
      "loss": -0.1221,
      "reward": 0.7705809772014618,
      "reward_std": 0.04761334415525198,
      "reward_std_all": 0.04499887488782406,
      "rewards-max/forensics_sinkhorn_reward": 0.9739186763763428,
      "rewards-max/forensics_strict_edge_reward": 0.3772052749991417,
      "rewards-max/forensics_tiou_reward": 0.9336400628089905,
      "rewards/forensics_sinkhorn_reward": 0.7559707760810852,
      "rewards/forensics_strict_edge_reward": 0.05485656438395381,
      "rewards/forensics_tiou_reward": 0.5558192580938339,
      "step": 941
    },
    {
      "completion_length": 23.75,
      "epoch": 2.890937019969278,
      "grad_norm": 9.088774681091309,
      "kl": 0.4013671875,
      "learning_rate": 2.7615384615384614e-07,
      "loss": -0.0524,
      "reward": 0.7798058390617371,
      "reward_std": 0.029822892509400845,
      "reward_std_all": 0.029829171486198902,
      "rewards-max/forensics_sinkhorn_reward": 0.9889178276062012,
      "rewards-max/forensics_strict_edge_reward": 0.5798844993114471,
      "rewards-max/forensics_tiou_reward": 0.975739449262619,
      "rewards/forensics_sinkhorn_reward": 0.8109532594680786,
      "rewards/forensics_strict_edge_reward": 0.07113531231880188,
      "rewards/forensics_tiou_reward": 0.6004447937011719,
      "step": 942
    },
    {
      "completion_length": 24.828125,
      "epoch": 2.8940092165898617,
      "grad_norm": 7.022214889526367,
      "kl": 0.4072265625,
      "learning_rate": 2.753846153846154e-07,
      "loss": -0.0223,
      "reward": 0.7857891321182251,
      "reward_std": 0.02076670713722706,
      "reward_std_all": 0.021747682243585587,
      "rewards-max/forensics_sinkhorn_reward": 0.9956993758678436,
      "rewards-max/forensics_strict_edge_reward": 0.9127408564090729,
      "rewards-max/forensics_tiou_reward": 0.9946455955505371,
      "rewards/forensics_sinkhorn_reward": 0.7990677058696747,
      "rewards/forensics_strict_edge_reward": 0.08753439411520958,
      "rewards/forensics_tiou_reward": 0.6402963697910309,
      "step": 943
    },
    {
      "completion_length": 23.46875,
      "epoch": 2.8970814132104454,
      "grad_norm": 7.078316688537598,
      "kl": 0.3984375,
      "learning_rate": 2.7461538461538457e-07,
      "loss": 0.0085,
      "reward": 0.7507763206958771,
      "reward_std": 0.011479094624519348,
      "reward_std_all": 0.03492068964987993,
      "rewards-max/forensics_sinkhorn_reward": 0.9874967038631439,
      "rewards-max/forensics_strict_edge_reward": 0.4383237212896347,
      "rewards-max/forensics_tiou_reward": 0.9703060388565063,
      "rewards/forensics_sinkhorn_reward": 0.7439064681529999,
      "rewards/forensics_strict_edge_reward": 0.06089804694056511,
      "rewards/forensics_tiou_reward": 0.5620500445365906,
      "step": 944
    },
    {
      "completion_length": 23.578125,
      "epoch": 2.900153609831029,
      "grad_norm": 12.670251846313477,
      "kl": 0.38671875,
      "learning_rate": 2.7384615384615383e-07,
      "loss": -0.056,
      "reward": 0.6974993944168091,
      "reward_std": 0.02259523910470307,
      "reward_std_all": 0.06366501376032829,
      "rewards-max/forensics_sinkhorn_reward": 0.9708169400691986,
      "rewards-max/forensics_strict_edge_reward": 0.28427761793136597,
      "rewards-max/forensics_tiou_reward": 0.931734710931778,
      "rewards/forensics_sinkhorn_reward": 0.6377937495708466,
      "rewards/forensics_strict_edge_reward": 0.02948484942317009,
      "rewards/forensics_tiou_reward": 0.40848876535892487,
      "step": 945
    },
    {
      "completion_length": 27.59375,
      "epoch": 2.903225806451613,
      "grad_norm": 6.5015716552734375,
      "kl": 0.408203125,
      "learning_rate": 2.7307692307692305e-07,
      "loss": -0.048,
      "reward": 0.7601517736911774,
      "reward_std": 0.04734482406638563,
      "reward_std_all": 0.04718559409957379,
      "rewards-max/forensics_sinkhorn_reward": 0.9863260388374329,
      "rewards-max/forensics_strict_edge_reward": 0.600974977016449,
      "rewards-max/forensics_tiou_reward": 0.9704040288925171,
      "rewards/forensics_sinkhorn_reward": 0.7866434156894684,
      "rewards/forensics_strict_edge_reward": 0.08861181326210499,
      "rewards/forensics_tiou_reward": 0.6508399546146393,
      "step": 946
    },
    {
      "completion_length": 27.640625,
      "epoch": 2.9062980030721968,
      "grad_norm": 7.8469672203063965,
      "kl": 0.533203125,
      "learning_rate": 2.723076923076923e-07,
      "loss": 0.0411,
      "reward": 0.7485096454620361,
      "reward_std": 0.05722767673432827,
      "reward_std_all": 0.057009387761354446,
      "rewards-max/forensics_sinkhorn_reward": 0.9827974140644073,
      "rewards-max/forensics_strict_edge_reward": 0.5568437725305557,
      "rewards-max/forensics_tiou_reward": 0.9569768905639648,
      "rewards/forensics_sinkhorn_reward": 0.7340617477893829,
      "rewards/forensics_strict_edge_reward": 0.07002835720777512,
      "rewards/forensics_tiou_reward": 0.5801519155502319,
      "step": 947
    },
    {
      "completion_length": 22.5625,
      "epoch": 2.9093701996927805,
      "grad_norm": 7.288337707519531,
      "kl": 0.541015625,
      "learning_rate": 2.7153846153846153e-07,
      "loss": -0.0466,
      "reward": 0.8053880929946899,
      "reward_std": 0.0020928585436195135,
      "reward_std_all": 0.0037056535948067904,
      "rewards-max/forensics_sinkhorn_reward": 0.9880036413669586,
      "rewards-max/forensics_strict_edge_reward": 0.6931992173194885,
      "rewards-max/forensics_tiou_reward": 0.9742040038108826,
      "rewards/forensics_sinkhorn_reward": 0.8502085208892822,
      "rewards/forensics_strict_edge_reward": 0.08226627111434937,
      "rewards/forensics_tiou_reward": 0.6450632810592651,
      "step": 948
    },
    {
      "completion_length": 25.984375,
      "epoch": 2.912442396313364,
      "grad_norm": 5.853859901428223,
      "kl": 0.369140625,
      "learning_rate": 2.7076923076923074e-07,
      "loss": -0.0424,
      "reward": 0.7896762490272522,
      "reward_std": 0.018619984621182084,
      "reward_std_all": 0.01945559773594141,
      "rewards-max/forensics_sinkhorn_reward": 0.9868594408035278,
      "rewards-max/forensics_strict_edge_reward": 0.7709200084209442,
      "rewards-max/forensics_tiou_reward": 0.971580296754837,
      "rewards/forensics_sinkhorn_reward": 0.8168374001979828,
      "rewards/forensics_strict_edge_reward": 0.08623242005705833,
      "rewards/forensics_tiou_reward": 0.6463737189769745,
      "step": 949
    },
    {
      "completion_length": 25.40625,
      "epoch": 2.9155145929339477,
      "grad_norm": 7.482161998748779,
      "kl": 0.41796875,
      "learning_rate": 2.7e-07,
      "loss": 0.0504,
      "reward": 0.7551541924476624,
      "reward_std": 0.06232131086289883,
      "reward_std_all": 0.059008123353123665,
      "rewards-max/forensics_sinkhorn_reward": 0.9813788533210754,
      "rewards-max/forensics_strict_edge_reward": 0.45834553241729736,
      "rewards-max/forensics_tiou_reward": 0.9572450220584869,
      "rewards/forensics_sinkhorn_reward": 0.7574819922447205,
      "rewards/forensics_strict_edge_reward": 0.06709631904959679,
      "rewards/forensics_tiou_reward": 0.565234899520874,
      "step": 950
    },
    {
      "completion_length": 28.21875,
      "epoch": 2.9185867895545314,
      "grad_norm": 5.921093463897705,
      "kl": 0.4951171875,
      "learning_rate": 2.692307692307692e-07,
      "loss": -0.0112,
      "reward": 0.7897881269454956,
      "reward_std": 0.026394356740638614,
      "reward_std_all": 0.025357982609421015,
      "rewards-max/forensics_sinkhorn_reward": 0.987270325422287,
      "rewards-max/forensics_strict_edge_reward": 0.5804109126329422,
      "rewards-max/forensics_tiou_reward": 0.9727413356304169,
      "rewards/forensics_sinkhorn_reward": 0.7242755591869354,
      "rewards/forensics_strict_edge_reward": 0.05192638747394085,
      "rewards/forensics_tiou_reward": 0.5805192291736603,
      "step": 951
    },
    {
      "completion_length": 22.109375,
      "epoch": 2.921658986175115,
      "grad_norm": 7.186092376708984,
      "kl": 0.4970703125,
      "learning_rate": 2.6846153846153844e-07,
      "loss": 0.0026,
      "reward": 0.7768523991107941,
      "reward_std": 0.016083284746855497,
      "reward_std_all": 0.02263793908059597,
      "rewards-max/forensics_sinkhorn_reward": 0.9731329083442688,
      "rewards-max/forensics_strict_edge_reward": 0.4449886232614517,
      "rewards-max/forensics_tiou_reward": 0.9382772743701935,
      "rewards/forensics_sinkhorn_reward": 0.7612816989421844,
      "rewards/forensics_strict_edge_reward": 0.03838028572499752,
      "rewards/forensics_tiou_reward": 0.5626814663410187,
      "step": 952
    },
    {
      "completion_length": 27.34375,
      "epoch": 2.924731182795699,
      "grad_norm": 6.612677574157715,
      "kl": 0.49609375,
      "learning_rate": 2.6769230769230765e-07,
      "loss": -0.0993,
      "reward": 0.8060334026813507,
      "reward_std": 0.0018150542746298015,
      "reward_std_all": 0.0033112120581790805,
      "rewards-max/forensics_sinkhorn_reward": 0.9921793639659882,
      "rewards-max/forensics_strict_edge_reward": 0.6005697548389435,
      "rewards-max/forensics_tiou_reward": 0.9844209551811218,
      "rewards/forensics_sinkhorn_reward": 0.864043802022934,
      "rewards/forensics_strict_edge_reward": 0.0970924124121666,
      "rewards/forensics_tiou_reward": 0.7179535031318665,
      "step": 953
    },
    {
      "completion_length": 24.25,
      "epoch": 2.9278033794162828,
      "grad_norm": 31.933067321777344,
      "kl": 2.244140625,
      "learning_rate": 2.669230769230769e-07,
      "loss": 0.167,
      "reward": 0.7136906087398529,
      "reward_std": 0.07978129759430885,
      "reward_std_all": 0.08789797499775887,
      "rewards-max/forensics_sinkhorn_reward": 0.9942738115787506,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9889454543590546,
      "rewards/forensics_sinkhorn_reward": 0.7094096839427948,
      "rewards/forensics_strict_edge_reward": 0.08302311971783638,
      "rewards/forensics_tiou_reward": 0.5196270197629929,
      "step": 954
    },
    {
      "completion_length": 27.03125,
      "epoch": 2.9308755760368665,
      "grad_norm": 7.3003644943237305,
      "kl": 0.392578125,
      "learning_rate": 2.6615384615384613e-07,
      "loss": 0.0494,
      "reward": 0.766807496547699,
      "reward_std": 0.021691851317882538,
      "reward_std_all": 0.032946537248790264,
      "rewards-max/forensics_sinkhorn_reward": 0.9821493923664093,
      "rewards-max/forensics_strict_edge_reward": 0.41928549110889435,
      "rewards-max/forensics_tiou_reward": 0.9569563865661621,
      "rewards/forensics_sinkhorn_reward": 0.7495186030864716,
      "rewards/forensics_strict_edge_reward": 0.03592958487570286,
      "rewards/forensics_tiou_reward": 0.5584875643253326,
      "step": 955
    },
    {
      "completion_length": 26.1875,
      "epoch": 2.93394777265745,
      "grad_norm": 6.441553592681885,
      "kl": 0.3837890625,
      "learning_rate": 2.653846153846154e-07,
      "loss": 0.0107,
      "reward": 0.7863792181015015,
      "reward_std": 0.024623381439596415,
      "reward_std_all": 0.025901693385094404,
      "rewards-max/forensics_sinkhorn_reward": 0.991363137960434,
      "rewards-max/forensics_strict_edge_reward": 0.43402300775051117,
      "rewards-max/forensics_tiou_reward": 0.9807591140270233,
      "rewards/forensics_sinkhorn_reward": 0.7876755595207214,
      "rewards/forensics_strict_edge_reward": 0.04858652874827385,
      "rewards/forensics_tiou_reward": 0.6352082788944244,
      "step": 956
    },
    {
      "completion_length": 24.328125,
      "epoch": 2.9370199692780337,
      "grad_norm": 7.38115119934082,
      "kl": 0.4560546875,
      "learning_rate": 2.646153846153846e-07,
      "loss": 0.0337,
      "reward": 0.7752572894096375,
      "reward_std": 0.011903659906238317,
      "reward_std_all": 0.023415496572852135,
      "rewards-max/forensics_sinkhorn_reward": 0.9834730625152588,
      "rewards-max/forensics_strict_edge_reward": 0.3070545494556427,
      "rewards-max/forensics_tiou_reward": 0.9637676775455475,
      "rewards/forensics_sinkhorn_reward": 0.712098240852356,
      "rewards/forensics_strict_edge_reward": 0.023076883517205715,
      "rewards/forensics_tiou_reward": 0.530754804611206,
      "step": 957
    },
    {
      "completion_length": 24.40625,
      "epoch": 2.9400921658986174,
      "grad_norm": 8.467168807983398,
      "kl": 0.4140625,
      "learning_rate": 2.638461538461538e-07,
      "loss": 0.0946,
      "reward": 0.7796480059623718,
      "reward_std": 0.033054254949092865,
      "reward_std_all": 0.03204926662147045,
      "rewards-max/forensics_sinkhorn_reward": 0.992323487997055,
      "rewards-max/forensics_strict_edge_reward": 0.41324345767498016,
      "rewards-max/forensics_tiou_reward": 0.9843902289867401,
      "rewards/forensics_sinkhorn_reward": 0.785795271396637,
      "rewards/forensics_strict_edge_reward": 0.04272174835205078,
      "rewards/forensics_tiou_reward": 0.5995562672615051,
      "step": 958
    },
    {
      "completion_length": 23.796875,
      "epoch": 2.943164362519201,
      "grad_norm": 6.719181060791016,
      "kl": 0.4296875,
      "learning_rate": 2.630769230769231e-07,
      "loss": 0.0054,
      "reward": 0.7816144824028015,
      "reward_std": 0.031712195836007595,
      "reward_std_all": 0.03078902466222644,
      "rewards-max/forensics_sinkhorn_reward": 0.9842667281627655,
      "rewards-max/forensics_strict_edge_reward": 0.5226984620094299,
      "rewards-max/forensics_tiou_reward": 0.9663922190666199,
      "rewards/forensics_sinkhorn_reward": 0.7835175395011902,
      "rewards/forensics_strict_edge_reward": 0.06520020589232445,
      "rewards/forensics_tiou_reward": 0.5714679062366486,
      "step": 959
    },
    {
      "completion_length": 27.140625,
      "epoch": 2.946236559139785,
      "grad_norm": 8.490221977233887,
      "kl": 0.37890625,
      "learning_rate": 2.623076923076923e-07,
      "loss": 0.0151,
      "reward": 0.7611661851406097,
      "reward_std": 0.004427110659889877,
      "reward_std_all": 0.026243186555802822,
      "rewards-max/forensics_sinkhorn_reward": 0.9792918860912323,
      "rewards-max/forensics_strict_edge_reward": 0.5199886560440063,
      "rewards-max/forensics_tiou_reward": 0.9428842961788177,
      "rewards/forensics_sinkhorn_reward": 0.7120723128318787,
      "rewards/forensics_strict_edge_reward": 0.040528031066060066,
      "rewards/forensics_tiou_reward": 0.5639036744832993,
      "step": 960
    },
    {
      "completion_length": 26.375,
      "epoch": 2.9493087557603688,
      "grad_norm": 6.36338996887207,
      "kl": 0.3447265625,
      "learning_rate": 2.615384615384615e-07,
      "loss": 0.0771,
      "reward": 0.7149235606193542,
      "reward_std": 0.08843020163476467,
      "reward_std_all": 0.08812452666461468,
      "rewards-max/forensics_sinkhorn_reward": 0.9782460629940033,
      "rewards-max/forensics_strict_edge_reward": 0.5713246315717697,
      "rewards-max/forensics_tiou_reward": 0.9513912498950958,
      "rewards/forensics_sinkhorn_reward": 0.6744065880775452,
      "rewards/forensics_strict_edge_reward": 0.06682351883500814,
      "rewards/forensics_tiou_reward": 0.4583921581506729,
      "step": 961
    },
    {
      "completion_length": 22.65625,
      "epoch": 2.9523809523809526,
      "grad_norm": 8.556811332702637,
      "kl": 0.4111328125,
      "learning_rate": 2.6076923076923073e-07,
      "loss": -0.0111,
      "reward": 0.7873998880386353,
      "reward_std": 0.021428808569908142,
      "reward_std_all": 0.021822846494615078,
      "rewards-max/forensics_sinkhorn_reward": 0.9926281571388245,
      "rewards-max/forensics_strict_edge_reward": 0.7047859728336334,
      "rewards-max/forensics_tiou_reward": 0.9860278367996216,
      "rewards/forensics_sinkhorn_reward": 0.7870790660381317,
      "rewards/forensics_strict_edge_reward": 0.0552319698035717,
      "rewards/forensics_tiou_reward": 0.6386372148990631,
      "step": 962
    },
    {
      "completion_length": 26.484375,
      "epoch": 2.955453149001536,
      "grad_norm": 7.227766513824463,
      "kl": 0.482421875,
      "learning_rate": 2.6e-07,
      "loss": 0.0363,
      "reward": 0.7827226519584656,
      "reward_std": 0.029045281931757927,
      "reward_std_all": 0.02954919170588255,
      "rewards-max/forensics_sinkhorn_reward": 0.9929300546646118,
      "rewards-max/forensics_strict_edge_reward": 0.32901202142238617,
      "rewards-max/forensics_tiou_reward": 0.9843437373638153,
      "rewards/forensics_sinkhorn_reward": 0.7543762624263763,
      "rewards/forensics_strict_edge_reward": 0.026136575266718864,
      "rewards/forensics_tiou_reward": 0.6581832766532898,
      "step": 963
    },
    {
      "completion_length": 27.859375,
      "epoch": 2.9585253456221197,
      "grad_norm": 6.22440767288208,
      "kl": 0.44140625,
      "learning_rate": 2.592307692307692e-07,
      "loss": -0.0083,
      "reward": 0.8039693832397461,
      "reward_std": 0.0040746001759544015,
      "reward_std_all": 0.005449066404253244,
      "rewards-max/forensics_sinkhorn_reward": 0.9767923355102539,
      "rewards-max/forensics_strict_edge_reward": 0.5371377021074295,
      "rewards-max/forensics_tiou_reward": 0.944997102022171,
      "rewards/forensics_sinkhorn_reward": 0.8277722299098969,
      "rewards/forensics_strict_edge_reward": 0.07228684611618519,
      "rewards/forensics_tiou_reward": 0.643252432346344,
      "step": 964
    },
    {
      "completion_length": 25.84375,
      "epoch": 2.9615975422427034,
      "grad_norm": 5.87302303314209,
      "kl": 0.2880859375,
      "learning_rate": 2.584615384615385e-07,
      "loss": -0.0931,
      "reward": 0.8047447800636292,
      "reward_std": 0.0032411408610641956,
      "reward_std_all": 0.004574760794639587,
      "rewards-max/forensics_sinkhorn_reward": 0.9937418401241302,
      "rewards-max/forensics_strict_edge_reward": 0.6516326665878296,
      "rewards-max/forensics_tiou_reward": 0.9889072477817535,
      "rewards/forensics_sinkhorn_reward": 0.8583214581012726,
      "rewards/forensics_strict_edge_reward": 0.11296408623456955,
      "rewards/forensics_tiou_reward": 0.7010172307491302,
      "step": 965
    },
    {
      "completion_length": 27.234375,
      "epoch": 2.964669738863287,
      "grad_norm": 7.214536190032959,
      "kl": 0.361328125,
      "learning_rate": 2.5769230769230764e-07,
      "loss": -0.0272,
      "reward": 0.7747893333435059,
      "reward_std": 0.03236547764390707,
      "reward_std_all": 0.033728089183568954,
      "rewards-max/forensics_sinkhorn_reward": 0.9856366813182831,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.970515787601471,
      "rewards/forensics_sinkhorn_reward": 0.7578179240226746,
      "rewards/forensics_strict_edge_reward": 0.04066935367882252,
      "rewards/forensics_tiou_reward": 0.5809662342071533,
      "step": 966
    },
    {
      "completion_length": 25.359375,
      "epoch": 2.967741935483871,
      "grad_norm": 6.55038595199585,
      "kl": 0.376953125,
      "learning_rate": 2.569230769230769e-07,
      "loss": 0.1348,
      "reward": 0.7571809589862823,
      "reward_std": 0.059105414897203445,
      "reward_std_all": 0.05643381178379059,
      "rewards-max/forensics_sinkhorn_reward": 0.99549800157547,
      "rewards-max/forensics_strict_edge_reward": 0.5913417637348175,
      "rewards-max/forensics_tiou_reward": 0.9947952926158905,
      "rewards/forensics_sinkhorn_reward": 0.7356389760971069,
      "rewards/forensics_strict_edge_reward": 0.032532365061342716,
      "rewards/forensics_tiou_reward": 0.5419754385948181,
      "step": 967
    },
    {
      "completion_length": 22.484375,
      "epoch": 2.9708141321044548,
      "grad_norm": 9.058650016784668,
      "kl": 0.3662109375,
      "learning_rate": 2.5615384615384617e-07,
      "loss": -0.0102,
      "reward": 0.7812623977661133,
      "reward_std": 0.03333643078804016,
      "reward_std_all": 0.03217343892902136,
      "rewards-max/forensics_sinkhorn_reward": 0.981688380241394,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9620749354362488,
      "rewards/forensics_sinkhorn_reward": 0.8155682682991028,
      "rewards/forensics_strict_edge_reward": 0.050751449540257454,
      "rewards/forensics_tiou_reward": 0.6056650876998901,
      "step": 968
    },
    {
      "completion_length": 26.953125,
      "epoch": 2.9738863287250386,
      "grad_norm": 6.032605171203613,
      "kl": 0.3896484375,
      "learning_rate": 2.553846153846154e-07,
      "loss": -0.0723,
      "reward": 0.79927197098732,
      "reward_std": 0.011207892093807459,
      "reward_std_all": 0.011938017792999744,
      "rewards-max/forensics_sinkhorn_reward": 0.979898601770401,
      "rewards-max/forensics_strict_edge_reward": 0.5015397369861603,
      "rewards-max/forensics_tiou_reward": 0.949306458234787,
      "rewards/forensics_sinkhorn_reward": 0.778743177652359,
      "rewards/forensics_strict_edge_reward": 0.05089171603322029,
      "rewards/forensics_tiou_reward": 0.6073808073997498,
      "step": 969
    },
    {
      "completion_length": 27.796875,
      "epoch": 2.976958525345622,
      "grad_norm": 8.431300163269043,
      "kl": 0.322265625,
      "learning_rate": 2.546153846153846e-07,
      "loss": 0.031,
      "reward": 0.7831595540046692,
      "reward_std": 0.022994506522081792,
      "reward_std_all": 0.024036089656874537,
      "rewards-max/forensics_sinkhorn_reward": 0.9914383292198181,
      "rewards-max/forensics_strict_edge_reward": 0.5511984527111053,
      "rewards-max/forensics_tiou_reward": 0.9819362759590149,
      "rewards/forensics_sinkhorn_reward": 0.7261736690998077,
      "rewards/forensics_strict_edge_reward": 0.06339588761329651,
      "rewards/forensics_tiou_reward": 0.599785715341568,
      "step": 970
    },
    {
      "completion_length": 26.578125,
      "epoch": 2.980030721966206,
      "grad_norm": 11.090087890625,
      "kl": 0.38671875,
      "learning_rate": 2.538461538461538e-07,
      "loss": 0.014,
      "reward": 0.7661500871181488,
      "reward_std": 0.012041724286973476,
      "reward_std_all": 0.027060219552367926,
      "rewards-max/forensics_sinkhorn_reward": 0.9881397783756256,
      "rewards-max/forensics_strict_edge_reward": 0.4632900059223175,
      "rewards-max/forensics_tiou_reward": 0.9767414629459381,
      "rewards/forensics_sinkhorn_reward": 0.7665494978427887,
      "rewards/forensics_strict_edge_reward": 0.06264027580618858,
      "rewards/forensics_tiou_reward": 0.5821575224399567,
      "step": 971
    },
    {
      "completion_length": 26.703125,
      "epoch": 2.9831029185867894,
      "grad_norm": 6.792198181152344,
      "kl": 0.4599609375,
      "learning_rate": 2.530769230769231e-07,
      "loss": 0.0417,
      "reward": 0.7870624959468842,
      "reward_std": 0.03300556092290208,
      "reward_std_all": 0.030502481386065483,
      "rewards-max/forensics_sinkhorn_reward": 0.9891430139541626,
      "rewards-max/forensics_strict_edge_reward": 0.8465203046798706,
      "rewards-max/forensics_tiou_reward": 0.9805768430233002,
      "rewards/forensics_sinkhorn_reward": 0.7942580878734589,
      "rewards/forensics_strict_edge_reward": 0.09986711293458939,
      "rewards/forensics_tiou_reward": 0.5933761298656464,
      "step": 972
    },
    {
      "completion_length": 24.578125,
      "epoch": 2.986175115207373,
      "grad_norm": 6.231260776519775,
      "kl": 0.4775390625,
      "learning_rate": 2.523076923076923e-07,
      "loss": 0.0672,
      "reward": 0.8031495213508606,
      "reward_std": 0.004301848472096026,
      "reward_std_all": 0.006126105319708586,
      "rewards-max/forensics_sinkhorn_reward": 0.9896897971630096,
      "rewards-max/forensics_strict_edge_reward": 0.5550253242254257,
      "rewards-max/forensics_tiou_reward": 0.9787240624427795,
      "rewards/forensics_sinkhorn_reward": 0.7878343462944031,
      "rewards/forensics_strict_edge_reward": 0.04367356467992067,
      "rewards/forensics_tiou_reward": 0.6120674014091492,
      "step": 973
    },
    {
      "completion_length": 21.375,
      "epoch": 2.989247311827957,
      "grad_norm": 7.2941508293151855,
      "kl": 0.35546875,
      "learning_rate": 2.515384615384615e-07,
      "loss": 0.0715,
      "reward": 0.7446863651275635,
      "reward_std": 0.05737078748643398,
      "reward_std_all": 0.06102371588349342,
      "rewards-max/forensics_sinkhorn_reward": 0.9763858318328857,
      "rewards-max/forensics_strict_edge_reward": 0.5484400391578674,
      "rewards-max/forensics_tiou_reward": 0.9488621354103088,
      "rewards/forensics_sinkhorn_reward": 0.7418661117553711,
      "rewards/forensics_strict_edge_reward": 0.04379190504550934,
      "rewards/forensics_tiou_reward": 0.554434597492218,
      "step": 974
    },
    {
      "completion_length": 24.640625,
      "epoch": 2.9923195084485408,
      "grad_norm": 6.956399917602539,
      "kl": 0.599609375,
      "learning_rate": 2.507692307692307e-07,
      "loss": 0.0546,
      "reward": 0.7706112861633301,
      "reward_std": 0.04014470428228378,
      "reward_std_all": 0.039259012788534164,
      "rewards-max/forensics_sinkhorn_reward": 0.9934130609035492,
      "rewards-max/forensics_strict_edge_reward": 0.48031625151634216,
      "rewards-max/forensics_tiou_reward": 0.987735241651535,
      "rewards/forensics_sinkhorn_reward": 0.804968535900116,
      "rewards/forensics_strict_edge_reward": 0.04459136538207531,
      "rewards/forensics_tiou_reward": 0.6130028069019318,
      "step": 975
    },
    {
      "completion_length": 25.828125,
      "epoch": 2.9953917050691246,
      "grad_norm": 7.122873306274414,
      "kl": 0.4296875,
      "learning_rate": 2.5e-07,
      "loss": -0.0271,
      "reward": 0.7990553081035614,
      "reward_std": 0.01222051156219095,
      "reward_std_all": 0.01286911393981427,
      "rewards-max/forensics_sinkhorn_reward": 0.9847772419452667,
      "rewards-max/forensics_strict_edge_reward": 0.48667749762535095,
      "rewards-max/forensics_tiou_reward": 0.9630177021026611,
      "rewards/forensics_sinkhorn_reward": 0.8091874122619629,
      "rewards/forensics_strict_edge_reward": 0.07528720423579216,
      "rewards/forensics_tiou_reward": 0.6652892231941223,
      "step": 976
    },
    {
      "completion_length": 25.265625,
      "epoch": 2.998463901689708,
      "grad_norm": 8.194101333618164,
      "kl": 0.384765625,
      "learning_rate": 2.492307692307692e-07,
      "loss": 0.0417,
      "reward": 0.7596315741539001,
      "reward_std": 0.05945532023906708,
      "reward_std_all": 0.05607696995139122,
      "rewards-max/forensics_sinkhorn_reward": 0.9782455563545227,
      "rewards-max/forensics_strict_edge_reward": 0.35571807622909546,
      "rewards-max/forensics_tiou_reward": 0.9459786415100098,
      "rewards/forensics_sinkhorn_reward": 0.7732966244220734,
      "rewards/forensics_strict_edge_reward": 0.049285768531262875,
      "rewards/forensics_tiou_reward": 0.5922656953334808,
      "step": 977
    },
    {
      "completion_length": 32.60000228881836,
      "epoch": 3.0,
      "grad_norm": 4.407153606414795,
      "kl": 0.46484375,
      "learning_rate": 2.4846153846153847e-07,
      "loss": 0.0352,
      "reward": 0.7672138810157776,
      "reward_std": 0.08442181348800659,
      "reward_std_all": 0.04711931571364403,
      "rewards-max/forensics_sinkhorn_reward": 0.9969732165336609,
      "rewards-max/forensics_strict_edge_reward": 0.740818202495575,
      "rewards-max/forensics_tiou_reward": 0.9983600974082947,
      "rewards/forensics_sinkhorn_reward": 0.7806196212768555,
      "rewards/forensics_strict_edge_reward": 0.14834266901016235,
      "rewards/forensics_tiou_reward": 0.6681763529777527,
      "step": 978
    },
    {
      "completion_length": 23.734375,
      "epoch": 3.003072196620584,
      "grad_norm": 7.39955997467041,
      "kl": 0.37890625,
      "learning_rate": 2.476923076923077e-07,
      "loss": -0.0534,
      "reward": 0.7850421369075775,
      "reward_std": 0.037368061020970345,
      "reward_std_all": 0.03419461473822594,
      "rewards-max/forensics_sinkhorn_reward": 0.9927559792995453,
      "rewards-max/forensics_strict_edge_reward": 0.564802497625351,
      "rewards-max/forensics_tiou_reward": 0.9859780669212341,
      "rewards/forensics_sinkhorn_reward": 0.8300704956054688,
      "rewards/forensics_strict_edge_reward": 0.07464264892041683,
      "rewards/forensics_tiou_reward": 0.6568388342857361,
      "step": 979
    },
    {
      "completion_length": 26.296875,
      "epoch": 3.0061443932411676,
      "grad_norm": 6.998391151428223,
      "kl": 0.435546875,
      "learning_rate": 2.469230769230769e-07,
      "loss": 0.0545,
      "reward": 0.7641227543354034,
      "reward_std": 0.05488484725356102,
      "reward_std_all": 0.05162941850721836,
      "rewards-max/forensics_sinkhorn_reward": 0.9765014052391052,
      "rewards-max/forensics_strict_edge_reward": 0.4985126703977585,
      "rewards-max/forensics_tiou_reward": 0.9474528133869171,
      "rewards/forensics_sinkhorn_reward": 0.7420503497123718,
      "rewards/forensics_strict_edge_reward": 0.056808119639754295,
      "rewards/forensics_tiou_reward": 0.5842968821525574,
      "step": 980
    },
    {
      "completion_length": 23.640625,
      "epoch": 3.0092165898617513,
      "grad_norm": 7.861020088195801,
      "kl": 0.5126953125,
      "learning_rate": 2.4615384615384616e-07,
      "loss": 0.0715,
      "reward": 0.7714274227619171,
      "reward_std": 0.03082568384706974,
      "reward_std_all": 0.03351936675608158,
      "rewards-max/forensics_sinkhorn_reward": 0.9905693233013153,
      "rewards-max/forensics_strict_edge_reward": 0.6976466774940491,
      "rewards-max/forensics_tiou_reward": 0.9784221947193146,
      "rewards/forensics_sinkhorn_reward": 0.7701433300971985,
      "rewards/forensics_strict_edge_reward": 0.06369226984679699,
      "rewards/forensics_tiou_reward": 0.5680620670318604,
      "step": 981
    },
    {
      "completion_length": 24.609375,
      "epoch": 3.0122887864823347,
      "grad_norm": 6.206663131713867,
      "kl": 0.4375,
      "learning_rate": 2.453846153846154e-07,
      "loss": -0.0587,
      "reward": 0.7907458245754242,
      "reward_std": 0.01980907516553998,
      "reward_std_all": 0.020196879282593727,
      "rewards-max/forensics_sinkhorn_reward": 0.9904648959636688,
      "rewards-max/forensics_strict_edge_reward": 0.5178692862391472,
      "rewards-max/forensics_tiou_reward": 0.9796192944049835,
      "rewards/forensics_sinkhorn_reward": 0.8659650683403015,
      "rewards/forensics_strict_edge_reward": 0.06531413830816746,
      "rewards/forensics_tiou_reward": 0.7185940146446228,
      "step": 982
    },
    {
      "completion_length": 23.546875,
      "epoch": 3.0153609831029184,
      "grad_norm": 8.50478744506836,
      "kl": 0.6962890625,
      "learning_rate": 2.446153846153846e-07,
      "loss": 0.0041,
      "reward": 0.8041102290153503,
      "reward_std": 0.0037507518427446485,
      "reward_std_all": 0.005293330061249435,
      "rewards-max/forensics_sinkhorn_reward": 0.9802899360656738,
      "rewards-max/forensics_strict_edge_reward": 0.4009152799844742,
      "rewards-max/forensics_tiou_reward": 0.9528968334197998,
      "rewards/forensics_sinkhorn_reward": 0.8204632997512817,
      "rewards/forensics_strict_edge_reward": 0.06210843846201897,
      "rewards/forensics_tiou_reward": 0.6403218805789948,
      "step": 983
    },
    {
      "completion_length": 22.890625,
      "epoch": 3.0184331797235022,
      "grad_norm": 8.52751636505127,
      "kl": 0.43359375,
      "learning_rate": 2.4384615384615386e-07,
      "loss": 0.0328,
      "reward": 0.776395320892334,
      "reward_std": 0.038150787353515625,
      "reward_std_all": 0.03692058566957712,
      "rewards-max/forensics_sinkhorn_reward": 0.9943661391735077,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9907462000846863,
      "rewards/forensics_sinkhorn_reward": 0.7529343962669373,
      "rewards/forensics_strict_edge_reward": 0.05728246457874775,
      "rewards/forensics_tiou_reward": 0.540978878736496,
      "step": 984
    },
    {
      "completion_length": 23.796875,
      "epoch": 3.021505376344086,
      "grad_norm": 8.581345558166504,
      "kl": 0.583984375,
      "learning_rate": 2.4307692307692307e-07,
      "loss": 0.0292,
      "reward": 0.7833983600139618,
      "reward_std": 0.020813061390072107,
      "reward_std_all": 0.02273112814873457,
      "rewards-max/forensics_sinkhorn_reward": 0.9876187741756439,
      "rewards-max/forensics_strict_edge_reward": 0.2731337174773216,
      "rewards-max/forensics_tiou_reward": 0.9620910286903381,
      "rewards/forensics_sinkhorn_reward": 0.7740613520145416,
      "rewards/forensics_strict_edge_reward": 0.019888165406882763,
      "rewards/forensics_tiou_reward": 0.5926266312599182,
      "step": 985
    },
    {
      "completion_length": 22.140625,
      "epoch": 3.02457757296467,
      "grad_norm": 6.86759614944458,
      "kl": 0.5078125,
      "learning_rate": 2.423076923076923e-07,
      "loss": -0.0966,
      "reward": 0.7960003018379211,
      "reward_std": 0.012277014902792871,
      "reward_std_all": 0.013180938316509128,
      "rewards-max/forensics_sinkhorn_reward": 0.9967097043991089,
      "rewards-max/forensics_strict_edge_reward": 0.8326062560081482,
      "rewards-max/forensics_tiou_reward": 0.9977988600730896,
      "rewards/forensics_sinkhorn_reward": 0.8342880606651306,
      "rewards/forensics_strict_edge_reward": 0.09538416191935539,
      "rewards/forensics_tiou_reward": 0.6682669520378113,
      "step": 986
    },
    {
      "completion_length": 24.515625,
      "epoch": 3.0276497695852536,
      "grad_norm": 7.662747383117676,
      "kl": 0.3642578125,
      "learning_rate": 2.415384615384615e-07,
      "loss": -0.0376,
      "reward": 0.8043858706951141,
      "reward_std": 0.0021742467069998384,
      "reward_std_all": 0.0042430926114320755,
      "rewards-max/forensics_sinkhorn_reward": 0.9835702776908875,
      "rewards-max/forensics_strict_edge_reward": 0.49686121940612793,
      "rewards-max/forensics_tiou_reward": 0.9587024748325348,
      "rewards/forensics_sinkhorn_reward": 0.803022176027298,
      "rewards/forensics_strict_edge_reward": 0.0736781470477581,
      "rewards/forensics_tiou_reward": 0.6037413775920868,
      "step": 987
    },
    {
      "completion_length": 23.140625,
      "epoch": 3.0307219662058373,
      "grad_norm": 6.812250137329102,
      "kl": 0.4423828125,
      "learning_rate": 2.4076923076923076e-07,
      "loss": -0.0345,
      "reward": 0.780838817358017,
      "reward_std": 0.03392329951748252,
      "reward_std_all": 0.032757511362433434,
      "rewards-max/forensics_sinkhorn_reward": 0.9949354827404022,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9938214719295502,
      "rewards/forensics_sinkhorn_reward": 0.797753095626831,
      "rewards/forensics_strict_edge_reward": 0.05704636126756668,
      "rewards/forensics_tiou_reward": 0.5844936072826385,
      "step": 988
    },
    {
      "completion_length": 26.921875,
      "epoch": 3.033794162826421,
      "grad_norm": 7.0921430587768555,
      "kl": 0.4052734375,
      "learning_rate": 2.4e-07,
      "loss": -0.0246,
      "reward": 0.7426756024360657,
      "reward_std": 0.07292237505316734,
      "reward_std_all": 0.07029929012060165,
      "rewards-max/forensics_sinkhorn_reward": 0.9769590497016907,
      "rewards-max/forensics_strict_edge_reward": 0.39209163188934326,
      "rewards-max/forensics_tiou_reward": 0.9473684728145599,
      "rewards/forensics_sinkhorn_reward": 0.7212226092815399,
      "rewards/forensics_strict_edge_reward": 0.04319887142628431,
      "rewards/forensics_tiou_reward": 0.5199029743671417,
      "step": 989
    },
    {
      "completion_length": 24.5625,
      "epoch": 3.0368663594470044,
      "grad_norm": 7.102407455444336,
      "kl": 0.443359375,
      "learning_rate": 2.392307692307692e-07,
      "loss": -0.0276,
      "reward": 0.7875242829322815,
      "reward_std": 0.001418077270500362,
      "reward_std_all": 0.012323561124503613,
      "rewards-max/forensics_sinkhorn_reward": 0.9851997494697571,
      "rewards-max/forensics_strict_edge_reward": 0.4883905351161957,
      "rewards-max/forensics_tiou_reward": 0.9599222540855408,
      "rewards/forensics_sinkhorn_reward": 0.776181548833847,
      "rewards/forensics_strict_edge_reward": 0.046539388597011566,
      "rewards/forensics_tiou_reward": 0.5633261501789093,
      "step": 990
    },
    {
      "completion_length": 24.75,
      "epoch": 3.0399385560675882,
      "grad_norm": 6.502134323120117,
      "kl": 0.3828125,
      "learning_rate": 2.3846153846153846e-07,
      "loss": -0.1254,
      "reward": 0.802092045545578,
      "reward_std": 0.008477915194816887,
      "reward_std_all": 0.009191694902256131,
      "rewards-max/forensics_sinkhorn_reward": 0.9872977137565613,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.9746153950691223,
      "rewards/forensics_sinkhorn_reward": 0.8786568343639374,
      "rewards/forensics_strict_edge_reward": 0.0719757117331028,
      "rewards/forensics_tiou_reward": 0.7374977469444275,
      "step": 991
    },
    {
      "completion_length": 26.78125,
      "epoch": 3.043010752688172,
      "grad_norm": 6.740666389465332,
      "kl": 0.4580078125,
      "learning_rate": 2.376923076923077e-07,
      "loss": -0.046,
      "reward": 0.801226019859314,
      "reward_std": 0.009722813265398145,
      "reward_std_all": 0.010140598053112626,
      "rewards-max/forensics_sinkhorn_reward": 0.9896663427352905,
      "rewards-max/forensics_strict_edge_reward": 0.7509788870811462,
      "rewards-max/forensics_tiou_reward": 0.9791575372219086,
      "rewards/forensics_sinkhorn_reward": 0.8051379323005676,
      "rewards/forensics_strict_edge_reward": 0.11989790201187134,
      "rewards/forensics_tiou_reward": 0.6912591457366943,
      "step": 992
    },
    {
      "completion_length": 26.59375,
      "epoch": 3.046082949308756,
      "grad_norm": 9.379669189453125,
      "kl": 0.4951171875,
      "learning_rate": 2.369230769230769e-07,
      "loss": -0.094,
      "reward": 0.7805063724517822,
      "reward_std": 0.02686687815003097,
      "reward_std_all": 0.027752208057790995,
      "rewards-max/forensics_sinkhorn_reward": 0.9838632643222809,
      "rewards-max/forensics_strict_edge_reward": 0.5136472433805466,
      "rewards-max/forensics_tiou_reward": 0.9648869037628174,
      "rewards/forensics_sinkhorn_reward": 0.7925541996955872,
      "rewards/forensics_strict_edge_reward": 0.06116946414113045,
      "rewards/forensics_tiou_reward": 0.6399107575416565,
      "step": 993
    },
    {
      "completion_length": 30.453125,
      "epoch": 3.0491551459293396,
      "grad_norm": 13.273797035217285,
      "kl": 0.583984375,
      "learning_rate": 2.3615384615384615e-07,
      "loss": 0.0711,
      "reward": 0.7575673758983612,
      "reward_std": 0.04511916730552912,
      "reward_std_all": 0.04697325639426708,
      "rewards-max/forensics_sinkhorn_reward": 0.9834852516651154,
      "rewards-max/forensics_strict_edge_reward": 0.32440951466560364,
      "rewards-max/forensics_tiou_reward": 0.9607594013214111,
      "rewards/forensics_sinkhorn_reward": 0.7187840342521667,
      "rewards/forensics_strict_edge_reward": 0.03582273609936237,
      "rewards/forensics_tiou_reward": 0.5549026131629944,
      "step": 994
    },
    {
      "completion_length": 23.484375,
      "epoch": 3.0522273425499233,
      "grad_norm": 8.589167594909668,
      "kl": 0.3642578125,
      "learning_rate": 2.3538461538461537e-07,
      "loss": -0.0565,
      "reward": 0.8013500571250916,
      "reward_std": 0.007202654494903982,
      "reward_std_all": 0.00874234875664115,
      "rewards-max/forensics_sinkhorn_reward": 0.9897874593734741,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.982524186372757,
      "rewards/forensics_sinkhorn_reward": 0.8021144866943359,
      "rewards/forensics_strict_edge_reward": 0.07462568953633308,
      "rewards/forensics_tiou_reward": 0.5895989537239075,
      "step": 995
    },
    {
      "completion_length": 23.515625,
      "epoch": 3.055299539170507,
      "grad_norm": 11.360891342163086,
      "kl": 0.4697265625,
      "learning_rate": 2.346153846153846e-07,
      "loss": 0.0147,
      "reward": 0.7889289259910583,
      "reward_std": 0.029947284841910005,
      "reward_std_all": 0.027908925665542483,
      "rewards-max/forensics_sinkhorn_reward": 0.9841114580631256,
      "rewards-max/forensics_strict_edge_reward": 0.7201169431209564,
      "rewards-max/forensics_tiou_reward": 0.9665648639202118,
      "rewards/forensics_sinkhorn_reward": 0.7890427112579346,
      "rewards/forensics_strict_edge_reward": 0.08936566486954689,
      "rewards/forensics_tiou_reward": 0.6032330989837646,
      "step": 996
    },
    {
      "completion_length": 23.765625,
      "epoch": 3.0583717357910905,
      "grad_norm": 7.3560919761657715,
      "kl": 0.3505859375,
      "learning_rate": 2.3384615384615382e-07,
      "loss": -0.0774,
      "reward": 0.7818110585212708,
      "reward_std": 0.02538723824545741,
      "reward_std_all": 0.026307882741093636,
      "rewards-max/forensics_sinkhorn_reward": 0.9871231019496918,
      "rewards-max/forensics_strict_edge_reward": 0.4268728941679001,
      "rewards-max/forensics_tiou_reward": 0.9707720577716827,
      "rewards/forensics_sinkhorn_reward": 0.7978023886680603,
      "rewards/forensics_strict_edge_reward": 0.05118349380791187,
      "rewards/forensics_tiou_reward": 0.6488046646118164,
      "step": 997
    },
    {
      "completion_length": 25.6875,
      "epoch": 3.0614439324116742,
      "grad_norm": 7.998345375061035,
      "kl": 0.376953125,
      "learning_rate": 2.3307692307692306e-07,
      "loss": 0.0994,
      "reward": 0.7849741280078888,
      "reward_std": 0.03687314875423908,
      "reward_std_all": 0.033808356150984764,
      "rewards-max/forensics_sinkhorn_reward": 0.9940810203552246,
      "rewards-max/forensics_strict_edge_reward": 0.6132497191429138,
      "rewards-max/forensics_tiou_reward": 0.9877854287624359,
      "rewards/forensics_sinkhorn_reward": 0.8013573288917542,
      "rewards/forensics_strict_edge_reward": 0.0665629543364048,
      "rewards/forensics_tiou_reward": 0.6194557249546051,
      "step": 998
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.064516129032258,
      "grad_norm": 9.392083168029785,
      "kl": 0.404296875,
      "learning_rate": 2.323076923076923e-07,
      "loss": -0.0392,
      "reward": 0.7952618300914764,
      "reward_std": 0.010670838004443794,
      "reward_std_all": 0.012326801544986665,
      "rewards-max/forensics_sinkhorn_reward": 0.9913001656532288,
      "rewards-max/forensics_strict_edge_reward": 0.6065306663513184,
      "rewards-max/forensics_tiou_reward": 0.9816692173480988,
      "rewards/forensics_sinkhorn_reward": 0.8192336857318878,
      "rewards/forensics_strict_edge_reward": 0.07318644225597382,
      "rewards/forensics_tiou_reward": 0.6976216435432434,
      "step": 999
    },
    {
      "completion_length": 25.453125,
      "epoch": 3.067588325652842,
      "grad_norm": 8.067549705505371,
      "kl": 0.423828125,
      "learning_rate": 2.3153846153846151e-07,
      "loss": -0.0785,
      "reward": 0.8009352684020996,
      "reward_std": 0.006283690046984702,
      "reward_std_all": 0.007794672856107354,
      "rewards-max/forensics_sinkhorn_reward": 0.9862496554851532,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9731225073337555,
      "rewards/forensics_sinkhorn_reward": 0.8374508619308472,
      "rewards/forensics_strict_edge_reward": 0.08537640795111656,
      "rewards/forensics_tiou_reward": 0.6889925301074982,
      "step": 1000
    },
    {
      "completion_length": 23.28125,
      "epoch": 3.0706605222734256,
      "grad_norm": 10.367080688476562,
      "kl": 0.5087890625,
      "learning_rate": 2.3076923076923078e-07,
      "loss": 0.0389,
      "reward": 0.7873983681201935,
      "reward_std": 0.021625862224027514,
      "reward_std_all": 0.021941534709185362,
      "rewards-max/forensics_sinkhorn_reward": 0.977985292673111,
      "rewards-max/forensics_strict_edge_reward": 0.45694398880004883,
      "rewards-max/forensics_tiou_reward": 0.9487713873386383,
      "rewards/forensics_sinkhorn_reward": 0.7711063325405121,
      "rewards/forensics_strict_edge_reward": 0.06693639606237411,
      "rewards/forensics_tiou_reward": 0.6300934851169586,
      "step": 1001
    },
    {
      "completion_length": 23.328125,
      "epoch": 3.0737327188940093,
      "grad_norm": 9.27846622467041,
      "kl": 0.3583984375,
      "learning_rate": 2.3e-07,
      "loss": -0.0887,
      "reward": 0.7896547615528107,
      "reward_std": 0.01431556005263701,
      "reward_std_all": 0.016811115434393287,
      "rewards-max/forensics_sinkhorn_reward": 0.989848792552948,
      "rewards-max/forensics_strict_edge_reward": 0.6005697548389435,
      "rewards-max/forensics_tiou_reward": 0.9759318232536316,
      "rewards/forensics_sinkhorn_reward": 0.7819881439208984,
      "rewards/forensics_strict_edge_reward": 0.06440357491374016,
      "rewards/forensics_tiou_reward": 0.5296167731285095,
      "step": 1002
    },
    {
      "completion_length": 25.078125,
      "epoch": 3.076804915514593,
      "grad_norm": 7.954104423522949,
      "kl": 0.3779296875,
      "learning_rate": 2.2923076923076924e-07,
      "loss": -0.0379,
      "reward": 0.8027710616588593,
      "reward_std": 0.0047733765677548945,
      "reward_std_all": 0.006350227864459157,
      "rewards-max/forensics_sinkhorn_reward": 0.9854883551597595,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9694820046424866,
      "rewards/forensics_sinkhorn_reward": 0.8395209908485413,
      "rewards/forensics_strict_edge_reward": 0.06201516464352608,
      "rewards/forensics_tiou_reward": 0.702109307050705,
      "step": 1003
    },
    {
      "completion_length": 26.0,
      "epoch": 3.0798771121351765,
      "grad_norm": 7.481289863586426,
      "kl": 0.5546875,
      "learning_rate": 2.2846153846153845e-07,
      "loss": -0.0677,
      "reward": 0.800006240606308,
      "reward_std": 0.009743118891492486,
      "reward_std_all": 0.010785710532218218,
      "rewards-max/forensics_sinkhorn_reward": 0.9820170998573303,
      "rewards-max/forensics_strict_edge_reward": 0.6146068125963211,
      "rewards-max/forensics_tiou_reward": 0.9631807208061218,
      "rewards/forensics_sinkhorn_reward": 0.7403691411018372,
      "rewards/forensics_strict_edge_reward": 0.08283961843699217,
      "rewards/forensics_tiou_reward": 0.5886313021183014,
      "step": 1004
    },
    {
      "completion_length": 24.3125,
      "epoch": 3.0829493087557602,
      "grad_norm": 7.752710819244385,
      "kl": 0.4150390625,
      "learning_rate": 2.276923076923077e-07,
      "loss": 0.0399,
      "reward": 0.7904161810874939,
      "reward_std": 0.023829021491110325,
      "reward_std_all": 0.023111945018172264,
      "rewards-max/forensics_sinkhorn_reward": 0.9928362965583801,
      "rewards-max/forensics_strict_edge_reward": 0.7309097647666931,
      "rewards-max/forensics_tiou_reward": 0.9875947535037994,
      "rewards/forensics_sinkhorn_reward": 0.7580822706222534,
      "rewards/forensics_strict_edge_reward": 0.0948774442076683,
      "rewards/forensics_tiou_reward": 0.6663148999214172,
      "step": 1005
    },
    {
      "completion_length": 23.828125,
      "epoch": 3.086021505376344,
      "grad_norm": 7.158941268920898,
      "kl": 0.427734375,
      "learning_rate": 2.269230769230769e-07,
      "loss": -0.0561,
      "reward": 0.8010926842689514,
      "reward_std": 0.009251065261196345,
      "reward_std_all": 0.010054531507194042,
      "rewards-max/forensics_sinkhorn_reward": 0.9872783422470093,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9677729308605194,
      "rewards/forensics_sinkhorn_reward": 0.836719423532486,
      "rewards/forensics_strict_edge_reward": 0.09240188077092171,
      "rewards/forensics_tiou_reward": 0.658466100692749,
      "step": 1006
    },
    {
      "completion_length": 28.796875,
      "epoch": 3.089093701996928,
      "grad_norm": 7.122160911560059,
      "kl": 0.4541015625,
      "learning_rate": 2.2615384615384614e-07,
      "loss": 0.0639,
      "reward": 0.7908376753330231,
      "reward_std": 0.027346539311110973,
      "reward_std_all": 0.02550286753103137,
      "rewards-max/forensics_sinkhorn_reward": 0.9930651187896729,
      "rewards-max/forensics_strict_edge_reward": 0.5447898209095001,
      "rewards-max/forensics_tiou_reward": 0.9853857457637787,
      "rewards/forensics_sinkhorn_reward": 0.8317540884017944,
      "rewards/forensics_strict_edge_reward": 0.07247474789619446,
      "rewards/forensics_tiou_reward": 0.6588665544986725,
      "step": 1007
    },
    {
      "completion_length": 24.890625,
      "epoch": 3.0921658986175116,
      "grad_norm": 8.563222885131836,
      "kl": 0.484375,
      "learning_rate": 2.2538461538461536e-07,
      "loss": 0.0546,
      "reward": 0.7087897062301636,
      "reward_std": 0.055515456944704056,
      "reward_std_all": 0.08013959601521492,
      "rewards-max/forensics_sinkhorn_reward": 0.9897396862506866,
      "rewards-max/forensics_strict_edge_reward": 0.625676691532135,
      "rewards-max/forensics_tiou_reward": 0.9798473715782166,
      "rewards/forensics_sinkhorn_reward": 0.7161904275417328,
      "rewards/forensics_strict_edge_reward": 0.08523857221007347,
      "rewards/forensics_tiou_reward": 0.6130083799362183,
      "step": 1008
    },
    {
      "completion_length": 23.6875,
      "epoch": 3.0952380952380953,
      "grad_norm": 8.152883529663086,
      "kl": 0.4599609375,
      "learning_rate": 2.246153846153846e-07,
      "loss": -0.024,
      "reward": 0.7805878818035126,
      "reward_std": 0.034985859878361225,
      "reward_std_all": 0.0334361270070076,
      "rewards-max/forensics_sinkhorn_reward": 0.976416289806366,
      "rewards-max/forensics_strict_edge_reward": 0.5080415308475494,
      "rewards-max/forensics_tiou_reward": 0.9334879517555237,
      "rewards/forensics_sinkhorn_reward": 0.7856386005878448,
      "rewards/forensics_strict_edge_reward": 0.06911065708845854,
      "rewards/forensics_tiou_reward": 0.617170125246048,
      "step": 1009
    },
    {
      "completion_length": 25.265625,
      "epoch": 3.098310291858679,
      "grad_norm": 7.160344123840332,
      "kl": 0.552734375,
      "learning_rate": 2.238461538461538e-07,
      "loss": -0.0304,
      "reward": 0.7962864339351654,
      "reward_std": 0.01735442056087777,
      "reward_std_all": 0.01714493054896593,
      "rewards-max/forensics_sinkhorn_reward": 0.9894949793815613,
      "rewards-max/forensics_strict_edge_reward": 0.5321264863014221,
      "rewards-max/forensics_tiou_reward": 0.9743541479110718,
      "rewards/forensics_sinkhorn_reward": 0.8281263411045074,
      "rewards/forensics_strict_edge_reward": 0.06742098927497864,
      "rewards/forensics_tiou_reward": 0.6621781885623932,
      "step": 1010
    },
    {
      "completion_length": 27.046875,
      "epoch": 3.1013824884792625,
      "grad_norm": 8.841757774353027,
      "kl": 0.6279296875,
      "learning_rate": 2.2307692307692308e-07,
      "loss": 0.0851,
      "reward": 0.7237746715545654,
      "reward_std": 0.05231238342821598,
      "reward_std_all": 0.0686332918703556,
      "rewards-max/forensics_sinkhorn_reward": 0.9766026735305786,
      "rewards-max/forensics_strict_edge_reward": 0.5402242243289948,
      "rewards-max/forensics_tiou_reward": 0.9391515254974365,
      "rewards/forensics_sinkhorn_reward": 0.7299756109714508,
      "rewards/forensics_strict_edge_reward": 0.0818423330783844,
      "rewards/forensics_tiou_reward": 0.59229975938797,
      "step": 1011
    },
    {
      "completion_length": 22.234375,
      "epoch": 3.1044546850998462,
      "grad_norm": 7.497313022613525,
      "kl": 0.4990234375,
      "learning_rate": 2.2230769230769232e-07,
      "loss": -0.0641,
      "reward": 0.7942339777946472,
      "reward_std": 0.021610441617667675,
      "reward_std_all": 0.020473730750381947,
      "rewards-max/forensics_sinkhorn_reward": 0.9896738231182098,
      "rewards-max/forensics_strict_edge_reward": 0.7101751863956451,
      "rewards-max/forensics_tiou_reward": 0.9783061742782593,
      "rewards/forensics_sinkhorn_reward": 0.8489674031734467,
      "rewards/forensics_strict_edge_reward": 0.09693951159715652,
      "rewards/forensics_tiou_reward": 0.6597901284694672,
      "step": 1012
    },
    {
      "completion_length": 25.3125,
      "epoch": 3.10752688172043,
      "grad_norm": 7.510137557983398,
      "kl": 0.48046875,
      "learning_rate": 2.2153846153846153e-07,
      "loss": -0.0881,
      "reward": 0.7631991803646088,
      "reward_std": 0.003649411373771727,
      "reward_std_all": 0.02476428635418415,
      "rewards-max/forensics_sinkhorn_reward": 0.9922003746032715,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.984605997800827,
      "rewards/forensics_sinkhorn_reward": 0.7833572924137115,
      "rewards/forensics_strict_edge_reward": 0.07269073650240898,
      "rewards/forensics_tiou_reward": 0.6108794212341309,
      "step": 1013
    },
    {
      "completion_length": 27.234375,
      "epoch": 3.110599078341014,
      "grad_norm": 8.838859558105469,
      "kl": 0.435546875,
      "learning_rate": 2.2076923076923077e-07,
      "loss": 0.0321,
      "reward": 0.7974775433540344,
      "reward_std": 0.012663547415286303,
      "reward_std_all": 0.013163530733436346,
      "rewards-max/forensics_sinkhorn_reward": 0.9897509217262268,
      "rewards-max/forensics_strict_edge_reward": 0.5885107517242432,
      "rewards-max/forensics_tiou_reward": 0.9796008765697479,
      "rewards/forensics_sinkhorn_reward": 0.8206277787685394,
      "rewards/forensics_strict_edge_reward": 0.08230933919548988,
      "rewards/forensics_tiou_reward": 0.6743416786193848,
      "step": 1014
    },
    {
      "completion_length": 25.828125,
      "epoch": 3.1136712749615976,
      "grad_norm": 7.039318084716797,
      "kl": 0.498046875,
      "learning_rate": 2.1999999999999998e-07,
      "loss": -0.0705,
      "reward": 0.7968482077121735,
      "reward_std": 0.018244216218590736,
      "reward_std_all": 0.017331491224467754,
      "rewards-max/forensics_sinkhorn_reward": 0.9953494668006897,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9943570494651794,
      "rewards/forensics_sinkhorn_reward": 0.8640268743038177,
      "rewards/forensics_strict_edge_reward": 0.10835880786180496,
      "rewards/forensics_tiou_reward": 0.7503436207771301,
      "step": 1015
    },
    {
      "completion_length": 26.734375,
      "epoch": 3.1167434715821813,
      "grad_norm": 8.963638305664062,
      "kl": 0.521484375,
      "learning_rate": 2.1923076923076922e-07,
      "loss": 0.0604,
      "reward": 0.7839497625827789,
      "reward_std": 0.027913257479667664,
      "reward_std_all": 0.027499329298734665,
      "rewards-max/forensics_sinkhorn_reward": 0.9841464459896088,
      "rewards-max/forensics_strict_edge_reward": 0.35550935566425323,
      "rewards-max/forensics_tiou_reward": 0.9663070738315582,
      "rewards/forensics_sinkhorn_reward": 0.8212552666664124,
      "rewards/forensics_strict_edge_reward": 0.04053772613406181,
      "rewards/forensics_tiou_reward": 0.7038279175758362,
      "step": 1016
    },
    {
      "completion_length": 27.75,
      "epoch": 3.119815668202765,
      "grad_norm": 7.527858734130859,
      "kl": 0.4560546875,
      "learning_rate": 2.1846153846153844e-07,
      "loss": 0.0565,
      "reward": 0.7905186414718628,
      "reward_std": 0.024228502064943314,
      "reward_std_all": 0.023454812821000814,
      "rewards-max/forensics_sinkhorn_reward": 0.9876415133476257,
      "rewards-max/forensics_strict_edge_reward": 0.4528360813856125,
      "rewards-max/forensics_tiou_reward": 0.9702747762203217,
      "rewards/forensics_sinkhorn_reward": 0.7782132625579834,
      "rewards/forensics_strict_edge_reward": 0.06589110195636749,
      "rewards/forensics_tiou_reward": 0.6347172260284424,
      "step": 1017
    },
    {
      "completion_length": 25.875,
      "epoch": 3.122887864823349,
      "grad_norm": 6.861810684204102,
      "kl": 0.466796875,
      "learning_rate": 2.1769230769230768e-07,
      "loss": -0.0411,
      "reward": 0.7951273918151855,
      "reward_std": 0.019154815934598446,
      "reward_std_all": 0.018736187368631363,
      "rewards-max/forensics_sinkhorn_reward": 0.9831738471984863,
      "rewards-max/forensics_strict_edge_reward": 0.5137558048591018,
      "rewards-max/forensics_tiou_reward": 0.9641115963459015,
      "rewards/forensics_sinkhorn_reward": 0.8312086462974548,
      "rewards/forensics_strict_edge_reward": 0.04627144732512534,
      "rewards/forensics_tiou_reward": 0.6690974235534668,
      "step": 1018
    },
    {
      "completion_length": 27.46875,
      "epoch": 3.1259600614439322,
      "grad_norm": 8.010222434997559,
      "kl": 0.3759765625,
      "learning_rate": 2.169230769230769e-07,
      "loss": -0.0789,
      "reward": 0.7997753620147705,
      "reward_std": 0.010946249996777624,
      "reward_std_all": 0.011553713819012046,
      "rewards-max/forensics_sinkhorn_reward": 0.9787947535514832,
      "rewards-max/forensics_strict_edge_reward": 0.5389202535152435,
      "rewards-max/forensics_tiou_reward": 0.9522608816623688,
      "rewards/forensics_sinkhorn_reward": 0.8071857392787933,
      "rewards/forensics_strict_edge_reward": 0.07264822348952293,
      "rewards/forensics_tiou_reward": 0.6368706524372101,
      "step": 1019
    },
    {
      "completion_length": 26.5625,
      "epoch": 3.129032258064516,
      "grad_norm": 9.150779724121094,
      "kl": 0.7578125,
      "learning_rate": 2.1615384615384613e-07,
      "loss": 0.0083,
      "reward": 0.8039962649345398,
      "reward_std": 0.004506773489993066,
      "reward_std_all": 0.005763927241787314,
      "rewards-max/forensics_sinkhorn_reward": 0.9923354685306549,
      "rewards-max/forensics_strict_edge_reward": 0.6706924736499786,
      "rewards-max/forensics_tiou_reward": 0.9847264885902405,
      "rewards/forensics_sinkhorn_reward": 0.811077892780304,
      "rewards/forensics_strict_edge_reward": 0.08039648458361626,
      "rewards/forensics_tiou_reward": 0.7048941552639008,
      "step": 1020
    },
    {
      "completion_length": 29.5,
      "epoch": 3.1321044546851,
      "grad_norm": 8.40793228149414,
      "kl": 0.4453125,
      "learning_rate": 2.153846153846154e-07,
      "loss": -0.0082,
      "reward": 0.7852519154548645,
      "reward_std": 0.02441895194351673,
      "reward_std_all": 0.02467519510537386,
      "rewards-max/forensics_sinkhorn_reward": 0.9728975296020508,
      "rewards-max/forensics_strict_edge_reward": 0.4251216650009155,
      "rewards-max/forensics_tiou_reward": 0.9383609592914581,
      "rewards/forensics_sinkhorn_reward": 0.7445433735847473,
      "rewards/forensics_strict_edge_reward": 0.043491654098033905,
      "rewards/forensics_tiou_reward": 0.5835034549236298,
      "step": 1021
    },
    {
      "completion_length": 25.625,
      "epoch": 3.1351766513056836,
      "grad_norm": 7.959000587463379,
      "kl": 0.4443359375,
      "learning_rate": 2.146153846153846e-07,
      "loss": -0.026,
      "reward": 0.7616203129291534,
      "reward_std": 0.042465690523386,
      "reward_std_all": 0.0447014644742012,
      "rewards-max/forensics_sinkhorn_reward": 0.9810871779918671,
      "rewards-max/forensics_strict_edge_reward": 0.5348145961761475,
      "rewards-max/forensics_tiou_reward": 0.9572597146034241,
      "rewards/forensics_sinkhorn_reward": 0.7166175544261932,
      "rewards/forensics_strict_edge_reward": 0.04715366940945387,
      "rewards/forensics_tiou_reward": 0.5567132830619812,
      "step": 1022
    },
    {
      "completion_length": 28.03125,
      "epoch": 3.1382488479262673,
      "grad_norm": 6.407602310180664,
      "kl": 0.4541015625,
      "learning_rate": 2.1384615384615385e-07,
      "loss": 0.0409,
      "reward": 0.791721522808075,
      "reward_std": 0.01652515446767211,
      "reward_std_all": 0.0172410411760211,
      "rewards-max/forensics_sinkhorn_reward": 0.9923664331436157,
      "rewards-max/forensics_strict_edge_reward": 0.7032848298549652,
      "rewards-max/forensics_tiou_reward": 0.9801067411899567,
      "rewards/forensics_sinkhorn_reward": 0.8112156391143799,
      "rewards/forensics_strict_edge_reward": 0.10440951958298683,
      "rewards/forensics_tiou_reward": 0.6350829005241394,
      "step": 1023
    },
    {
      "completion_length": 24.53125,
      "epoch": 3.141321044546851,
      "grad_norm": 7.8304057121276855,
      "kl": 0.501953125,
      "learning_rate": 2.1307692307692307e-07,
      "loss": -0.0292,
      "reward": 0.7951012551784515,
      "reward_std": 0.019867890048772097,
      "reward_std_all": 0.018828388303518295,
      "rewards-max/forensics_sinkhorn_reward": 0.9944486916065216,
      "rewards-max/forensics_strict_edge_reward": 0.6212178468704224,
      "rewards-max/forensics_tiou_reward": 0.9891217947006226,
      "rewards/forensics_sinkhorn_reward": 0.8054540753364563,
      "rewards/forensics_strict_edge_reward": 0.09171435609459877,
      "rewards/forensics_tiou_reward": 0.6648297905921936,
      "step": 1024
    },
    {
      "completion_length": 21.96875,
      "epoch": 3.144393241167435,
      "grad_norm": 8.047429084777832,
      "kl": 0.5361328125,
      "learning_rate": 2.123076923076923e-07,
      "loss": -0.0894,
      "reward": 0.7993150353431702,
      "reward_std": 0.013320078141987324,
      "reward_std_all": 0.013224279507994652,
      "rewards-max/forensics_sinkhorn_reward": 0.9945515990257263,
      "rewards-max/forensics_strict_edge_reward": 0.7511312067508698,
      "rewards-max/forensics_tiou_reward": 0.9918400347232819,
      "rewards/forensics_sinkhorn_reward": 0.847673237323761,
      "rewards/forensics_strict_edge_reward": 0.10642437636852264,
      "rewards/forensics_tiou_reward": 0.700612872838974,
      "step": 1025
    },
    {
      "completion_length": 29.484375,
      "epoch": 3.1474654377880182,
      "grad_norm": 9.559612274169922,
      "kl": 0.4951171875,
      "learning_rate": 2.1153846153846152e-07,
      "loss": -0.0061,
      "reward": 0.7965729832649231,
      "reward_std": 0.01087632030248642,
      "reward_std_all": 0.012356589548289776,
      "rewards-max/forensics_sinkhorn_reward": 0.9901724457740784,
      "rewards-max/forensics_strict_edge_reward": 0.42256493866443634,
      "rewards-max/forensics_tiou_reward": 0.9772316217422485,
      "rewards/forensics_sinkhorn_reward": 0.7895000278949738,
      "rewards/forensics_strict_edge_reward": 0.05761254206299782,
      "rewards/forensics_tiou_reward": 0.6426801681518555,
      "step": 1026
    },
    {
      "completion_length": 25.828125,
      "epoch": 3.150537634408602,
      "grad_norm": 6.496703624725342,
      "kl": 0.4619140625,
      "learning_rate": 2.1076923076923076e-07,
      "loss": -0.0763,
      "reward": 0.8024469316005707,
      "reward_std": 0.006370052462443709,
      "reward_std_all": 0.0075488006696105,
      "rewards-max/forensics_sinkhorn_reward": 0.9797661602497101,
      "rewards-max/forensics_strict_edge_reward": 0.5676676407456398,
      "rewards-max/forensics_tiou_reward": 0.9550109803676605,
      "rewards/forensics_sinkhorn_reward": 0.7898756861686707,
      "rewards/forensics_strict_edge_reward": 0.08730533625930548,
      "rewards/forensics_tiou_reward": 0.6169053018093109,
      "step": 1027
    },
    {
      "completion_length": 21.671875,
      "epoch": 3.153609831029186,
      "grad_norm": 11.313554763793945,
      "kl": 0.3896484375,
      "learning_rate": 2.0999999999999997e-07,
      "loss": 0.0344,
      "reward": 0.7951144576072693,
      "reward_std": 0.017032704316079617,
      "reward_std_all": 0.016992981545627117,
      "rewards-max/forensics_sinkhorn_reward": 0.976348340511322,
      "rewards-max/forensics_strict_edge_reward": 0.32120537757873535,
      "rewards-max/forensics_tiou_reward": 0.9328241050243378,
      "rewards/forensics_sinkhorn_reward": 0.8111256957054138,
      "rewards/forensics_strict_edge_reward": 0.03889681678265333,
      "rewards/forensics_tiou_reward": 0.6164595186710358,
      "step": 1028
    },
    {
      "completion_length": 26.234375,
      "epoch": 3.1566820276497696,
      "grad_norm": 7.787617206573486,
      "kl": 0.546875,
      "learning_rate": 2.0923076923076921e-07,
      "loss": 0.0664,
      "reward": 0.7625203728675842,
      "reward_std": 0.051581552252173424,
      "reward_std_all": 0.05168035812675953,
      "rewards-max/forensics_sinkhorn_reward": 0.9878175556659698,
      "rewards-max/forensics_strict_edge_reward": 0.6042353510856628,
      "rewards-max/forensics_tiou_reward": 0.9682635962963104,
      "rewards/forensics_sinkhorn_reward": 0.7505797743797302,
      "rewards/forensics_strict_edge_reward": 0.052894992753863335,
      "rewards/forensics_tiou_reward": 0.5822676420211792,
      "step": 1029
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.1597542242703534,
      "grad_norm": 7.523263931274414,
      "kl": 0.455078125,
      "learning_rate": 2.0846153846153843e-07,
      "loss": -0.0401,
      "reward": 0.786542683839798,
      "reward_std": 0.018694974947720766,
      "reward_std_all": 0.02028077165596187,
      "rewards-max/forensics_sinkhorn_reward": 0.9971720278263092,
      "rewards-max/forensics_strict_edge_reward": 0.9442227482795715,
      "rewards-max/forensics_tiou_reward": 0.9967693984508514,
      "rewards/forensics_sinkhorn_reward": 0.7989981770515442,
      "rewards/forensics_strict_edge_reward": 0.10329161956906319,
      "rewards/forensics_tiou_reward": 0.6516062915325165,
      "step": 1030
    },
    {
      "completion_length": 24.671875,
      "epoch": 3.162826420890937,
      "grad_norm": 9.976314544677734,
      "kl": 0.40234375,
      "learning_rate": 2.076923076923077e-07,
      "loss": -0.0557,
      "reward": 0.8050013184547424,
      "reward_std": 0.0036927188048139215,
      "reward_std_all": 0.004984004306606948,
      "rewards-max/forensics_sinkhorn_reward": 0.995241641998291,
      "rewards-max/forensics_strict_edge_reward": 0.7565081417560577,
      "rewards-max/forensics_tiou_reward": 0.9934422671794891,
      "rewards/forensics_sinkhorn_reward": 0.8603909909725189,
      "rewards/forensics_strict_edge_reward": 0.12456881999969482,
      "rewards/forensics_tiou_reward": 0.7086225748062134,
      "step": 1031
    },
    {
      "completion_length": 28.21875,
      "epoch": 3.165898617511521,
      "grad_norm": 7.202932357788086,
      "kl": 0.5078125,
      "learning_rate": 2.0692307692307694e-07,
      "loss": -0.0617,
      "reward": 0.7595867216587067,
      "reward_std": 0.022458245512098074,
      "reward_std_all": 0.0364353246986866,
      "rewards-max/forensics_sinkhorn_reward": 0.9914971888065338,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9847341775894165,
      "rewards/forensics_sinkhorn_reward": 0.7999809086322784,
      "rewards/forensics_strict_edge_reward": 0.1109415590763092,
      "rewards/forensics_tiou_reward": 0.6391288936138153,
      "step": 1032
    },
    {
      "completion_length": 24.53125,
      "epoch": 3.1689708141321047,
      "grad_norm": 8.916236877441406,
      "kl": 0.3740234375,
      "learning_rate": 2.0615384615384615e-07,
      "loss": 0.0072,
      "reward": 0.7720115482807159,
      "reward_std": 0.04578505456447601,
      "reward_std_all": 0.043148916214704514,
      "rewards-max/forensics_sinkhorn_reward": 0.9929500222206116,
      "rewards-max/forensics_strict_edge_reward": 0.6237462311983109,
      "rewards-max/forensics_tiou_reward": 0.9864647388458252,
      "rewards/forensics_sinkhorn_reward": 0.7831440567970276,
      "rewards/forensics_strict_edge_reward": 0.05293551832437515,
      "rewards/forensics_tiou_reward": 0.5658978521823883,
      "step": 1033
    },
    {
      "completion_length": 28.78125,
      "epoch": 3.172043010752688,
      "grad_norm": 9.234286308288574,
      "kl": 0.57421875,
      "learning_rate": 2.053846153846154e-07,
      "loss": 0.0332,
      "reward": 0.7691978812217712,
      "reward_std": 0.044982677325606346,
      "reward_std_all": 0.04327051714062691,
      "rewards-max/forensics_sinkhorn_reward": 0.9960377216339111,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9953659772872925,
      "rewards/forensics_sinkhorn_reward": 0.7814270853996277,
      "rewards/forensics_strict_edge_reward": 0.10195168852806091,
      "rewards/forensics_tiou_reward": 0.6588990986347198,
      "step": 1034
    },
    {
      "completion_length": 23.234375,
      "epoch": 3.175115207373272,
      "grad_norm": 9.37611198425293,
      "kl": 0.3818359375,
      "learning_rate": 2.046153846153846e-07,
      "loss": -0.0749,
      "reward": 0.7957993447780609,
      "reward_std": 0.017902206163853407,
      "reward_std_all": 0.017684859223663807,
      "rewards-max/forensics_sinkhorn_reward": 0.9859788119792938,
      "rewards-max/forensics_strict_edge_reward": 0.40730567276477814,
      "rewards-max/forensics_tiou_reward": 0.9704178273677826,
      "rewards/forensics_sinkhorn_reward": 0.8551211953163147,
      "rewards/forensics_strict_edge_reward": 0.04103003814816475,
      "rewards/forensics_tiou_reward": 0.6806952953338623,
      "step": 1035
    },
    {
      "completion_length": 25.96875,
      "epoch": 3.1781874039938556,
      "grad_norm": 9.137595176696777,
      "kl": 0.35546875,
      "learning_rate": 2.0384615384615384e-07,
      "loss": 0.1172,
      "reward": 0.7389089465141296,
      "reward_std": 0.02165909670293331,
      "reward_std_all": 0.04498104192316532,
      "rewards-max/forensics_sinkhorn_reward": 0.9930850267410278,
      "rewards-max/forensics_strict_edge_reward": 0.6524691581726074,
      "rewards-max/forensics_tiou_reward": 0.9866218566894531,
      "rewards/forensics_sinkhorn_reward": 0.7273980379104614,
      "rewards/forensics_strict_edge_reward": 0.058980418369174004,
      "rewards/forensics_tiou_reward": 0.5292454063892365,
      "step": 1036
    },
    {
      "completion_length": 31.953125,
      "epoch": 3.1812596006144394,
      "grad_norm": 7.11990213394165,
      "kl": 0.404296875,
      "learning_rate": 2.0307692307692306e-07,
      "loss": 0.0109,
      "reward": 0.7907060980796814,
      "reward_std": 0.01718732068547979,
      "reward_std_all": 0.01805585518013686,
      "rewards-max/forensics_sinkhorn_reward": 0.9924259185791016,
      "rewards-max/forensics_strict_edge_reward": 0.5955832898616791,
      "rewards-max/forensics_tiou_reward": 0.9854381680488586,
      "rewards/forensics_sinkhorn_reward": 0.8043520748615265,
      "rewards/forensics_strict_edge_reward": 0.07398675754666328,
      "rewards/forensics_tiou_reward": 0.6506545543670654,
      "step": 1037
    },
    {
      "completion_length": 28.71875,
      "epoch": 3.184331797235023,
      "grad_norm": 7.945591926574707,
      "kl": 0.43359375,
      "learning_rate": 2.023076923076923e-07,
      "loss": -0.0524,
      "reward": 0.7929233610630035,
      "reward_std": 0.015222099493257701,
      "reward_std_all": 0.016291818930767477,
      "rewards-max/forensics_sinkhorn_reward": 0.9790216088294983,
      "rewards-max/forensics_strict_edge_reward": 0.5889547169208527,
      "rewards-max/forensics_tiou_reward": 0.948736846446991,
      "rewards/forensics_sinkhorn_reward": 0.798314243555069,
      "rewards/forensics_strict_edge_reward": 0.06933434121310711,
      "rewards/forensics_tiou_reward": 0.6505659818649292,
      "step": 1038
    },
    {
      "completion_length": 26.109375,
      "epoch": 3.187403993855607,
      "grad_norm": 7.343130588531494,
      "kl": 0.58203125,
      "learning_rate": 2.015384615384615e-07,
      "loss": -0.087,
      "reward": 0.783380389213562,
      "reward_std": 0.01868082315195352,
      "reward_std_all": 0.021352474461309612,
      "rewards-max/forensics_sinkhorn_reward": 0.9849674105644226,
      "rewards-max/forensics_strict_edge_reward": 0.5244060158729553,
      "rewards-max/forensics_tiou_reward": 0.9643359780311584,
      "rewards/forensics_sinkhorn_reward": 0.83094921708107,
      "rewards/forensics_strict_edge_reward": 0.07075317576527596,
      "rewards/forensics_tiou_reward": 0.644146740436554,
      "step": 1039
    },
    {
      "completion_length": 27.8125,
      "epoch": 3.1904761904761907,
      "grad_norm": 6.751374244689941,
      "kl": 0.3896484375,
      "learning_rate": 2.0076923076923075e-07,
      "loss": -0.0406,
      "reward": 0.7788733243942261,
      "reward_std": 0.032281752210110426,
      "reward_std_all": 0.031818469520658255,
      "rewards-max/forensics_sinkhorn_reward": 0.9902163445949554,
      "rewards-max/forensics_strict_edge_reward": 0.8360384404659271,
      "rewards-max/forensics_tiou_reward": 0.9798360466957092,
      "rewards/forensics_sinkhorn_reward": 0.7824916541576385,
      "rewards/forensics_strict_edge_reward": 0.09777095168828964,
      "rewards/forensics_tiou_reward": 0.6030389070510864,
      "step": 1040
    },
    {
      "completion_length": 27.671875,
      "epoch": 3.193548387096774,
      "grad_norm": 7.601099491119385,
      "kl": 0.4423828125,
      "learning_rate": 2e-07,
      "loss": 0.0817,
      "reward": 0.7792418897151947,
      "reward_std": 0.039284782484173775,
      "reward_std_all": 0.036674944683909416,
      "rewards-max/forensics_sinkhorn_reward": 0.9864098131656647,
      "rewards-max/forensics_strict_edge_reward": 0.5401631891727448,
      "rewards-max/forensics_tiou_reward": 0.9688520431518555,
      "rewards/forensics_sinkhorn_reward": 0.7881095707416534,
      "rewards/forensics_strict_edge_reward": 0.057745885103940964,
      "rewards/forensics_tiou_reward": 0.6416893005371094,
      "step": 1041
    },
    {
      "completion_length": 28.34375,
      "epoch": 3.196620583717358,
      "grad_norm": 7.261686325073242,
      "kl": 0.4833984375,
      "learning_rate": 1.9923076923076923e-07,
      "loss": 0.1408,
      "reward": 0.7368608117103577,
      "reward_std": 0.0528164878487587,
      "reward_std_all": 0.06530371494591236,
      "rewards-max/forensics_sinkhorn_reward": 0.9905888736248016,
      "rewards-max/forensics_strict_edge_reward": 0.6377167105674744,
      "rewards-max/forensics_tiou_reward": 0.9813041090965271,
      "rewards/forensics_sinkhorn_reward": 0.7468223571777344,
      "rewards/forensics_strict_edge_reward": 0.10251970402896404,
      "rewards/forensics_tiou_reward": 0.6021969616413116,
      "step": 1042
    },
    {
      "completion_length": 23.140625,
      "epoch": 3.1996927803379416,
      "grad_norm": 8.836498260498047,
      "kl": 0.373046875,
      "learning_rate": 1.9846153846153844e-07,
      "loss": -0.0467,
      "reward": 0.7938726246356964,
      "reward_std": 0.012298714835196733,
      "reward_std_all": 0.013860814273357391,
      "rewards-max/forensics_sinkhorn_reward": 0.9936729371547699,
      "rewards-max/forensics_strict_edge_reward": 0.8360384404659271,
      "rewards-max/forensics_tiou_reward": 0.9888424575328827,
      "rewards/forensics_sinkhorn_reward": 0.8214543461799622,
      "rewards/forensics_strict_edge_reward": 0.136975459754467,
      "rewards/forensics_tiou_reward": 0.6235316097736359,
      "step": 1043
    },
    {
      "completion_length": 28.6875,
      "epoch": 3.2027649769585254,
      "grad_norm": 9.06728744506836,
      "kl": 0.529296875,
      "learning_rate": 1.9769230769230768e-07,
      "loss": -0.0229,
      "reward": 0.7908191680908203,
      "reward_std": 0.015278979437425733,
      "reward_std_all": 0.016674889717251062,
      "rewards-max/forensics_sinkhorn_reward": 0.9941563606262207,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9902341663837433,
      "rewards/forensics_sinkhorn_reward": 0.8324616551399231,
      "rewards/forensics_strict_edge_reward": 0.12559890002012253,
      "rewards/forensics_tiou_reward": 0.6781806349754333,
      "step": 1044
    },
    {
      "completion_length": 23.0625,
      "epoch": 3.205837173579109,
      "grad_norm": 7.453777313232422,
      "kl": 0.4580078125,
      "learning_rate": 1.9692307692307693e-07,
      "loss": 0.0296,
      "reward": 0.7318911850452423,
      "reward_std": 0.03968910127878189,
      "reward_std_all": 0.05640373192727566,
      "rewards-max/forensics_sinkhorn_reward": 0.9880678355693817,
      "rewards-max/forensics_strict_edge_reward": 0.7053695321083069,
      "rewards-max/forensics_tiou_reward": 0.9735333025455475,
      "rewards/forensics_sinkhorn_reward": 0.6960986256599426,
      "rewards/forensics_strict_edge_reward": 0.07600412331521511,
      "rewards/forensics_tiou_reward": 0.5502915680408478,
      "step": 1045
    },
    {
      "completion_length": 25.921875,
      "epoch": 3.208909370199693,
      "grad_norm": 7.434966564178467,
      "kl": 0.4697265625,
      "learning_rate": 1.9615384615384614e-07,
      "loss": 0.13,
      "reward": 0.7503799796104431,
      "reward_std": 0.05397051014006138,
      "reward_std_all": 0.055094173178076744,
      "rewards-max/forensics_sinkhorn_reward": 0.9875594675540924,
      "rewards-max/forensics_strict_edge_reward": 0.5322106778621674,
      "rewards-max/forensics_tiou_reward": 0.9690609276294708,
      "rewards/forensics_sinkhorn_reward": 0.7119312286376953,
      "rewards/forensics_strict_edge_reward": 0.04452528432011604,
      "rewards/forensics_tiou_reward": 0.582271546125412,
      "step": 1046
    },
    {
      "completion_length": 23.703125,
      "epoch": 3.2119815668202767,
      "grad_norm": 22.440324783325195,
      "kl": 0.8818359375,
      "learning_rate": 1.9538461538461538e-07,
      "loss": 0.0052,
      "reward": 0.7985417246818542,
      "reward_std": 0.01340500672813505,
      "reward_std_all": 0.013599355472251773,
      "rewards-max/forensics_sinkhorn_reward": 0.9962644875049591,
      "rewards-max/forensics_strict_edge_reward": 0.7589184939861298,
      "rewards-max/forensics_tiou_reward": 0.9933993816375732,
      "rewards/forensics_sinkhorn_reward": 0.8095277845859528,
      "rewards/forensics_strict_edge_reward": 0.13167321495711803,
      "rewards/forensics_tiou_reward": 0.6234591007232666,
      "step": 1047
    },
    {
      "completion_length": 23.828125,
      "epoch": 3.21505376344086,
      "grad_norm": 12.084380149841309,
      "kl": 1.0380859375,
      "learning_rate": 1.946153846153846e-07,
      "loss": 0.1475,
      "reward": 0.7711828052997589,
      "reward_std": 0.0436903927475214,
      "reward_std_all": 0.041963886469602585,
      "rewards-max/forensics_sinkhorn_reward": 0.9769070744514465,
      "rewards-max/forensics_strict_edge_reward": 0.6284713298082352,
      "rewards-max/forensics_tiou_reward": 0.9398075342178345,
      "rewards/forensics_sinkhorn_reward": 0.786780059337616,
      "rewards/forensics_strict_edge_reward": 0.09889369830489159,
      "rewards/forensics_tiou_reward": 0.5744982063770294,
      "step": 1048
    },
    {
      "completion_length": 26.171875,
      "epoch": 3.218125960061444,
      "grad_norm": 8.100576400756836,
      "kl": 0.4833984375,
      "learning_rate": 1.9384615384615383e-07,
      "loss": 0.077,
      "reward": 0.7726084291934967,
      "reward_std": 0.04966719442745671,
      "reward_std_all": 0.04589614819269627,
      "rewards-max/forensics_sinkhorn_reward": 0.9886212050914764,
      "rewards-max/forensics_strict_edge_reward": 0.6573699712753296,
      "rewards-max/forensics_tiou_reward": 0.9758254885673523,
      "rewards/forensics_sinkhorn_reward": 0.7576882541179657,
      "rewards/forensics_strict_edge_reward": 0.06975927576422691,
      "rewards/forensics_tiou_reward": 0.561726838350296,
      "step": 1049
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.2211981566820276,
      "grad_norm": 7.687995433807373,
      "kl": 0.537109375,
      "learning_rate": 1.9307692307692307e-07,
      "loss": -0.037,
      "reward": 0.7923834323883057,
      "reward_std": 0.022885290789417922,
      "reward_std_all": 0.02169196493923664,
      "rewards-max/forensics_sinkhorn_reward": 0.9934666752815247,
      "rewards-max/forensics_strict_edge_reward": 0.595534548163414,
      "rewards-max/forensics_tiou_reward": 0.9868558943271637,
      "rewards/forensics_sinkhorn_reward": 0.852967381477356,
      "rewards/forensics_strict_edge_reward": 0.09649091213941574,
      "rewards/forensics_tiou_reward": 0.7067160308361053,
      "step": 1050
    },
    {
      "completion_length": 24.734375,
      "epoch": 3.2242703533026114,
      "grad_norm": 8.872733116149902,
      "kl": 0.4033203125,
      "learning_rate": 1.9230769230769231e-07,
      "loss": -0.0034,
      "reward": 0.7599896490573883,
      "reward_std": 0.008998723234981298,
      "reward_std_all": 0.03008402418345213,
      "rewards-max/forensics_sinkhorn_reward": 0.9883653819561005,
      "rewards-max/forensics_strict_edge_reward": 0.5452546775341034,
      "rewards-max/forensics_tiou_reward": 0.9743146002292633,
      "rewards/forensics_sinkhorn_reward": 0.7676591277122498,
      "rewards/forensics_strict_edge_reward": 0.07925636693835258,
      "rewards/forensics_tiou_reward": 0.6591472923755646,
      "step": 1051
    },
    {
      "completion_length": 25.984375,
      "epoch": 3.227342549923195,
      "grad_norm": 14.128067970275879,
      "kl": 1.0,
      "learning_rate": 1.9153846153846153e-07,
      "loss": -0.0006,
      "reward": 0.7833864092826843,
      "reward_std": 0.035699755884706974,
      "reward_std_all": 0.0332126384600997,
      "rewards-max/forensics_sinkhorn_reward": 0.994008868932724,
      "rewards-max/forensics_strict_edge_reward": 0.6274522840976715,
      "rewards-max/forensics_tiou_reward": 0.985971987247467,
      "rewards/forensics_sinkhorn_reward": 0.7939279973506927,
      "rewards/forensics_strict_edge_reward": 0.07734471932053566,
      "rewards/forensics_tiou_reward": 0.6292480230331421,
      "step": 1052
    },
    {
      "completion_length": 27.984375,
      "epoch": 3.230414746543779,
      "grad_norm": 7.504277229309082,
      "kl": 0.396484375,
      "learning_rate": 1.9076923076923077e-07,
      "loss": -0.049,
      "reward": 0.7901023030281067,
      "reward_std": 0.026059972820803523,
      "reward_std_all": 0.024961286573670805,
      "rewards-max/forensics_sinkhorn_reward": 0.9943835735321045,
      "rewards-max/forensics_strict_edge_reward": 0.6714634150266647,
      "rewards-max/forensics_tiou_reward": 0.9900472462177277,
      "rewards/forensics_sinkhorn_reward": 0.8499102890491486,
      "rewards/forensics_strict_edge_reward": 0.08408949337899685,
      "rewards/forensics_tiou_reward": 0.6899988353252411,
      "step": 1053
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.2334869431643627,
      "grad_norm": 7.521747589111328,
      "kl": 0.365234375,
      "learning_rate": 1.8999999999999998e-07,
      "loss": 0.0349,
      "reward": 0.7968805134296417,
      "reward_std": 0.014686715323477983,
      "reward_std_all": 0.014918138971552253,
      "rewards-max/forensics_sinkhorn_reward": 0.9794013202190399,
      "rewards-max/forensics_strict_edge_reward": 0.3603787049651146,
      "rewards-max/forensics_tiou_reward": 0.946301132440567,
      "rewards/forensics_sinkhorn_reward": 0.8112502098083496,
      "rewards/forensics_strict_edge_reward": 0.06973514333367348,
      "rewards/forensics_tiou_reward": 0.6658637821674347,
      "step": 1054
    },
    {
      "completion_length": 24.96875,
      "epoch": 3.236559139784946,
      "grad_norm": 7.22048807144165,
      "kl": 0.5068359375,
      "learning_rate": 1.8923076923076922e-07,
      "loss": 0.0498,
      "reward": 0.7805387377738953,
      "reward_std": 0.041923449374735355,
      "reward_std_all": 0.038642133586108685,
      "rewards-max/forensics_sinkhorn_reward": 0.9825919270515442,
      "rewards-max/forensics_strict_edge_reward": 0.4219362586736679,
      "rewards-max/forensics_tiou_reward": 0.9604517221450806,
      "rewards/forensics_sinkhorn_reward": 0.7852011919021606,
      "rewards/forensics_strict_edge_reward": 0.0548232588917017,
      "rewards/forensics_tiou_reward": 0.6292471289634705,
      "step": 1055
    },
    {
      "completion_length": 25.34375,
      "epoch": 3.23963133640553,
      "grad_norm": 6.836135387420654,
      "kl": 0.3603515625,
      "learning_rate": 1.8846153846153846e-07,
      "loss": -0.0076,
      "reward": 0.7733533680438995,
      "reward_std": 0.028456755680963397,
      "reward_std_all": 0.034142215736210346,
      "rewards-max/forensics_sinkhorn_reward": 0.9862486720085144,
      "rewards-max/forensics_strict_edge_reward": 0.5901642739772797,
      "rewards-max/forensics_tiou_reward": 0.972326785326004,
      "rewards/forensics_sinkhorn_reward": 0.7473717331886292,
      "rewards/forensics_strict_edge_reward": 0.03671798203140497,
      "rewards/forensics_tiou_reward": 0.5737969279289246,
      "step": 1056
    },
    {
      "completion_length": 27.75,
      "epoch": 3.2427035330261136,
      "grad_norm": 8.582878112792969,
      "kl": 0.505859375,
      "learning_rate": 1.8769230769230767e-07,
      "loss": -0.0687,
      "reward": 0.7944774031639099,
      "reward_std": 0.019709271378815174,
      "reward_std_all": 0.019298993982374668,
      "rewards-max/forensics_sinkhorn_reward": 0.9938655197620392,
      "rewards-max/forensics_strict_edge_reward": 0.5525192320346832,
      "rewards-max/forensics_tiou_reward": 0.9871877729892731,
      "rewards/forensics_sinkhorn_reward": 0.801464170217514,
      "rewards/forensics_strict_edge_reward": 0.08104685693979263,
      "rewards/forensics_tiou_reward": 0.6569502949714661,
      "step": 1057
    },
    {
      "completion_length": 25.21875,
      "epoch": 3.2457757296466974,
      "grad_norm": 7.357495307922363,
      "kl": 0.421875,
      "learning_rate": 1.8692307692307692e-07,
      "loss": -0.0129,
      "reward": 0.7979715764522552,
      "reward_std": 0.010073726996779442,
      "reward_std_all": 0.011487816460430622,
      "rewards-max/forensics_sinkhorn_reward": 0.9598365128040314,
      "rewards-max/forensics_strict_edge_reward": 0.2183588519692421,
      "rewards-max/forensics_tiou_reward": 0.9085713922977448,
      "rewards/forensics_sinkhorn_reward": 0.7912038862705231,
      "rewards/forensics_strict_edge_reward": 0.028560731559991837,
      "rewards/forensics_tiou_reward": 0.6206274926662445,
      "step": 1058
    },
    {
      "completion_length": 25.25,
      "epoch": 3.248847926267281,
      "grad_norm": 7.812993049621582,
      "kl": 0.357421875,
      "learning_rate": 1.8615384615384613e-07,
      "loss": 0.0112,
      "reward": 0.7966318428516388,
      "reward_std": 0.015994985296856612,
      "reward_std_all": 0.016053782543167472,
      "rewards-max/forensics_sinkhorn_reward": 0.9901972413063049,
      "rewards-max/forensics_strict_edge_reward": 0.6652968376874924,
      "rewards-max/forensics_tiou_reward": 0.9792750775814056,
      "rewards/forensics_sinkhorn_reward": 0.7982811331748962,
      "rewards/forensics_strict_edge_reward": 0.05946521274745464,
      "rewards/forensics_tiou_reward": 0.6229669451713562,
      "step": 1059
    },
    {
      "completion_length": 24.0625,
      "epoch": 3.251920122887865,
      "grad_norm": 8.898226737976074,
      "kl": 0.5966796875,
      "learning_rate": 1.853846153846154e-07,
      "loss": 0.0601,
      "reward": 0.7494568824768066,
      "reward_std": 0.0437060808762908,
      "reward_std_all": 0.05229992885142565,
      "rewards-max/forensics_sinkhorn_reward": 0.9916248023509979,
      "rewards-max/forensics_strict_edge_reward": 0.3701426535844803,
      "rewards-max/forensics_tiou_reward": 0.9824292361736298,
      "rewards/forensics_sinkhorn_reward": 0.7164702713489532,
      "rewards/forensics_strict_edge_reward": 0.04688766226172447,
      "rewards/forensics_tiou_reward": 0.5927717089653015,
      "step": 1060
    },
    {
      "completion_length": 24.8125,
      "epoch": 3.2549923195084487,
      "grad_norm": 7.734612464904785,
      "kl": 0.443359375,
      "learning_rate": 1.846153846153846e-07,
      "loss": -0.0104,
      "reward": 0.7939915060997009,
      "reward_std": 0.020884276600554585,
      "reward_std_all": 0.019643154344521463,
      "rewards-max/forensics_sinkhorn_reward": 0.9975062608718872,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8366240859031677,
      "rewards/forensics_strict_edge_reward": 0.1571243330836296,
      "rewards/forensics_tiou_reward": 0.669546902179718,
      "step": 1061
    },
    {
      "completion_length": 22.96875,
      "epoch": 3.258064516129032,
      "grad_norm": 9.06340217590332,
      "kl": 0.5673828125,
      "learning_rate": 1.8384615384615385e-07,
      "loss": -0.0824,
      "reward": 0.7628699243068695,
      "reward_std": 0.017940467223525047,
      "reward_std_all": 0.03262057341635227,
      "rewards-max/forensics_sinkhorn_reward": 0.9903722107410431,
      "rewards-max/forensics_strict_edge_reward": 0.7769311368465424,
      "rewards-max/forensics_tiou_reward": 0.9815578758716583,
      "rewards/forensics_sinkhorn_reward": 0.7666159272193909,
      "rewards/forensics_strict_edge_reward": 0.08590240031480789,
      "rewards/forensics_tiou_reward": 0.5707451701164246,
      "step": 1062
    },
    {
      "completion_length": 22.515625,
      "epoch": 3.261136712749616,
      "grad_norm": 9.78145980834961,
      "kl": 0.3603515625,
      "learning_rate": 1.8307692307692306e-07,
      "loss": -0.0811,
      "reward": 0.7687021195888519,
      "reward_std": 0.012860799324698746,
      "reward_std_all": 0.026432067388668656,
      "rewards-max/forensics_sinkhorn_reward": 0.9929730892181396,
      "rewards-max/forensics_strict_edge_reward": 0.6095658242702484,
      "rewards-max/forensics_tiou_reward": 0.9874280095100403,
      "rewards/forensics_sinkhorn_reward": 0.78209188580513,
      "rewards/forensics_strict_edge_reward": 0.07860651053488255,
      "rewards/forensics_tiou_reward": 0.6406514048576355,
      "step": 1063
    },
    {
      "completion_length": 27.28125,
      "epoch": 3.2642089093701996,
      "grad_norm": 7.077240467071533,
      "kl": 0.4873046875,
      "learning_rate": 1.823076923076923e-07,
      "loss": 0.0253,
      "reward": 0.7874098122119904,
      "reward_std": 0.02357808593660593,
      "reward_std_all": 0.023539231158792973,
      "rewards-max/forensics_sinkhorn_reward": 0.994440346956253,
      "rewards-max/forensics_strict_edge_reward": 0.6364709585905075,
      "rewards-max/forensics_tiou_reward": 0.9904276132583618,
      "rewards/forensics_sinkhorn_reward": 0.8019729554653168,
      "rewards/forensics_strict_edge_reward": 0.06930016726255417,
      "rewards/forensics_tiou_reward": 0.651838630437851,
      "step": 1064
    },
    {
      "completion_length": 27.046875,
      "epoch": 3.2672811059907834,
      "grad_norm": 8.180957794189453,
      "kl": 0.462890625,
      "learning_rate": 1.8153846153846152e-07,
      "loss": 0.0203,
      "reward": 0.7998310923576355,
      "reward_std": 0.011671071872115135,
      "reward_std_all": 0.011965508572757244,
      "rewards-max/forensics_sinkhorn_reward": 0.9932990074157715,
      "rewards-max/forensics_strict_edge_reward": 0.7457418143749237,
      "rewards-max/forensics_tiou_reward": 0.9867348372936249,
      "rewards/forensics_sinkhorn_reward": 0.8219655454158783,
      "rewards/forensics_strict_edge_reward": 0.10190966352820396,
      "rewards/forensics_tiou_reward": 0.6426216959953308,
      "step": 1065
    },
    {
      "completion_length": 22.609375,
      "epoch": 3.270353302611367,
      "grad_norm": 8.281448364257812,
      "kl": 0.474609375,
      "learning_rate": 1.8076923076923076e-07,
      "loss": -0.0644,
      "reward": 0.8055005371570587,
      "reward_std": 0.001298719027545303,
      "reward_std_all": 0.003386460361070931,
      "rewards-max/forensics_sinkhorn_reward": 0.988578587770462,
      "rewards-max/forensics_strict_edge_reward": 0.47726282477378845,
      "rewards-max/forensics_tiou_reward": 0.9752153754234314,
      "rewards/forensics_sinkhorn_reward": 0.8582761883735657,
      "rewards/forensics_strict_edge_reward": 0.0511620594188571,
      "rewards/forensics_tiou_reward": 0.6686933636665344,
      "step": 1066
    },
    {
      "completion_length": 26.90625,
      "epoch": 3.273425499231951,
      "grad_norm": 7.132514476776123,
      "kl": 0.474609375,
      "learning_rate": 1.8e-07,
      "loss": -0.0309,
      "reward": 0.7941299974918365,
      "reward_std": 0.020920903945807368,
      "reward_std_all": 0.020158675499260426,
      "rewards-max/forensics_sinkhorn_reward": 0.9912763237953186,
      "rewards-max/forensics_strict_edge_reward": 0.6761750876903534,
      "rewards-max/forensics_tiou_reward": 0.981611967086792,
      "rewards/forensics_sinkhorn_reward": 0.8164132237434387,
      "rewards/forensics_strict_edge_reward": 0.08579843863844872,
      "rewards/forensics_tiou_reward": 0.6360439956188202,
      "step": 1067
    },
    {
      "completion_length": 25.828125,
      "epoch": 3.2764976958525347,
      "grad_norm": 7.850934982299805,
      "kl": 0.5478515625,
      "learning_rate": 1.792307692307692e-07,
      "loss": 0.0205,
      "reward": 0.789993554353714,
      "reward_std": 0.024304023012518883,
      "reward_std_all": 0.023223754949867725,
      "rewards-max/forensics_sinkhorn_reward": 0.991643875837326,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9850738048553467,
      "rewards/forensics_sinkhorn_reward": 0.8038020431995392,
      "rewards/forensics_strict_edge_reward": 0.14014934748411179,
      "rewards/forensics_tiou_reward": 0.6654897034168243,
      "step": 1068
    },
    {
      "completion_length": 25.671875,
      "epoch": 3.279569892473118,
      "grad_norm": 7.7883100509643555,
      "kl": 0.5263671875,
      "learning_rate": 1.7846153846153845e-07,
      "loss": 0.0541,
      "reward": 0.7576049566268921,
      "reward_std": 0.02745223417878151,
      "reward_std_all": 0.0392691376619041,
      "rewards-max/forensics_sinkhorn_reward": 0.9924815893173218,
      "rewards-max/forensics_strict_edge_reward": 0.6928144693374634,
      "rewards-max/forensics_tiou_reward": 0.9852930009365082,
      "rewards/forensics_sinkhorn_reward": 0.7876647710800171,
      "rewards/forensics_strict_edge_reward": 0.11652587354183197,
      "rewards/forensics_tiou_reward": 0.6407597064971924,
      "step": 1069
    },
    {
      "completion_length": 29.1875,
      "epoch": 3.282642089093702,
      "grad_norm": 9.26845645904541,
      "kl": 0.5947265625,
      "learning_rate": 1.776923076923077e-07,
      "loss": 0.1012,
      "reward": 0.7721708416938782,
      "reward_std": 0.04651814606040716,
      "reward_std_all": 0.043795693665742874,
      "rewards-max/forensics_sinkhorn_reward": 0.9829877614974976,
      "rewards-max/forensics_strict_edge_reward": 0.36225126683712006,
      "rewards-max/forensics_tiou_reward": 0.9600865840911865,
      "rewards/forensics_sinkhorn_reward": 0.7779169678688049,
      "rewards/forensics_strict_edge_reward": 0.05757945403456688,
      "rewards/forensics_tiou_reward": 0.5793627500534058,
      "step": 1070
    },
    {
      "completion_length": 27.578125,
      "epoch": 3.2857142857142856,
      "grad_norm": 9.633122444152832,
      "kl": 0.4453125,
      "learning_rate": 1.7692307692307693e-07,
      "loss": 0.0765,
      "reward": 0.7902079224586487,
      "reward_std": 0.020195297431200743,
      "reward_std_all": 0.02017670008353889,
      "rewards-max/forensics_sinkhorn_reward": 0.9911814332008362,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.983416348695755,
      "rewards/forensics_sinkhorn_reward": 0.7902039885520935,
      "rewards/forensics_strict_edge_reward": 0.08187959343194962,
      "rewards/forensics_tiou_reward": 0.6167282462120056,
      "step": 1071
    },
    {
      "completion_length": 25.203125,
      "epoch": 3.2887864823348694,
      "grad_norm": 7.372955322265625,
      "kl": 0.4423828125,
      "learning_rate": 1.7615384615384615e-07,
      "loss": 0.0382,
      "reward": 0.779734879732132,
      "reward_std": 0.021962295286357403,
      "reward_std_all": 0.026624776422977448,
      "rewards-max/forensics_sinkhorn_reward": 0.9848269820213318,
      "rewards-max/forensics_strict_edge_reward": 0.47726282477378845,
      "rewards-max/forensics_tiou_reward": 0.9671243727207184,
      "rewards/forensics_sinkhorn_reward": 0.7650039792060852,
      "rewards/forensics_strict_edge_reward": 0.06722372025251389,
      "rewards/forensics_tiou_reward": 0.5913327038288116,
      "step": 1072
    },
    {
      "completion_length": 26.953125,
      "epoch": 3.291858678955453,
      "grad_norm": 69.25697326660156,
      "kl": 3.6171875,
      "learning_rate": 1.7538461538461539e-07,
      "loss": 0.0949,
      "reward": 0.8056444525718689,
      "reward_std": 0.00293520325794816,
      "reward_std_all": 0.0043257789220660925,
      "rewards-max/forensics_sinkhorn_reward": 0.9886492490768433,
      "rewards-max/forensics_strict_edge_reward": 0.7421593070030212,
      "rewards-max/forensics_tiou_reward": 0.9776163995265961,
      "rewards/forensics_sinkhorn_reward": 0.8974408805370331,
      "rewards/forensics_strict_edge_reward": 0.09134914353489876,
      "rewards/forensics_tiou_reward": 0.78237184882164,
      "step": 1073
    },
    {
      "completion_length": 27.953125,
      "epoch": 3.294930875576037,
      "grad_norm": 8.456840515136719,
      "kl": 0.734375,
      "learning_rate": 1.746153846153846e-07,
      "loss": 0.0612,
      "reward": 0.7898341119289398,
      "reward_std": 0.023105221334844828,
      "reward_std_all": 0.02270533423870802,
      "rewards-max/forensics_sinkhorn_reward": 0.9872857332229614,
      "rewards-max/forensics_strict_edge_reward": 0.35669709742069244,
      "rewards-max/forensics_tiou_reward": 0.9693578779697418,
      "rewards/forensics_sinkhorn_reward": 0.806176632642746,
      "rewards/forensics_strict_edge_reward": 0.049113355576992035,
      "rewards/forensics_tiou_reward": 0.6373355090618134,
      "step": 1074
    },
    {
      "completion_length": 30.125,
      "epoch": 3.2980030721966207,
      "grad_norm": 8.37224006652832,
      "kl": 0.5078125,
      "learning_rate": 1.7384615384615384e-07,
      "loss": -0.0136,
      "reward": 0.7774884700775146,
      "reward_std": 0.03899433743208647,
      "reward_std_all": 0.03732973150908947,
      "rewards-max/forensics_sinkhorn_reward": 0.9786566197872162,
      "rewards-max/forensics_strict_edge_reward": 0.39551177620887756,
      "rewards-max/forensics_tiou_reward": 0.9544362127780914,
      "rewards/forensics_sinkhorn_reward": 0.6937329471111298,
      "rewards/forensics_strict_edge_reward": 0.06024149805307388,
      "rewards/forensics_tiou_reward": 0.5396005511283875,
      "step": 1075
    },
    {
      "completion_length": 22.84375,
      "epoch": 3.3010752688172045,
      "grad_norm": 8.56030559539795,
      "kl": 0.3837890625,
      "learning_rate": 1.7307692307692305e-07,
      "loss": 0.0157,
      "reward": 0.7921704053878784,
      "reward_std": 0.009961967123672366,
      "reward_std_all": 0.014688286930322647,
      "rewards-max/forensics_sinkhorn_reward": 0.9833047091960907,
      "rewards-max/forensics_strict_edge_reward": 0.6398303955793381,
      "rewards-max/forensics_tiou_reward": 0.9638921320438385,
      "rewards/forensics_sinkhorn_reward": 0.7950811386108398,
      "rewards/forensics_strict_edge_reward": 0.08110261708498001,
      "rewards/forensics_tiou_reward": 0.642160952091217,
      "step": 1076
    },
    {
      "completion_length": 22.875,
      "epoch": 3.3041474654377883,
      "grad_norm": 6.574634075164795,
      "kl": 0.466796875,
      "learning_rate": 1.723076923076923e-07,
      "loss": -0.031,
      "reward": 0.7643943727016449,
      "reward_std": 0.018812560476362705,
      "reward_std_all": 0.03165324218571186,
      "rewards-max/forensics_sinkhorn_reward": 0.9880736172199249,
      "rewards-max/forensics_strict_edge_reward": 0.5776711404323578,
      "rewards-max/forensics_tiou_reward": 0.973606139421463,
      "rewards/forensics_sinkhorn_reward": 0.8011943995952606,
      "rewards/forensics_strict_edge_reward": 0.07666691392660141,
      "rewards/forensics_tiou_reward": 0.603025883436203,
      "step": 1077
    },
    {
      "completion_length": 25.5625,
      "epoch": 3.3072196620583716,
      "grad_norm": 8.19904613494873,
      "kl": 0.431640625,
      "learning_rate": 1.7153846153846153e-07,
      "loss": -0.064,
      "reward": 0.7386575639247894,
      "reward_std": 0.08704955875873566,
      "reward_std_all": 0.08109144680202007,
      "rewards-max/forensics_sinkhorn_reward": 0.9927375614643097,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.9818525910377502,
      "rewards/forensics_sinkhorn_reward": 0.7618061900138855,
      "rewards/forensics_strict_edge_reward": 0.06741883605718613,
      "rewards/forensics_tiou_reward": 0.6261454522609711,
      "step": 1078
    },
    {
      "completion_length": 29.34375,
      "epoch": 3.3102918586789554,
      "grad_norm": 7.9241533279418945,
      "kl": 0.4169921875,
      "learning_rate": 1.7076923076923075e-07,
      "loss": 0.0323,
      "reward": 0.793099045753479,
      "reward_std": 0.021999423974193633,
      "reward_std_all": 0.021133945789188147,
      "rewards-max/forensics_sinkhorn_reward": 0.9924285709857941,
      "rewards-max/forensics_strict_edge_reward": 0.4462113678455353,
      "rewards-max/forensics_tiou_reward": 0.9856492280960083,
      "rewards/forensics_sinkhorn_reward": 0.7674096822738647,
      "rewards/forensics_strict_edge_reward": 0.06687301024794579,
      "rewards/forensics_tiou_reward": 0.6466571390628815,
      "step": 1079
    },
    {
      "completion_length": 22.625,
      "epoch": 3.313364055299539,
      "grad_norm": 9.771493911743164,
      "kl": 0.4296875,
      "learning_rate": 1.7000000000000001e-07,
      "loss": 0.0864,
      "reward": 0.7548881769180298,
      "reward_std": 0.037656369619071484,
      "reward_std_all": 0.044503542594611645,
      "rewards-max/forensics_sinkhorn_reward": 0.9924426376819611,
      "rewards-max/forensics_strict_edge_reward": 0.6540074050426483,
      "rewards-max/forensics_tiou_reward": 0.9850364923477173,
      "rewards/forensics_sinkhorn_reward": 0.7261826992034912,
      "rewards/forensics_strict_edge_reward": 0.056017402559518814,
      "rewards/forensics_tiou_reward": 0.5519436597824097,
      "step": 1080
    },
    {
      "completion_length": 28.328125,
      "epoch": 3.316436251920123,
      "grad_norm": 8.992802619934082,
      "kl": 0.4287109375,
      "learning_rate": 1.6923076923076923e-07,
      "loss": 0.0363,
      "reward": 0.7580351829528809,
      "reward_std": 0.02878668811172247,
      "reward_std_all": 0.038517991080880165,
      "rewards-max/forensics_sinkhorn_reward": 0.9762799739837646,
      "rewards-max/forensics_strict_edge_reward": 0.38380059599876404,
      "rewards-max/forensics_tiou_reward": 0.9454112946987152,
      "rewards/forensics_sinkhorn_reward": 0.7321815490722656,
      "rewards/forensics_strict_edge_reward": 0.0717119611799717,
      "rewards/forensics_tiou_reward": 0.5722134709358215,
      "step": 1081
    },
    {
      "completion_length": 25.953125,
      "epoch": 3.3195084485407067,
      "grad_norm": 11.51883602142334,
      "kl": 0.43359375,
      "learning_rate": 1.6846153846153847e-07,
      "loss": -0.0812,
      "reward": 0.7605903744697571,
      "reward_std": 0.021875786129385233,
      "reward_std_all": 0.03343642922118306,
      "rewards-max/forensics_sinkhorn_reward": 0.9844020009040833,
      "rewards-max/forensics_strict_edge_reward": 0.5595431923866272,
      "rewards-max/forensics_tiou_reward": 0.9611313939094543,
      "rewards/forensics_sinkhorn_reward": 0.737979382276535,
      "rewards/forensics_strict_edge_reward": 0.05375925451517105,
      "rewards/forensics_tiou_reward": 0.5952294766902924,
      "step": 1082
    },
    {
      "completion_length": 22.546875,
      "epoch": 3.3225806451612905,
      "grad_norm": 8.514543533325195,
      "kl": 0.34375,
      "learning_rate": 1.6769230769230768e-07,
      "loss": 0.0024,
      "reward": 0.7398945093154907,
      "reward_std": 0.04021686781197786,
      "reward_std_all": 0.05405120924115181,
      "rewards-max/forensics_sinkhorn_reward": 0.989675372838974,
      "rewards-max/forensics_strict_edge_reward": 0.5675406754016876,
      "rewards-max/forensics_tiou_reward": 0.9773009121417999,
      "rewards/forensics_sinkhorn_reward": 0.749515950679779,
      "rewards/forensics_strict_edge_reward": 0.07792562246322632,
      "rewards/forensics_tiou_reward": 0.5789288729429245,
      "step": 1083
    },
    {
      "completion_length": 29.328125,
      "epoch": 3.3256528417818743,
      "grad_norm": 8.304734230041504,
      "kl": 0.408203125,
      "learning_rate": 1.6692307692307692e-07,
      "loss": 0.0709,
      "reward": 0.7622956931591034,
      "reward_std": 0.03805242292582989,
      "reward_std_all": 0.04488877393305302,
      "rewards-max/forensics_sinkhorn_reward": 0.985608696937561,
      "rewards-max/forensics_strict_edge_reward": 0.3873155415058136,
      "rewards-max/forensics_tiou_reward": 0.9670977294445038,
      "rewards/forensics_sinkhorn_reward": 0.7530462443828583,
      "rewards/forensics_strict_edge_reward": 0.05213763564825058,
      "rewards/forensics_tiou_reward": 0.5572361350059509,
      "step": 1084
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.3287250384024576,
      "grad_norm": 11.948490142822266,
      "kl": 0.4345703125,
      "learning_rate": 1.6615384615384613e-07,
      "loss": 0.0922,
      "reward": 0.7808832824230194,
      "reward_std": 0.044045548886060715,
      "reward_std_all": 0.039729729294776917,
      "rewards-max/forensics_sinkhorn_reward": 0.9934375584125519,
      "rewards-max/forensics_strict_edge_reward": 0.40001481771469116,
      "rewards-max/forensics_tiou_reward": 0.9864121675491333,
      "rewards/forensics_sinkhorn_reward": 0.7799926996231079,
      "rewards/forensics_strict_edge_reward": 0.05454499553889036,
      "rewards/forensics_tiou_reward": 0.6412388980388641,
      "step": 1085
    },
    {
      "completion_length": 30.3125,
      "epoch": 3.3317972350230414,
      "grad_norm": 8.929883003234863,
      "kl": 0.634765625,
      "learning_rate": 1.6538461538461538e-07,
      "loss": -0.0548,
      "reward": 0.7840549945831299,
      "reward_std": 0.03250792529433966,
      "reward_std_all": 0.030625230632722378,
      "rewards-max/forensics_sinkhorn_reward": 0.9865650236606598,
      "rewards-max/forensics_strict_edge_reward": 0.47294433414936066,
      "rewards-max/forensics_tiou_reward": 0.969944566488266,
      "rewards/forensics_sinkhorn_reward": 0.8248119056224823,
      "rewards/forensics_strict_edge_reward": 0.056223105639219284,
      "rewards/forensics_tiou_reward": 0.6804829835891724,
      "step": 1086
    },
    {
      "completion_length": 27.671875,
      "epoch": 3.334869431643625,
      "grad_norm": 6.713168144226074,
      "kl": 0.4501953125,
      "learning_rate": 1.646153846153846e-07,
      "loss": -0.0632,
      "reward": 0.7565176784992218,
      "reward_std": 0.06044504884630442,
      "reward_std_all": 0.05770028568804264,
      "rewards-max/forensics_sinkhorn_reward": 0.992956131696701,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9874931275844574,
      "rewards/forensics_sinkhorn_reward": 0.7487687170505524,
      "rewards/forensics_strict_edge_reward": 0.08503212407231331,
      "rewards/forensics_tiou_reward": 0.5743398368358612,
      "step": 1087
    },
    {
      "completion_length": 23.8125,
      "epoch": 3.337941628264209,
      "grad_norm": 7.988343715667725,
      "kl": 0.3984375,
      "learning_rate": 1.6384615384615383e-07,
      "loss": -0.0798,
      "reward": 0.7937840521335602,
      "reward_std": 0.015393548994325101,
      "reward_std_all": 0.016140258288942277,
      "rewards-max/forensics_sinkhorn_reward": 0.9933119118213654,
      "rewards-max/forensics_strict_edge_reward": 0.7292781174182892,
      "rewards-max/forensics_tiou_reward": 0.9858996868133545,
      "rewards/forensics_sinkhorn_reward": 0.7931574881076813,
      "rewards/forensics_strict_edge_reward": 0.0856020413339138,
      "rewards/forensics_tiou_reward": 0.6264292299747467,
      "step": 1088
    },
    {
      "completion_length": 23.671875,
      "epoch": 3.3410138248847927,
      "grad_norm": 13.866528511047363,
      "kl": 0.4482421875,
      "learning_rate": 1.6307692307692307e-07,
      "loss": -0.0425,
      "reward": 0.8044065535068512,
      "reward_std": 0.004652605624869466,
      "reward_std_all": 0.005699865520000458,
      "rewards-max/forensics_sinkhorn_reward": 0.9945617914199829,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9905121922492981,
      "rewards/forensics_sinkhorn_reward": 0.850996345281601,
      "rewards/forensics_strict_edge_reward": 0.08794458955526352,
      "rewards/forensics_tiou_reward": 0.7381398677825928,
      "step": 1089
    },
    {
      "completion_length": 28.609375,
      "epoch": 3.3440860215053765,
      "grad_norm": 8.338822364807129,
      "kl": 0.357421875,
      "learning_rate": 1.623076923076923e-07,
      "loss": -0.062,
      "reward": 0.8009228408336639,
      "reward_std": 0.011419119080528617,
      "reward_std_all": 0.011240103049203753,
      "rewards-max/forensics_sinkhorn_reward": 0.9941853284835815,
      "rewards-max/forensics_strict_edge_reward": 0.663480818271637,
      "rewards-max/forensics_tiou_reward": 0.9900148808956146,
      "rewards/forensics_sinkhorn_reward": 0.8606568872928619,
      "rewards/forensics_strict_edge_reward": 0.118174709379673,
      "rewards/forensics_tiou_reward": 0.7276288866996765,
      "step": 1090
    },
    {
      "completion_length": 23.828125,
      "epoch": 3.3471582181259603,
      "grad_norm": 9.840143203735352,
      "kl": 0.546875,
      "learning_rate": 1.6153846153846155e-07,
      "loss": 0.0066,
      "reward": 0.7967840433120728,
      "reward_std": 0.017312193987891078,
      "reward_std_all": 0.01665100804530084,
      "rewards-max/forensics_sinkhorn_reward": 0.9895962774753571,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.9808943569660187,
      "rewards/forensics_sinkhorn_reward": 0.8557414412498474,
      "rewards/forensics_strict_edge_reward": 0.07642491906881332,
      "rewards/forensics_tiou_reward": 0.6802087724208832,
      "step": 1091
    },
    {
      "completion_length": 23.1875,
      "epoch": 3.3502304147465436,
      "grad_norm": 6.691363334655762,
      "kl": 0.515625,
      "learning_rate": 1.6076923076923076e-07,
      "loss": -0.0117,
      "reward": 0.7850500643253326,
      "reward_std": 0.02313167741522193,
      "reward_std_all": 0.023823609575629234,
      "rewards-max/forensics_sinkhorn_reward": 0.9893096685409546,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9800632894039154,
      "rewards/forensics_sinkhorn_reward": 0.8119780719280243,
      "rewards/forensics_strict_edge_reward": 0.08670113235712051,
      "rewards/forensics_tiou_reward": 0.6323740184307098,
      "step": 1092
    },
    {
      "completion_length": 27.828125,
      "epoch": 3.3533026113671274,
      "grad_norm": 9.629284858703613,
      "kl": 0.513671875,
      "learning_rate": 1.6e-07,
      "loss": 0.0642,
      "reward": 0.7503706812858582,
      "reward_std": 0.08025797829031944,
      "reward_std_all": 0.07316887378692627,
      "rewards-max/forensics_sinkhorn_reward": 0.9853541254997253,
      "rewards-max/forensics_strict_edge_reward": 0.6762872636318207,
      "rewards-max/forensics_tiou_reward": 0.9689873456954956,
      "rewards/forensics_sinkhorn_reward": 0.7378414869308472,
      "rewards/forensics_strict_edge_reward": 0.09734094515442848,
      "rewards/forensics_tiou_reward": 0.6186569333076477,
      "step": 1093
    },
    {
      "completion_length": 26.390625,
      "epoch": 3.356374807987711,
      "grad_norm": 6.861050128936768,
      "kl": 0.552734375,
      "learning_rate": 1.5923076923076922e-07,
      "loss": 0.115,
      "reward": 0.7564835548400879,
      "reward_std": 0.05773073062300682,
      "reward_std_all": 0.05722612328827381,
      "rewards-max/forensics_sinkhorn_reward": 0.9933605790138245,
      "rewards-max/forensics_strict_edge_reward": 0.5787447243928909,
      "rewards-max/forensics_tiou_reward": 0.9878523349761963,
      "rewards/forensics_sinkhorn_reward": 0.7444321811199188,
      "rewards/forensics_strict_edge_reward": 0.07475166767835617,
      "rewards/forensics_tiou_reward": 0.5715235769748688,
      "step": 1094
    },
    {
      "completion_length": 25.390625,
      "epoch": 3.359447004608295,
      "grad_norm": 7.9626288414001465,
      "kl": 0.40234375,
      "learning_rate": 1.5846153846153846e-07,
      "loss": 0.017,
      "reward": 0.7852164804935455,
      "reward_std": 0.027113661635667086,
      "reward_std_all": 0.02663758583366871,
      "rewards-max/forensics_sinkhorn_reward": 0.9964086413383484,
      "rewards-max/forensics_strict_edge_reward": 0.6215263903141022,
      "rewards-max/forensics_tiou_reward": 0.9960500299930573,
      "rewards/forensics_sinkhorn_reward": 0.8068370819091797,
      "rewards/forensics_strict_edge_reward": 0.0682503879070282,
      "rewards/forensics_tiou_reward": 0.6421956121921539,
      "step": 1095
    },
    {
      "completion_length": 28.21875,
      "epoch": 3.3625192012288787,
      "grad_norm": 6.956463813781738,
      "kl": 0.34765625,
      "learning_rate": 1.5769230769230767e-07,
      "loss": 0.051,
      "reward": 0.7901366949081421,
      "reward_std": 0.01731564709916711,
      "reward_std_all": 0.01842503249645233,
      "rewards-max/forensics_sinkhorn_reward": 0.9864631593227386,
      "rewards-max/forensics_strict_edge_reward": 0.5060584545135498,
      "rewards-max/forensics_tiou_reward": 0.9689970910549164,
      "rewards/forensics_sinkhorn_reward": 0.7630384564399719,
      "rewards/forensics_strict_edge_reward": 0.058561019599437714,
      "rewards/forensics_tiou_reward": 0.6295673251152039,
      "step": 1096
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.3655913978494625,
      "grad_norm": 8.148737907409668,
      "kl": 0.3779296875,
      "learning_rate": 1.569230769230769e-07,
      "loss": 0.0207,
      "reward": 0.7810601592063904,
      "reward_std": 0.039004402118735015,
      "reward_std_all": 0.036080574383959174,
      "rewards-max/forensics_sinkhorn_reward": 0.9948690235614777,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9912803173065186,
      "rewards/forensics_sinkhorn_reward": 0.8103465437889099,
      "rewards/forensics_strict_edge_reward": 0.1331135258078575,
      "rewards/forensics_tiou_reward": 0.6664192378520966,
      "step": 1097
    },
    {
      "completion_length": 27.28125,
      "epoch": 3.3686635944700463,
      "grad_norm": 8.947790145874023,
      "kl": 0.501953125,
      "learning_rate": 1.5615384615384612e-07,
      "loss": -0.1179,
      "reward": 0.7787047922611237,
      "reward_std": 0.03103689756244421,
      "reward_std_all": 0.031092561781406403,
      "rewards-max/forensics_sinkhorn_reward": 0.9893652200698853,
      "rewards-max/forensics_strict_edge_reward": 0.5475890934467316,
      "rewards-max/forensics_tiou_reward": 0.9806084036827087,
      "rewards/forensics_sinkhorn_reward": 0.7952981889247894,
      "rewards/forensics_strict_edge_reward": 0.057704782113432884,
      "rewards/forensics_tiou_reward": 0.6645872592926025,
      "step": 1098
    },
    {
      "completion_length": 22.03125,
      "epoch": 3.3717357910906296,
      "grad_norm": 8.13018798828125,
      "kl": 0.388671875,
      "learning_rate": 1.5538461538461536e-07,
      "loss": -0.1128,
      "reward": 0.8047339022159576,
      "reward_std": 0.0018142957706004381,
      "reward_std_all": 0.003926790552213788,
      "rewards-max/forensics_sinkhorn_reward": 0.9924096167087555,
      "rewards-max/forensics_strict_edge_reward": 0.7004553079605103,
      "rewards-max/forensics_tiou_reward": 0.9815574288368225,
      "rewards/forensics_sinkhorn_reward": 0.8512057363986969,
      "rewards/forensics_strict_edge_reward": 0.0639827623963356,
      "rewards/forensics_tiou_reward": 0.6581927537918091,
      "step": 1099
    },
    {
      "completion_length": 24.140625,
      "epoch": 3.3748079877112134,
      "grad_norm": 7.850218296051025,
      "kl": 0.4423828125,
      "learning_rate": 1.5461538461538463e-07,
      "loss": 0.1614,
      "reward": 0.7636809647083282,
      "reward_std": 0.0635006669908762,
      "reward_std_all": 0.05768118426203728,
      "rewards-max/forensics_sinkhorn_reward": 0.9967075288295746,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.997629851102829,
      "rewards/forensics_sinkhorn_reward": 0.7162302434444427,
      "rewards/forensics_strict_edge_reward": 0.09166135638952255,
      "rewards/forensics_tiou_reward": 0.5606272518634796,
      "step": 1100
    },
    {
      "completion_length": 23.265625,
      "epoch": 3.377880184331797,
      "grad_norm": 8.185954093933105,
      "kl": 0.4619140625,
      "learning_rate": 1.5384615384615385e-07,
      "loss": 0.0562,
      "reward": 0.7871285676956177,
      "reward_std": 0.0253401268273592,
      "reward_std_all": 0.024936790578067303,
      "rewards-max/forensics_sinkhorn_reward": 0.9924351274967194,
      "rewards-max/forensics_strict_edge_reward": 0.476627953350544,
      "rewards-max/forensics_tiou_reward": 0.9826513826847076,
      "rewards/forensics_sinkhorn_reward": 0.776526153087616,
      "rewards/forensics_strict_edge_reward": 0.08930057659745216,
      "rewards/forensics_tiou_reward": 0.6096388101577759,
      "step": 1101
    },
    {
      "completion_length": 22.9375,
      "epoch": 3.380952380952381,
      "grad_norm": 9.252159118652344,
      "kl": 0.431640625,
      "learning_rate": 1.5307692307692309e-07,
      "loss": 0.0819,
      "reward": 0.7606096863746643,
      "reward_std": 0.0354679049924016,
      "reward_std_all": 0.043031640350818634,
      "rewards-max/forensics_sinkhorn_reward": 0.9598517715930939,
      "rewards-max/forensics_strict_edge_reward": 0.31992475129663944,
      "rewards-max/forensics_tiou_reward": 0.8931084275245667,
      "rewards/forensics_sinkhorn_reward": 0.7171742022037506,
      "rewards/forensics_strict_edge_reward": 0.03238247742410749,
      "rewards/forensics_tiou_reward": 0.4856417626142502,
      "step": 1102
    },
    {
      "completion_length": 26.609375,
      "epoch": 3.3840245775729647,
      "grad_norm": 7.221450328826904,
      "kl": 0.419921875,
      "learning_rate": 1.523076923076923e-07,
      "loss": 0.0092,
      "reward": 0.7892346680164337,
      "reward_std": 0.02720430912449956,
      "reward_std_all": 0.02535225637257099,
      "rewards-max/forensics_sinkhorn_reward": 0.9914194047451019,
      "rewards-max/forensics_strict_edge_reward": 0.784106969833374,
      "rewards-max/forensics_tiou_reward": 0.984106034040451,
      "rewards/forensics_sinkhorn_reward": 0.8443799614906311,
      "rewards/forensics_strict_edge_reward": 0.14104197174310684,
      "rewards/forensics_tiou_reward": 0.7117729187011719,
      "step": 1103
    },
    {
      "completion_length": 26.5,
      "epoch": 3.3870967741935485,
      "grad_norm": 7.9405975341796875,
      "kl": 0.625,
      "learning_rate": 1.5153846153846154e-07,
      "loss": 0.0558,
      "reward": 0.7592863738536835,
      "reward_std": 0.03104826621711254,
      "reward_std_all": 0.042551870457828045,
      "rewards-max/forensics_sinkhorn_reward": 0.9906629323959351,
      "rewards-max/forensics_strict_edge_reward": 0.7335375547409058,
      "rewards-max/forensics_tiou_reward": 0.9806074798107147,
      "rewards/forensics_sinkhorn_reward": 0.795221596956253,
      "rewards/forensics_strict_edge_reward": 0.08623328804969788,
      "rewards/forensics_tiou_reward": 0.6214011609554291,
      "step": 1104
    },
    {
      "completion_length": 25.59375,
      "epoch": 3.3901689708141323,
      "grad_norm": 7.707510948181152,
      "kl": 0.4208984375,
      "learning_rate": 1.5076923076923075e-07,
      "loss": -0.034,
      "reward": 0.7984350323677063,
      "reward_std": 0.012614715844392776,
      "reward_std_all": 0.013113278895616531,
      "rewards-max/forensics_sinkhorn_reward": 0.9843058586120605,
      "rewards-max/forensics_strict_edge_reward": 0.5019287317991257,
      "rewards-max/forensics_tiou_reward": 0.9621241092681885,
      "rewards/forensics_sinkhorn_reward": 0.8036203980445862,
      "rewards/forensics_strict_edge_reward": 0.07056773453950882,
      "rewards/forensics_tiou_reward": 0.5941226780414581,
      "step": 1105
    },
    {
      "completion_length": 26.578125,
      "epoch": 3.3932411674347156,
      "grad_norm": 6.857783317565918,
      "kl": 0.470703125,
      "learning_rate": 1.5e-07,
      "loss": -0.0054,
      "reward": 0.7827543318271637,
      "reward_std": 0.039904242381453514,
      "reward_std_all": 0.036576706916093826,
      "rewards-max/forensics_sinkhorn_reward": 0.9852939546108246,
      "rewards-max/forensics_strict_edge_reward": 0.5504324436187744,
      "rewards-max/forensics_tiou_reward": 0.9667387008666992,
      "rewards/forensics_sinkhorn_reward": 0.803038626909256,
      "rewards/forensics_strict_edge_reward": 0.06379745900630951,
      "rewards/forensics_tiou_reward": 0.6230219304561615,
      "step": 1106
    },
    {
      "completion_length": 21.703125,
      "epoch": 3.3963133640552994,
      "grad_norm": 8.974260330200195,
      "kl": 0.4541015625,
      "learning_rate": 1.492307692307692e-07,
      "loss": -0.0105,
      "reward": 0.7648390531539917,
      "reward_std": 0.040388076566159725,
      "reward_std_all": 0.04165405849926174,
      "rewards-max/forensics_sinkhorn_reward": 0.9932601749897003,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9878800511360168,
      "rewards/forensics_sinkhorn_reward": 0.809389591217041,
      "rewards/forensics_strict_edge_reward": 0.06733800284564495,
      "rewards/forensics_tiou_reward": 0.6339763104915619,
      "step": 1107
    },
    {
      "completion_length": 24.15625,
      "epoch": 3.399385560675883,
      "grad_norm": 9.192548751831055,
      "kl": 0.3828125,
      "learning_rate": 1.4846153846153845e-07,
      "loss": -0.0427,
      "reward": 0.7982712388038635,
      "reward_std": 0.014818658004514873,
      "reward_std_all": 0.014451855211518705,
      "rewards-max/forensics_sinkhorn_reward": 0.9973844885826111,
      "rewards-max/forensics_strict_edge_reward": 0.8683927059173584,
      "rewards-max/forensics_tiou_reward": 0.9986400008201599,
      "rewards/forensics_sinkhorn_reward": 0.8322708010673523,
      "rewards/forensics_strict_edge_reward": 0.10088291019201279,
      "rewards/forensics_tiou_reward": 0.7072187066078186,
      "step": 1108
    },
    {
      "completion_length": 26.484375,
      "epoch": 3.402457757296467,
      "grad_norm": 7.680929183959961,
      "kl": 0.5400390625,
      "learning_rate": 1.476923076923077e-07,
      "loss": -0.0298,
      "reward": 0.802320271730423,
      "reward_std": 0.007205248926766217,
      "reward_std_all": 0.008238040725700557,
      "rewards-max/forensics_sinkhorn_reward": 0.9856469631195068,
      "rewards-max/forensics_strict_edge_reward": 0.5684361159801483,
      "rewards-max/forensics_tiou_reward": 0.9694900214672089,
      "rewards/forensics_sinkhorn_reward": 0.8278471529483795,
      "rewards/forensics_strict_edge_reward": 0.07864643074572086,
      "rewards/forensics_tiou_reward": 0.6753920018672943,
      "step": 1109
    },
    {
      "completion_length": 22.34375,
      "epoch": 3.4055299539170507,
      "grad_norm": 8.289118766784668,
      "kl": 0.40625,
      "learning_rate": 1.4692307692307693e-07,
      "loss": -0.0149,
      "reward": 0.7951543927192688,
      "reward_std": 0.012257817201316357,
      "reward_std_all": 0.0135529525578022,
      "rewards-max/forensics_sinkhorn_reward": 0.9908548593521118,
      "rewards-max/forensics_strict_edge_reward": 0.6350628435611725,
      "rewards-max/forensics_tiou_reward": 0.9800599813461304,
      "rewards/forensics_sinkhorn_reward": 0.7988518178462982,
      "rewards/forensics_strict_edge_reward": 0.08873161487281322,
      "rewards/forensics_tiou_reward": 0.5952627658843994,
      "step": 1110
    },
    {
      "completion_length": 23.84375,
      "epoch": 3.4086021505376345,
      "grad_norm": 9.528609275817871,
      "kl": 0.38671875,
      "learning_rate": 1.4615384615384617e-07,
      "loss": -0.0999,
      "reward": 0.8009613156318665,
      "reward_std": 0.008220568648539484,
      "reward_std_all": 0.009473632089793682,
      "rewards-max/forensics_sinkhorn_reward": 0.9964370727539062,
      "rewards-max/forensics_strict_edge_reward": 0.7513054013252258,
      "rewards-max/forensics_tiou_reward": 0.9958148300647736,
      "rewards/forensics_sinkhorn_reward": 0.7866269946098328,
      "rewards/forensics_strict_edge_reward": 0.0645233802497387,
      "rewards/forensics_tiou_reward": 0.6116162836551666,
      "step": 1111
    },
    {
      "completion_length": 24.5,
      "epoch": 3.4116743471582183,
      "grad_norm": 7.872464179992676,
      "kl": 0.408203125,
      "learning_rate": 1.4538461538461538e-07,
      "loss": -0.0127,
      "reward": 0.7923924028873444,
      "reward_std": 0.024458322674036026,
      "reward_std_all": 0.023158671334385872,
      "rewards-max/forensics_sinkhorn_reward": 0.9813950955867767,
      "rewards-max/forensics_strict_edge_reward": 0.37908462435007095,
      "rewards-max/forensics_tiou_reward": 0.9555761516094208,
      "rewards/forensics_sinkhorn_reward": 0.8166280388832092,
      "rewards/forensics_strict_edge_reward": 0.06713112723082304,
      "rewards/forensics_tiou_reward": 0.6879339814186096,
      "step": 1112
    },
    {
      "completion_length": 24.859375,
      "epoch": 3.4147465437788016,
      "grad_norm": 6.6710405349731445,
      "kl": 0.4931640625,
      "learning_rate": 1.4461538461538462e-07,
      "loss": -0.0046,
      "reward": 0.7651699781417847,
      "reward_std": 0.013950809894595295,
      "reward_std_all": 0.028759855777025223,
      "rewards-max/forensics_sinkhorn_reward": 0.9914788603782654,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9837238192558289,
      "rewards/forensics_sinkhorn_reward": 0.8276321291923523,
      "rewards/forensics_strict_edge_reward": 0.09554711543023586,
      "rewards/forensics_tiou_reward": 0.6682678461074829,
      "step": 1113
    },
    {
      "completion_length": 25.109375,
      "epoch": 3.4178187403993854,
      "grad_norm": 8.704486846923828,
      "kl": 0.3564453125,
      "learning_rate": 1.4384615384615384e-07,
      "loss": 0.0286,
      "reward": 0.7926593124866486,
      "reward_std": 0.013716128189116716,
      "reward_std_all": 0.015233521349728107,
      "rewards-max/forensics_sinkhorn_reward": 0.9854435324668884,
      "rewards-max/forensics_strict_edge_reward": 0.3822198510169983,
      "rewards-max/forensics_tiou_reward": 0.9665828049182892,
      "rewards/forensics_sinkhorn_reward": 0.8061532974243164,
      "rewards/forensics_strict_edge_reward": 0.03865431807935238,
      "rewards/forensics_tiou_reward": 0.6191215515136719,
      "step": 1114
    },
    {
      "completion_length": 28.578125,
      "epoch": 3.420890937019969,
      "grad_norm": 7.77090311050415,
      "kl": 0.4853515625,
      "learning_rate": 1.4307692307692308e-07,
      "loss": -0.0413,
      "reward": 0.8054200410842896,
      "reward_std": 0.0022417795844376087,
      "reward_std_all": 0.003803479252383113,
      "rewards-max/forensics_sinkhorn_reward": 0.9934550523757935,
      "rewards-max/forensics_strict_edge_reward": 0.6712121218442917,
      "rewards-max/forensics_tiou_reward": 0.9892013370990753,
      "rewards/forensics_sinkhorn_reward": 0.8352220952510834,
      "rewards/forensics_strict_edge_reward": 0.1095220036804676,
      "rewards/forensics_tiou_reward": 0.6980368196964264,
      "step": 1115
    },
    {
      "completion_length": 23.890625,
      "epoch": 3.423963133640553,
      "grad_norm": 7.034942150115967,
      "kl": 0.4375,
      "learning_rate": 1.423076923076923e-07,
      "loss": 0.0509,
      "reward": 0.7791152596473694,
      "reward_std": 0.02176364162005484,
      "reward_std_all": 0.025517484871670604,
      "rewards-max/forensics_sinkhorn_reward": 0.9905494749546051,
      "rewards-max/forensics_strict_edge_reward": 0.5289284288883209,
      "rewards-max/forensics_tiou_reward": 0.9795706570148468,
      "rewards/forensics_sinkhorn_reward": 0.7735245525836945,
      "rewards/forensics_strict_edge_reward": 0.046041738241910934,
      "rewards/forensics_tiou_reward": 0.5663398206233978,
      "step": 1116
    },
    {
      "completion_length": 25.1875,
      "epoch": 3.4270353302611367,
      "grad_norm": 8.329172134399414,
      "kl": 0.45703125,
      "learning_rate": 1.4153846153846153e-07,
      "loss": -0.0633,
      "reward": 0.782946914434433,
      "reward_std": 0.02168169105425477,
      "reward_std_all": 0.023055811412632465,
      "rewards-max/forensics_sinkhorn_reward": 0.9880597591400146,
      "rewards-max/forensics_strict_edge_reward": 0.5227911174297333,
      "rewards-max/forensics_tiou_reward": 0.9742622673511505,
      "rewards/forensics_sinkhorn_reward": 0.7851434051990509,
      "rewards/forensics_strict_edge_reward": 0.09537636488676071,
      "rewards/forensics_tiou_reward": 0.5782394707202911,
      "step": 1117
    },
    {
      "completion_length": 27.90625,
      "epoch": 3.4301075268817205,
      "grad_norm": 8.163063049316406,
      "kl": 0.4267578125,
      "learning_rate": 1.4076923076923074e-07,
      "loss": -0.0739,
      "reward": 0.787116289138794,
      "reward_std": 0.0015853287186473608,
      "reward_std_all": 0.012604403542354703,
      "rewards-max/forensics_sinkhorn_reward": 0.9893041253089905,
      "rewards-max/forensics_strict_edge_reward": 0.6293683648109436,
      "rewards-max/forensics_tiou_reward": 0.9782234132289886,
      "rewards/forensics_sinkhorn_reward": 0.8231620490550995,
      "rewards/forensics_strict_edge_reward": 0.08299234509468079,
      "rewards/forensics_tiou_reward": 0.6890841126441956,
      "step": 1118
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.4331797235023043,
      "grad_norm": 8.348344802856445,
      "kl": 0.41796875,
      "learning_rate": 1.4e-07,
      "loss": -0.0425,
      "reward": 0.7673836052417755,
      "reward_std": 0.046221038326621056,
      "reward_std_all": 0.044799549505114555,
      "rewards-max/forensics_sinkhorn_reward": 0.9868068397045135,
      "rewards-max/forensics_strict_edge_reward": 0.5755647718906403,
      "rewards-max/forensics_tiou_reward": 0.9705443680286407,
      "rewards/forensics_sinkhorn_reward": 0.7463714778423309,
      "rewards/forensics_strict_edge_reward": 0.039337120950222015,
      "rewards/forensics_tiou_reward": 0.5267618596553802,
      "step": 1119
    },
    {
      "completion_length": 24.03125,
      "epoch": 3.436251920122888,
      "grad_norm": 7.879317760467529,
      "kl": 0.4013671875,
      "learning_rate": 1.3923076923076922e-07,
      "loss": 0.0452,
      "reward": 0.7939440906047821,
      "reward_std": 0.020724066998809576,
      "reward_std_all": 0.020142854191362858,
      "rewards-max/forensics_sinkhorn_reward": 0.9875826835632324,
      "rewards-max/forensics_strict_edge_reward": 0.702418714761734,
      "rewards-max/forensics_tiou_reward": 0.9668895602226257,
      "rewards/forensics_sinkhorn_reward": 0.8104462027549744,
      "rewards/forensics_strict_edge_reward": 0.061603402718901634,
      "rewards/forensics_tiou_reward": 0.6124120354652405,
      "step": 1120
    },
    {
      "completion_length": 24.859375,
      "epoch": 3.439324116743472,
      "grad_norm": 8.031440734863281,
      "kl": 0.55078125,
      "learning_rate": 1.3846153846153846e-07,
      "loss": 0.0261,
      "reward": 0.7765249311923981,
      "reward_std": 0.01771678775548935,
      "reward_std_all": 0.023432331858202815,
      "rewards-max/forensics_sinkhorn_reward": 0.9749414324760437,
      "rewards-max/forensics_strict_edge_reward": 0.4419972822070122,
      "rewards-max/forensics_tiou_reward": 0.9411161541938782,
      "rewards/forensics_sinkhorn_reward": 0.7795314788818359,
      "rewards/forensics_strict_edge_reward": 0.04940391331911087,
      "rewards/forensics_tiou_reward": 0.6084334850311279,
      "step": 1121
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.442396313364055,
      "grad_norm": 8.569167137145996,
      "kl": 0.3388671875,
      "learning_rate": 1.376923076923077e-07,
      "loss": 0.0205,
      "reward": 0.7704135179519653,
      "reward_std": 0.02069495152682066,
      "reward_std_all": 0.030371317639946938,
      "rewards-max/forensics_sinkhorn_reward": 0.9878958761692047,
      "rewards-max/forensics_strict_edge_reward": 0.581572100520134,
      "rewards-max/forensics_tiou_reward": 0.9749914705753326,
      "rewards/forensics_sinkhorn_reward": 0.8139127194881439,
      "rewards/forensics_strict_edge_reward": 0.0952078141272068,
      "rewards/forensics_tiou_reward": 0.7071995735168457,
      "step": 1122
    },
    {
      "completion_length": 26.359375,
      "epoch": 3.445468509984639,
      "grad_norm": 7.516648769378662,
      "kl": 0.4326171875,
      "learning_rate": 1.3692307692307692e-07,
      "loss": -0.0301,
      "reward": 0.7925404906272888,
      "reward_std": 0.017749551683664322,
      "reward_std_all": 0.018316288478672504,
      "rewards-max/forensics_sinkhorn_reward": 0.9927717447280884,
      "rewards-max/forensics_strict_edge_reward": 0.5770428776741028,
      "rewards-max/forensics_tiou_reward": 0.9851188659667969,
      "rewards/forensics_sinkhorn_reward": 0.7676661014556885,
      "rewards/forensics_strict_edge_reward": 0.07676218822598457,
      "rewards/forensics_tiou_reward": 0.615290492773056,
      "step": 1123
    },
    {
      "completion_length": 26.1875,
      "epoch": 3.4485407066052227,
      "grad_norm": 8.323768615722656,
      "kl": 0.4150390625,
      "learning_rate": 1.3615384615384616e-07,
      "loss": -0.0431,
      "reward": 0.782441109418869,
      "reward_std": 0.0197512274608016,
      "reward_std_all": 0.022522523533552885,
      "rewards-max/forensics_sinkhorn_reward": 0.9900469183921814,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9794794023036957,
      "rewards/forensics_sinkhorn_reward": 0.8205967247486115,
      "rewards/forensics_strict_edge_reward": 0.07645111531019211,
      "rewards/forensics_tiou_reward": 0.6558985412120819,
      "step": 1124
    },
    {
      "completion_length": 23.1875,
      "epoch": 3.4516129032258065,
      "grad_norm": 8.238324165344238,
      "kl": 0.40625,
      "learning_rate": 1.3538461538461537e-07,
      "loss": -0.021,
      "reward": 0.7908506989479065,
      "reward_std": 0.027659797109663486,
      "reward_std_all": 0.02571201976388693,
      "rewards-max/forensics_sinkhorn_reward": 0.9780865907669067,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9519418478012085,
      "rewards/forensics_sinkhorn_reward": 0.8533013463020325,
      "rewards/forensics_strict_edge_reward": 0.06309754401445389,
      "rewards/forensics_tiou_reward": 0.6953431665897369,
      "step": 1125
    },
    {
      "completion_length": 25.484375,
      "epoch": 3.4546850998463903,
      "grad_norm": 8.889511108398438,
      "kl": 0.4609375,
      "learning_rate": 1.346153846153846e-07,
      "loss": -0.0065,
      "reward": 0.7979070246219635,
      "reward_std": 0.01587723009288311,
      "reward_std_all": 0.015412768349051476,
      "rewards-max/forensics_sinkhorn_reward": 0.9963768124580383,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9951414167881012,
      "rewards/forensics_sinkhorn_reward": 0.8413996696472168,
      "rewards/forensics_strict_edge_reward": 0.11698826402425766,
      "rewards/forensics_tiou_reward": 0.6999365091323853,
      "step": 1126
    },
    {
      "completion_length": 26.625,
      "epoch": 3.457757296466974,
      "grad_norm": 8.408763885498047,
      "kl": 0.5615234375,
      "learning_rate": 1.3384615384615383e-07,
      "loss": 0.1072,
      "reward": 0.7760159373283386,
      "reward_std": 0.037501854822039604,
      "reward_std_all": 0.03626501187682152,
      "rewards-max/forensics_sinkhorn_reward": 0.9899296462535858,
      "rewards-max/forensics_strict_edge_reward": 0.6568913459777832,
      "rewards-max/forensics_tiou_reward": 0.9797503352165222,
      "rewards/forensics_sinkhorn_reward": 0.7485355734825134,
      "rewards/forensics_strict_edge_reward": 0.0694685373455286,
      "rewards/forensics_tiou_reward": 0.5757424831390381,
      "step": 1127
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.460829493087558,
      "grad_norm": 8.73733139038086,
      "kl": 0.685546875,
      "learning_rate": 1.3307692307692307e-07,
      "loss": 0.0511,
      "reward": 0.7844443917274475,
      "reward_std": 0.0376122510060668,
      "reward_std_all": 0.03445722348988056,
      "rewards-max/forensics_sinkhorn_reward": 0.989098459482193,
      "rewards-max/forensics_strict_edge_reward": 0.4468829333782196,
      "rewards-max/forensics_tiou_reward": 0.9746043086051941,
      "rewards/forensics_sinkhorn_reward": 0.7673924267292023,
      "rewards/forensics_strict_edge_reward": 0.05133785307407379,
      "rewards/forensics_tiou_reward": 0.6469413638114929,
      "step": 1128
    },
    {
      "completion_length": 23.15625,
      "epoch": 3.463901689708141,
      "grad_norm": 8.418989181518555,
      "kl": 0.361328125,
      "learning_rate": 1.323076923076923e-07,
      "loss": -0.0699,
      "reward": 0.7985966503620148,
      "reward_std": 0.014861484640277922,
      "reward_std_all": 0.014604117488488555,
      "rewards-max/forensics_sinkhorn_reward": 0.9863660633563995,
      "rewards-max/forensics_strict_edge_reward": 0.5067107081413269,
      "rewards-max/forensics_tiou_reward": 0.9657301306724548,
      "rewards/forensics_sinkhorn_reward": 0.872837632894516,
      "rewards/forensics_strict_edge_reward": 0.07511766254901886,
      "rewards/forensics_tiou_reward": 0.7458710968494415,
      "step": 1129
    },
    {
      "completion_length": 27.765625,
      "epoch": 3.466973886328725,
      "grad_norm": 7.9461894035339355,
      "kl": 0.4228515625,
      "learning_rate": 1.3153846153846155e-07,
      "loss": 0.0642,
      "reward": 0.7837408185005188,
      "reward_std": 0.03865964896976948,
      "reward_std_all": 0.03497983142733574,
      "rewards-max/forensics_sinkhorn_reward": 0.994858056306839,
      "rewards-max/forensics_strict_edge_reward": 0.6657527685165405,
      "rewards-max/forensics_tiou_reward": 0.9908345639705658,
      "rewards/forensics_sinkhorn_reward": 0.7990245819091797,
      "rewards/forensics_strict_edge_reward": 0.07835564389824867,
      "rewards/forensics_tiou_reward": 0.6849492490291595,
      "step": 1130
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.4700460829493087,
      "grad_norm": 10.53582763671875,
      "kl": 0.5419921875,
      "learning_rate": 1.3076923076923076e-07,
      "loss": -0.0778,
      "reward": 0.7870196104049683,
      "reward_std": 0.01841866225004196,
      "reward_std_all": 0.019907933194190264,
      "rewards-max/forensics_sinkhorn_reward": 0.9934196174144745,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9862222075462341,
      "rewards/forensics_sinkhorn_reward": 0.828908771276474,
      "rewards/forensics_strict_edge_reward": 0.0870414562523365,
      "rewards/forensics_tiou_reward": 0.649721622467041,
      "step": 1131
    },
    {
      "completion_length": 25.890625,
      "epoch": 3.4731182795698925,
      "grad_norm": 11.682414054870605,
      "kl": 0.4990234375,
      "learning_rate": 1.3e-07,
      "loss": -0.0109,
      "reward": 0.778573215007782,
      "reward_std": 0.03446589264785871,
      "reward_std_all": 0.03366163210012019,
      "rewards-max/forensics_sinkhorn_reward": 0.9898640513420105,
      "rewards-max/forensics_strict_edge_reward": 0.3799305483698845,
      "rewards-max/forensics_tiou_reward": 0.9795471131801605,
      "rewards/forensics_sinkhorn_reward": 0.7515264451503754,
      "rewards/forensics_strict_edge_reward": 0.04063994251191616,
      "rewards/forensics_tiou_reward": 0.578070268034935,
      "step": 1132
    },
    {
      "completion_length": 25.1875,
      "epoch": 3.4761904761904763,
      "grad_norm": 7.839033603668213,
      "kl": 0.466796875,
      "learning_rate": 1.2923076923076924e-07,
      "loss": -0.0198,
      "reward": 0.7959770560264587,
      "reward_std": 0.019398603588342667,
      "reward_std_all": 0.018379514571279287,
      "rewards-max/forensics_sinkhorn_reward": 0.9947424232959747,
      "rewards-max/forensics_strict_edge_reward": 0.7742990553379059,
      "rewards-max/forensics_tiou_reward": 0.9894998371601105,
      "rewards/forensics_sinkhorn_reward": 0.8241877257823944,
      "rewards/forensics_strict_edge_reward": 0.14802943170070648,
      "rewards/forensics_tiou_reward": 0.671680748462677,
      "step": 1133
    },
    {
      "completion_length": 25.5625,
      "epoch": 3.47926267281106,
      "grad_norm": 8.029900550842285,
      "kl": 0.4208984375,
      "learning_rate": 1.2846153846153845e-07,
      "loss": 0.0624,
      "reward": 0.7909009754657745,
      "reward_std": 0.018283777870237827,
      "reward_std_all": 0.019085456617176533,
      "rewards-max/forensics_sinkhorn_reward": 0.9813326299190521,
      "rewards-max/forensics_strict_edge_reward": 0.5368446409702301,
      "rewards-max/forensics_tiou_reward": 0.955970972776413,
      "rewards/forensics_sinkhorn_reward": 0.7898927628993988,
      "rewards/forensics_strict_edge_reward": 0.06884752959012985,
      "rewards/forensics_tiou_reward": 0.6275484561920166,
      "step": 1134
    },
    {
      "completion_length": 29.65625,
      "epoch": 3.482334869431644,
      "grad_norm": 7.739442348480225,
      "kl": 0.49609375,
      "learning_rate": 1.276923076923077e-07,
      "loss": 0.0797,
      "reward": 0.7618014812469482,
      "reward_std": 0.04243449680507183,
      "reward_std_all": 0.04323923168703914,
      "rewards-max/forensics_sinkhorn_reward": 0.9912085235118866,
      "rewards-max/forensics_strict_edge_reward": 0.38722455501556396,
      "rewards-max/forensics_tiou_reward": 0.9822240173816681,
      "rewards/forensics_sinkhorn_reward": 0.7375030219554901,
      "rewards/forensics_strict_edge_reward": 0.043155910447239876,
      "rewards/forensics_tiou_reward": 0.5687306374311447,
      "step": 1135
    },
    {
      "completion_length": 23.9375,
      "epoch": 3.485407066052227,
      "grad_norm": 8.725235939025879,
      "kl": 0.49609375,
      "learning_rate": 1.269230769230769e-07,
      "loss": -0.0256,
      "reward": 0.79377081990242,
      "reward_std": 0.012137641198933125,
      "reward_std_all": 0.014034921769052744,
      "rewards-max/forensics_sinkhorn_reward": 0.9793184101581573,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9500771164894104,
      "rewards/forensics_sinkhorn_reward": 0.7850413918495178,
      "rewards/forensics_strict_edge_reward": 0.05794152617454529,
      "rewards/forensics_tiou_reward": 0.6267277598381042,
      "step": 1136
    },
    {
      "completion_length": 21.984375,
      "epoch": 3.488479262672811,
      "grad_norm": 8.285262107849121,
      "kl": 0.4775390625,
      "learning_rate": 1.2615384615384615e-07,
      "loss": -0.0846,
      "reward": 0.7881108522415161,
      "reward_std": 0.010258188704028726,
      "reward_std_all": 0.014827391132712364,
      "rewards-max/forensics_sinkhorn_reward": 0.9847680628299713,
      "rewards-max/forensics_strict_edge_reward": 0.477401964366436,
      "rewards-max/forensics_tiou_reward": 0.9659362733364105,
      "rewards/forensics_sinkhorn_reward": 0.8037382066249847,
      "rewards/forensics_strict_edge_reward": 0.0568801867775619,
      "rewards/forensics_tiou_reward": 0.617699533700943,
      "step": 1137
    },
    {
      "completion_length": 24.484375,
      "epoch": 3.4915514592933947,
      "grad_norm": 9.176933288574219,
      "kl": 0.421875,
      "learning_rate": 1.2538461538461536e-07,
      "loss": -0.0155,
      "reward": 0.7806496620178223,
      "reward_std": 0.043684495612978935,
      "reward_std_all": 0.04002850130200386,
      "rewards-max/forensics_sinkhorn_reward": 0.9861780107021332,
      "rewards-max/forensics_strict_edge_reward": 0.4450889080762863,
      "rewards-max/forensics_tiou_reward": 0.971940666437149,
      "rewards/forensics_sinkhorn_reward": 0.7910829186439514,
      "rewards/forensics_strict_edge_reward": 0.04371308535337448,
      "rewards/forensics_tiou_reward": 0.6383158266544342,
      "step": 1138
    },
    {
      "completion_length": 25.390625,
      "epoch": 3.4946236559139785,
      "grad_norm": 7.331421375274658,
      "kl": 0.564453125,
      "learning_rate": 1.246153846153846e-07,
      "loss": -0.0576,
      "reward": 0.8036662042140961,
      "reward_std": 0.005612343200482428,
      "reward_std_all": 0.006523843389004469,
      "rewards-max/forensics_sinkhorn_reward": 0.9939583837985992,
      "rewards-max/forensics_strict_edge_reward": 0.6306980103254318,
      "rewards-max/forensics_tiou_reward": 0.9881081581115723,
      "rewards/forensics_sinkhorn_reward": 0.8305543959140778,
      "rewards/forensics_strict_edge_reward": 0.0987914390861988,
      "rewards/forensics_tiou_reward": 0.6777902841567993,
      "step": 1139
    },
    {
      "completion_length": 26.890625,
      "epoch": 3.4976958525345623,
      "grad_norm": 9.859292030334473,
      "kl": 0.4541015625,
      "learning_rate": 1.2384615384615384e-07,
      "loss": -0.035,
      "reward": 0.7936199903488159,
      "reward_std": 0.011555912380572408,
      "reward_std_all": 0.013854745659045875,
      "rewards-max/forensics_sinkhorn_reward": 0.981376439332962,
      "rewards-max/forensics_strict_edge_reward": 0.4712212085723877,
      "rewards-max/forensics_tiou_reward": 0.9606176912784576,
      "rewards/forensics_sinkhorn_reward": 0.7732287347316742,
      "rewards/forensics_strict_edge_reward": 0.053860776126384735,
      "rewards/forensics_tiou_reward": 0.6141356229782104,
      "step": 1140
    },
    {
      "completion_length": 21.125,
      "epoch": 3.500768049155146,
      "grad_norm": 7.731374740600586,
      "kl": 0.3857421875,
      "learning_rate": 1.2307692307692308e-07,
      "loss": -0.0741,
      "reward": 0.7632586658000946,
      "reward_std": 0.048219192773103714,
      "reward_std_all": 0.047999924048781395,
      "rewards-max/forensics_sinkhorn_reward": 0.9969899356365204,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9983034729957581,
      "rewards/forensics_sinkhorn_reward": 0.8156192302703857,
      "rewards/forensics_strict_edge_reward": 0.13389545306563377,
      "rewards/forensics_tiou_reward": 0.6638227105140686,
      "step": 1141
    },
    {
      "completion_length": 27.453125,
      "epoch": 3.50384024577573,
      "grad_norm": 7.999854564666748,
      "kl": 0.541015625,
      "learning_rate": 1.223076923076923e-07,
      "loss": 0.0223,
      "reward": 0.7982360124588013,
      "reward_std": 0.014993124874308705,
      "reward_std_all": 0.01462882524356246,
      "rewards-max/forensics_sinkhorn_reward": 0.9935970306396484,
      "rewards-max/forensics_strict_edge_reward": 0.686236709356308,
      "rewards-max/forensics_tiou_reward": 0.9880185723304749,
      "rewards/forensics_sinkhorn_reward": 0.8024011254310608,
      "rewards/forensics_strict_edge_reward": 0.10129913687705994,
      "rewards/forensics_tiou_reward": 0.7084723114967346,
      "step": 1142
    },
    {
      "completion_length": 25.3125,
      "epoch": 3.506912442396313,
      "grad_norm": 9.284214973449707,
      "kl": 0.64453125,
      "learning_rate": 1.2153846153846154e-07,
      "loss": -0.0317,
      "reward": 0.7951057255268097,
      "reward_std": 0.019107726402580738,
      "reward_std_all": 0.018560919910669327,
      "rewards-max/forensics_sinkhorn_reward": 0.9896596968173981,
      "rewards-max/forensics_strict_edge_reward": 0.5159863531589508,
      "rewards-max/forensics_tiou_reward": 0.9759138524532318,
      "rewards/forensics_sinkhorn_reward": 0.8046525120735168,
      "rewards/forensics_strict_edge_reward": 0.058333514258265495,
      "rewards/forensics_tiou_reward": 0.6582117676734924,
      "step": 1143
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.509984639016897,
      "grad_norm": 8.545984268188477,
      "kl": 0.474609375,
      "learning_rate": 1.2076923076923075e-07,
      "loss": -0.0928,
      "reward": 0.8039357662200928,
      "reward_std": 0.004380193422548473,
      "reward_std_all": 0.0057986893225461245,
      "rewards-max/forensics_sinkhorn_reward": 0.9933499693870544,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9904227256774902,
      "rewards/forensics_sinkhorn_reward": 0.86898472905159,
      "rewards/forensics_strict_edge_reward": 0.09480119310319424,
      "rewards/forensics_tiou_reward": 0.6715805232524872,
      "step": 1144
    },
    {
      "completion_length": 25.625,
      "epoch": 3.5130568356374807,
      "grad_norm": 8.863175392150879,
      "kl": 0.4267578125,
      "learning_rate": 1.2e-07,
      "loss": 0.0108,
      "reward": 0.7739493548870087,
      "reward_std": 0.03519422188401222,
      "reward_std_all": 0.034873565658926964,
      "rewards-max/forensics_sinkhorn_reward": 0.9870084822177887,
      "rewards-max/forensics_strict_edge_reward": 0.6716773808002472,
      "rewards-max/forensics_tiou_reward": 0.9711186587810516,
      "rewards/forensics_sinkhorn_reward": 0.8203856348991394,
      "rewards/forensics_strict_edge_reward": 0.11172222346067429,
      "rewards/forensics_tiou_reward": 0.6751187443733215,
      "step": 1145
    },
    {
      "completion_length": 24.765625,
      "epoch": 3.5161290322580645,
      "grad_norm": 9.357490539550781,
      "kl": 0.4560546875,
      "learning_rate": 1.1923076923076923e-07,
      "loss": 0.0814,
      "reward": 0.7544525861740112,
      "reward_std": 0.04668652079999447,
      "reward_std_all": 0.048605579882860184,
      "rewards-max/forensics_sinkhorn_reward": 0.9784355759620667,
      "rewards-max/forensics_strict_edge_reward": 0.33200711756944656,
      "rewards-max/forensics_tiou_reward": 0.9429546296596527,
      "rewards/forensics_sinkhorn_reward": 0.7249603867530823,
      "rewards/forensics_strict_edge_reward": 0.060335053130984306,
      "rewards/forensics_tiou_reward": 0.5610135495662689,
      "step": 1146
    },
    {
      "completion_length": 28.34375,
      "epoch": 3.5192012288786483,
      "grad_norm": 8.208331108093262,
      "kl": 0.462890625,
      "learning_rate": 1.1846153846153846e-07,
      "loss": -0.0179,
      "reward": 0.7437717318534851,
      "reward_std": 0.0724023450165987,
      "reward_std_all": 0.07450412958860397,
      "rewards-max/forensics_sinkhorn_reward": 0.9953894913196564,
      "rewards-max/forensics_strict_edge_reward": 0.5984092578291893,
      "rewards-max/forensics_tiou_reward": 0.9932521879673004,
      "rewards/forensics_sinkhorn_reward": 0.7571132183074951,
      "rewards/forensics_strict_edge_reward": 0.055312881246209145,
      "rewards/forensics_tiou_reward": 0.5930323600769043,
      "step": 1147
    },
    {
      "completion_length": 27.625,
      "epoch": 3.522273425499232,
      "grad_norm": 7.180883884429932,
      "kl": 0.44921875,
      "learning_rate": 1.1769230769230768e-07,
      "loss": -0.001,
      "reward": 0.7731605470180511,
      "reward_std": 0.03269229456782341,
      "reward_std_all": 0.033548000268638134,
      "rewards-max/forensics_sinkhorn_reward": 0.99082350730896,
      "rewards-max/forensics_strict_edge_reward": 0.5101649016141891,
      "rewards-max/forensics_tiou_reward": 0.9746334850788116,
      "rewards/forensics_sinkhorn_reward": 0.8107199966907501,
      "rewards/forensics_strict_edge_reward": 0.05223125033080578,
      "rewards/forensics_tiou_reward": 0.6598890125751495,
      "step": 1148
    },
    {
      "completion_length": 25.828125,
      "epoch": 3.525345622119816,
      "grad_norm": 9.371110916137695,
      "kl": 0.5244140625,
      "learning_rate": 1.1692307692307691e-07,
      "loss": 0.0608,
      "reward": 0.76213338971138,
      "reward_std": 0.034347291104495525,
      "reward_std_all": 0.041699089109897614,
      "rewards-max/forensics_sinkhorn_reward": 0.9859951138496399,
      "rewards-max/forensics_strict_edge_reward": 0.5394621193408966,
      "rewards-max/forensics_tiou_reward": 0.9686523377895355,
      "rewards/forensics_sinkhorn_reward": 0.7948162853717804,
      "rewards/forensics_strict_edge_reward": 0.08548712637275457,
      "rewards/forensics_tiou_reward": 0.6213328838348389,
      "step": 1149
    },
    {
      "completion_length": 24.359375,
      "epoch": 3.528417818740399,
      "grad_norm": 9.636282920837402,
      "kl": 0.427734375,
      "learning_rate": 1.1615384615384615e-07,
      "loss": -0.0682,
      "reward": 0.7689150273799896,
      "reward_std": 0.043824577704072,
      "reward_std_all": 0.04255103878676891,
      "rewards-max/forensics_sinkhorn_reward": 0.9863269925117493,
      "rewards-max/forensics_strict_edge_reward": 0.42791682481765747,
      "rewards-max/forensics_tiou_reward": 0.9676614999771118,
      "rewards/forensics_sinkhorn_reward": 0.8024688065052032,
      "rewards/forensics_strict_edge_reward": 0.052301716059446335,
      "rewards/forensics_tiou_reward": 0.6589296162128448,
      "step": 1150
    },
    {
      "completion_length": 25.125,
      "epoch": 3.531490015360983,
      "grad_norm": 10.960152626037598,
      "kl": 0.576171875,
      "learning_rate": 1.1538461538461539e-07,
      "loss": 0.0385,
      "reward": 0.7763620615005493,
      "reward_std": 0.02085340252961032,
      "reward_std_all": 0.0252184591954574,
      "rewards-max/forensics_sinkhorn_reward": 0.9893417656421661,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9766556322574615,
      "rewards/forensics_sinkhorn_reward": 0.8040758073329926,
      "rewards/forensics_strict_edge_reward": 0.049588073045015335,
      "rewards/forensics_tiou_reward": 0.6226461827754974,
      "step": 1151
    },
    {
      "completion_length": 22.703125,
      "epoch": 3.5345622119815667,
      "grad_norm": 11.256126403808594,
      "kl": 0.365234375,
      "learning_rate": 1.1461538461538462e-07,
      "loss": -0.0153,
      "reward": 0.7880650758743286,
      "reward_std": 0.01855997648090124,
      "reward_std_all": 0.01975778117775917,
      "rewards-max/forensics_sinkhorn_reward": 0.9935403168201447,
      "rewards-max/forensics_strict_edge_reward": 0.8469520509243011,
      "rewards-max/forensics_tiou_reward": 0.9877664148807526,
      "rewards/forensics_sinkhorn_reward": 0.8154328167438507,
      "rewards/forensics_strict_edge_reward": 0.0633673295378685,
      "rewards/forensics_tiou_reward": 0.6385392546653748,
      "step": 1152
    },
    {
      "completion_length": 25.671875,
      "epoch": 3.5376344086021505,
      "grad_norm": 14.450815200805664,
      "kl": 0.8642578125,
      "learning_rate": 1.1384615384615384e-07,
      "loss": 0.0782,
      "reward": 0.778027206659317,
      "reward_std": 0.037520166486501694,
      "reward_std_all": 0.035853090696036816,
      "rewards-max/forensics_sinkhorn_reward": 0.9808283746242523,
      "rewards-max/forensics_strict_edge_reward": 0.4251216650009155,
      "rewards-max/forensics_tiou_reward": 0.9564506411552429,
      "rewards/forensics_sinkhorn_reward": 0.7725835740566254,
      "rewards/forensics_strict_edge_reward": 0.03492183331400156,
      "rewards/forensics_tiou_reward": 0.6206232011318207,
      "step": 1153
    },
    {
      "completion_length": 23.984375,
      "epoch": 3.5407066052227343,
      "grad_norm": 7.498956680297852,
      "kl": 0.517578125,
      "learning_rate": 1.1307692307692307e-07,
      "loss": -0.0558,
      "reward": 0.7949649095535278,
      "reward_std": 0.013177823391743004,
      "reward_std_all": 0.014166401699185371,
      "rewards-max/forensics_sinkhorn_reward": 0.9919137954711914,
      "rewards-max/forensics_strict_edge_reward": 0.47706422209739685,
      "rewards-max/forensics_tiou_reward": 0.9821053743362427,
      "rewards/forensics_sinkhorn_reward": 0.8572929203510284,
      "rewards/forensics_strict_edge_reward": 0.06494747661054134,
      "rewards/forensics_tiou_reward": 0.6747323274612427,
      "step": 1154
    },
    {
      "completion_length": 25.421875,
      "epoch": 3.543778801843318,
      "grad_norm": 7.6792755126953125,
      "kl": 0.4580078125,
      "learning_rate": 1.123076923076923e-07,
      "loss": -0.0096,
      "reward": 0.7904193997383118,
      "reward_std": 0.029188170097768307,
      "reward_std_all": 0.026670588180422783,
      "rewards-max/forensics_sinkhorn_reward": 0.9928356111049652,
      "rewards-max/forensics_strict_edge_reward": 0.7205979228019714,
      "rewards-max/forensics_tiou_reward": 0.987729549407959,
      "rewards/forensics_sinkhorn_reward": 0.8313622772693634,
      "rewards/forensics_strict_edge_reward": 0.1325140856206417,
      "rewards/forensics_tiou_reward": 0.7025973498821259,
      "step": 1155
    },
    {
      "completion_length": 25.640625,
      "epoch": 3.546850998463902,
      "grad_norm": 8.44534683227539,
      "kl": 0.5,
      "learning_rate": 1.1153846153846154e-07,
      "loss": -0.0371,
      "reward": 0.8009666204452515,
      "reward_std": 0.009343780111521482,
      "reward_std_all": 0.010213032132014632,
      "rewards-max/forensics_sinkhorn_reward": 0.9842934310436249,
      "rewards-max/forensics_strict_edge_reward": 0.5536971986293793,
      "rewards-max/forensics_tiou_reward": 0.9628590941429138,
      "rewards/forensics_sinkhorn_reward": 0.8455798029899597,
      "rewards/forensics_strict_edge_reward": 0.05662534385919571,
      "rewards/forensics_tiou_reward": 0.6674586534500122,
      "step": 1156
    },
    {
      "completion_length": 28.25,
      "epoch": 3.549923195084485,
      "grad_norm": 8.908517837524414,
      "kl": 0.541015625,
      "learning_rate": 1.1076923076923077e-07,
      "loss": -0.0126,
      "reward": 0.792080819606781,
      "reward_std": 0.009988749050535262,
      "reward_std_all": 0.014125150628387928,
      "rewards-max/forensics_sinkhorn_reward": 0.9913254380226135,
      "rewards-max/forensics_strict_edge_reward": 0.5243985205888748,
      "rewards-max/forensics_tiou_reward": 0.9832257032394409,
      "rewards/forensics_sinkhorn_reward": 0.8747739493846893,
      "rewards/forensics_strict_edge_reward": 0.08102768659591675,
      "rewards/forensics_tiou_reward": 0.7705710828304291,
      "step": 1157
    },
    {
      "completion_length": 25.78125,
      "epoch": 3.5529953917050694,
      "grad_norm": 8.21705150604248,
      "kl": 0.544921875,
      "learning_rate": 1.0999999999999999e-07,
      "loss": 0.0553,
      "reward": 0.7886147797107697,
      "reward_std": 0.024837405420839787,
      "reward_std_all": 0.024127833545207977,
      "rewards-max/forensics_sinkhorn_reward": 0.9903479814529419,
      "rewards-max/forensics_strict_edge_reward": 0.5925651341676712,
      "rewards-max/forensics_tiou_reward": 0.9796778857707977,
      "rewards/forensics_sinkhorn_reward": 0.8125556707382202,
      "rewards/forensics_strict_edge_reward": 0.12243695929646492,
      "rewards/forensics_tiou_reward": 0.6913808286190033,
      "step": 1158
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.5560675883256527,
      "grad_norm": 10.06222915649414,
      "kl": 0.53125,
      "learning_rate": 1.0923076923076922e-07,
      "loss": 0.0,
      "reward": 0.777726024389267,
      "reward_std": 0.03630607621744275,
      "reward_std_all": 0.035765042062848806,
      "rewards-max/forensics_sinkhorn_reward": 0.9924125671386719,
      "rewards-max/forensics_strict_edge_reward": 0.6816803216934204,
      "rewards-max/forensics_tiou_reward": 0.984258383512497,
      "rewards/forensics_sinkhorn_reward": 0.8231797814369202,
      "rewards/forensics_strict_edge_reward": 0.10404732078313828,
      "rewards/forensics_tiou_reward": 0.639162540435791,
      "step": 1159
    },
    {
      "completion_length": 26.390625,
      "epoch": 3.5591397849462365,
      "grad_norm": 8.599841117858887,
      "kl": 0.5166015625,
      "learning_rate": 1.0846153846153845e-07,
      "loss": 0.0093,
      "reward": 0.7843006253242493,
      "reward_std": 0.03757887799292803,
      "reward_std_all": 0.034500883892178535,
      "rewards-max/forensics_sinkhorn_reward": 0.9858013689517975,
      "rewards-max/forensics_strict_edge_reward": 0.601691871881485,
      "rewards-max/forensics_tiou_reward": 0.972105860710144,
      "rewards/forensics_sinkhorn_reward": 0.7792790532112122,
      "rewards/forensics_strict_edge_reward": 0.09122958034276962,
      "rewards/forensics_tiou_reward": 0.6216439306735992,
      "step": 1160
    },
    {
      "completion_length": 21.46875,
      "epoch": 3.5622119815668203,
      "grad_norm": 9.823382377624512,
      "kl": 0.3896484375,
      "learning_rate": 1.076923076923077e-07,
      "loss": -0.0146,
      "reward": 0.7889575064182281,
      "reward_std": 0.030629001557826996,
      "reward_std_all": 0.028369427658617496,
      "rewards-max/forensics_sinkhorn_reward": 0.9932341575622559,
      "rewards-max/forensics_strict_edge_reward": 0.8187680542469025,
      "rewards-max/forensics_tiou_reward": 0.987594872713089,
      "rewards/forensics_sinkhorn_reward": 0.8359127938747406,
      "rewards/forensics_strict_edge_reward": 0.0960003063082695,
      "rewards/forensics_tiou_reward": 0.6408507823944092,
      "step": 1161
    },
    {
      "completion_length": 24.421875,
      "epoch": 3.565284178187404,
      "grad_norm": 8.968926429748535,
      "kl": 0.37109375,
      "learning_rate": 1.0692307692307693e-07,
      "loss": 0.008,
      "reward": 0.7961700260639191,
      "reward_std": 0.01812779193278402,
      "reward_std_all": 0.017551927594467998,
      "rewards-max/forensics_sinkhorn_reward": 0.9866506159305573,
      "rewards-max/forensics_strict_edge_reward": 0.5147627294063568,
      "rewards-max/forensics_tiou_reward": 0.9674694240093231,
      "rewards/forensics_sinkhorn_reward": 0.8611807525157928,
      "rewards/forensics_strict_edge_reward": 0.07928584516048431,
      "rewards/forensics_tiou_reward": 0.684426486492157,
      "step": 1162
    },
    {
      "completion_length": 23.359375,
      "epoch": 3.568356374807988,
      "grad_norm": 7.5303497314453125,
      "kl": 0.4970703125,
      "learning_rate": 1.0615384615384615e-07,
      "loss": 0.0453,
      "reward": 0.7865349054336548,
      "reward_std": 0.02589453646214679,
      "reward_std_all": 0.02548144047614187,
      "rewards-max/forensics_sinkhorn_reward": 0.9879404008388519,
      "rewards-max/forensics_strict_edge_reward": 0.742465078830719,
      "rewards-max/forensics_tiou_reward": 0.9745467007160187,
      "rewards/forensics_sinkhorn_reward": 0.7380824685096741,
      "rewards/forensics_strict_edge_reward": 0.11331260949373245,
      "rewards/forensics_tiou_reward": 0.6078698635101318,
      "step": 1163
    },
    {
      "completion_length": 24.25,
      "epoch": 3.571428571428571,
      "grad_norm": 9.3292818069458,
      "kl": 0.509765625,
      "learning_rate": 1.0538461538461538e-07,
      "loss": 0.0101,
      "reward": 0.7921194434165955,
      "reward_std": 0.01963196974247694,
      "reward_std_all": 0.019716907292604446,
      "rewards-max/forensics_sinkhorn_reward": 0.9554194808006287,
      "rewards-max/forensics_strict_edge_reward": 0.33789213187992573,
      "rewards-max/forensics_tiou_reward": 0.8958330452442169,
      "rewards/forensics_sinkhorn_reward": 0.7810589969158173,
      "rewards/forensics_strict_edge_reward": 0.043614841997623444,
      "rewards/forensics_tiou_reward": 0.600823312997818,
      "step": 1164
    },
    {
      "completion_length": 30.546875,
      "epoch": 3.5745007680491554,
      "grad_norm": 7.898451328277588,
      "kl": 0.3994140625,
      "learning_rate": 1.0461538461538461e-07,
      "loss": -0.0029,
      "reward": 0.7882652580738068,
      "reward_std": 0.023177077062427998,
      "reward_std_all": 0.022878890857100487,
      "rewards-max/forensics_sinkhorn_reward": 0.9931254386901855,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9887362718582153,
      "rewards/forensics_sinkhorn_reward": 0.7456497550010681,
      "rewards/forensics_strict_edge_reward": 0.08493928238749504,
      "rewards/forensics_tiou_reward": 0.605524867773056,
      "step": 1165
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.5775729646697387,
      "grad_norm": 9.307470321655273,
      "kl": 0.4775390625,
      "learning_rate": 1.0384615384615385e-07,
      "loss": -0.0164,
      "reward": 0.788379043340683,
      "reward_std": 0.024045919068157673,
      "reward_std_all": 0.02393151354044676,
      "rewards-max/forensics_sinkhorn_reward": 0.9911156892776489,
      "rewards-max/forensics_strict_edge_reward": 0.5805842876434326,
      "rewards-max/forensics_tiou_reward": 0.981065571308136,
      "rewards/forensics_sinkhorn_reward": 0.800552099943161,
      "rewards/forensics_strict_edge_reward": 0.07166705280542374,
      "rewards/forensics_tiou_reward": 0.587947815656662,
      "step": 1166
    },
    {
      "completion_length": 22.859375,
      "epoch": 3.5806451612903225,
      "grad_norm": 8.567479133605957,
      "kl": 0.5810546875,
      "learning_rate": 1.0307692307692307e-07,
      "loss": -0.0404,
      "reward": 0.7813123166561127,
      "reward_std": 0.023760398849844933,
      "reward_std_all": 0.025694568641483784,
      "rewards-max/forensics_sinkhorn_reward": 0.9925469756126404,
      "rewards-max/forensics_strict_edge_reward": 0.5826255530118942,
      "rewards-max/forensics_tiou_reward": 0.9850448966026306,
      "rewards/forensics_sinkhorn_reward": 0.7992275953292847,
      "rewards/forensics_strict_edge_reward": 0.06006435118615627,
      "rewards/forensics_tiou_reward": 0.6088287234306335,
      "step": 1167
    },
    {
      "completion_length": 30.0,
      "epoch": 3.5837173579109063,
      "grad_norm": 8.277417182922363,
      "kl": 0.3974609375,
      "learning_rate": 1.023076923076923e-07,
      "loss": -0.0618,
      "reward": 0.7985826730728149,
      "reward_std": 0.0133603346766904,
      "reward_std_all": 0.013578484300523996,
      "rewards-max/forensics_sinkhorn_reward": 0.9951393604278564,
      "rewards-max/forensics_strict_edge_reward": 0.7500158548355103,
      "rewards-max/forensics_tiou_reward": 0.9923431873321533,
      "rewards/forensics_sinkhorn_reward": 0.8117110133171082,
      "rewards/forensics_strict_edge_reward": 0.07591696083545685,
      "rewards/forensics_tiou_reward": 0.6863480508327484,
      "step": 1168
    },
    {
      "completion_length": 28.203125,
      "epoch": 3.58678955453149,
      "grad_norm": 7.963021278381348,
      "kl": 0.4443359375,
      "learning_rate": 1.0153846153846153e-07,
      "loss": -0.1166,
      "reward": 0.7758894264698029,
      "reward_std": 0.002288520336151123,
      "reward_std_all": 0.01839398918673396,
      "rewards-max/forensics_sinkhorn_reward": 0.9971861839294434,
      "rewards-max/forensics_strict_edge_reward": 0.7244620621204376,
      "rewards-max/forensics_tiou_reward": 0.9988830983638763,
      "rewards/forensics_sinkhorn_reward": 0.8428139984607697,
      "rewards/forensics_strict_edge_reward": 0.0999619048088789,
      "rewards/forensics_tiou_reward": 0.7017987966537476,
      "step": 1169
    },
    {
      "completion_length": 24.203125,
      "epoch": 3.589861751152074,
      "grad_norm": 8.061790466308594,
      "kl": 0.4345703125,
      "learning_rate": 1.0076923076923076e-07,
      "loss": -0.0495,
      "reward": 0.803200751543045,
      "reward_std": 0.006200834177434444,
      "reward_std_all": 0.007238322286866605,
      "rewards-max/forensics_sinkhorn_reward": 0.9943070113658905,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.991768479347229,
      "rewards/forensics_sinkhorn_reward": 0.8646206855773926,
      "rewards/forensics_strict_edge_reward": 0.09646748006343842,
      "rewards/forensics_tiou_reward": 0.696079820394516,
      "step": 1170
    },
    {
      "completion_length": 24.625,
      "epoch": 3.592933947772657,
      "grad_norm": 10.511825561523438,
      "kl": 0.4052734375,
      "learning_rate": 1e-07,
      "loss": 0.0484,
      "reward": 0.7765588164329529,
      "reward_std": 0.020398101769387722,
      "reward_std_all": 0.027669217437505722,
      "rewards-max/forensics_sinkhorn_reward": 0.9780500829219818,
      "rewards-max/forensics_strict_edge_reward": 0.33124835789203644,
      "rewards-max/forensics_tiou_reward": 0.9483762085437775,
      "rewards/forensics_sinkhorn_reward": 0.7965174615383148,
      "rewards/forensics_strict_edge_reward": 0.04210494086146355,
      "rewards/forensics_tiou_reward": 0.6126288175582886,
      "step": 1171
    },
    {
      "completion_length": 24.46875,
      "epoch": 3.5960061443932414,
      "grad_norm": 8.892409324645996,
      "kl": 0.58203125,
      "learning_rate": 9.923076923076922e-08,
      "loss": -0.0137,
      "reward": 0.7669646739959717,
      "reward_std": 0.03413468645885587,
      "reward_std_all": 0.03712216857820749,
      "rewards-max/forensics_sinkhorn_reward": 0.9915926456451416,
      "rewards-max/forensics_strict_edge_reward": 0.39324072003364563,
      "rewards-max/forensics_tiou_reward": 0.9803325533866882,
      "rewards/forensics_sinkhorn_reward": 0.7984854876995087,
      "rewards/forensics_strict_edge_reward": 0.04482343606650829,
      "rewards/forensics_tiou_reward": 0.6272444725036621,
      "step": 1172
    },
    {
      "completion_length": 23.546875,
      "epoch": 3.5990783410138247,
      "grad_norm": 10.076227188110352,
      "kl": 0.439453125,
      "learning_rate": 9.846153846153846e-08,
      "loss": 0.0708,
      "reward": 0.7695167660713196,
      "reward_std": 0.031192901311442256,
      "reward_std_all": 0.037987057119607925,
      "rewards-max/forensics_sinkhorn_reward": 0.9868482649326324,
      "rewards-max/forensics_strict_edge_reward": 0.3477187380194664,
      "rewards-max/forensics_tiou_reward": 0.9619519710540771,
      "rewards/forensics_sinkhorn_reward": 0.8085373342037201,
      "rewards/forensics_strict_edge_reward": 0.036758753936737776,
      "rewards/forensics_tiou_reward": 0.651108592748642,
      "step": 1173
    },
    {
      "completion_length": 22.796875,
      "epoch": 3.6021505376344085,
      "grad_norm": 9.035984992980957,
      "kl": 0.564453125,
      "learning_rate": 9.769230769230769e-08,
      "loss": 0.0383,
      "reward": 0.7844422161579132,
      "reward_std": 0.02890144381672144,
      "reward_std_all": 0.028055557515472174,
      "rewards-max/forensics_sinkhorn_reward": 0.9890916645526886,
      "rewards-max/forensics_strict_edge_reward": 0.5984183251857758,
      "rewards-max/forensics_tiou_reward": 0.9751665890216827,
      "rewards/forensics_sinkhorn_reward": 0.8028226494789124,
      "rewards/forensics_strict_edge_reward": 0.10580594651401043,
      "rewards/forensics_tiou_reward": 0.6013359874486923,
      "step": 1174
    },
    {
      "completion_length": 23.0625,
      "epoch": 3.6052227342549923,
      "grad_norm": 10.143426895141602,
      "kl": 0.556640625,
      "learning_rate": 9.692307692307692e-08,
      "loss": -0.0552,
      "reward": 0.8025139570236206,
      "reward_std": 0.007240869337692857,
      "reward_std_all": 0.008212605956941843,
      "rewards-max/forensics_sinkhorn_reward": 0.9950074851512909,
      "rewards-max/forensics_strict_edge_reward": 0.8146918714046478,
      "rewards-max/forensics_tiou_reward": 0.9919725358486176,
      "rewards/forensics_sinkhorn_reward": 0.8262183368206024,
      "rewards/forensics_strict_edge_reward": 0.08949650079011917,
      "rewards/forensics_tiou_reward": 0.7046573758125305,
      "step": 1175
    },
    {
      "completion_length": 25.203125,
      "epoch": 3.608294930875576,
      "grad_norm": 10.166105270385742,
      "kl": 0.962890625,
      "learning_rate": 9.615384615384616e-08,
      "loss": 0.0416,
      "reward": 0.7736578583717346,
      "reward_std": 0.056600723415613174,
      "reward_std_all": 0.0509840901941061,
      "rewards-max/forensics_sinkhorn_reward": 0.9969730377197266,
      "rewards-max/forensics_strict_edge_reward": 0.8021849691867828,
      "rewards-max/forensics_tiou_reward": 0.9971610009670258,
      "rewards/forensics_sinkhorn_reward": 0.791524201631546,
      "rewards/forensics_strict_edge_reward": 0.09453132748603821,
      "rewards/forensics_tiou_reward": 0.634416937828064,
      "step": 1176
    },
    {
      "completion_length": 24.359375,
      "epoch": 3.61136712749616,
      "grad_norm": 8.4871244430542,
      "kl": 0.5048828125,
      "learning_rate": 9.538461538461538e-08,
      "loss": -0.0497,
      "reward": 0.7982012629508972,
      "reward_std": 0.013122296542860568,
      "reward_std_all": 0.013632943271659315,
      "rewards-max/forensics_sinkhorn_reward": 0.9916003048419952,
      "rewards-max/forensics_strict_edge_reward": 0.6052689850330353,
      "rewards-max/forensics_tiou_reward": 0.9823501706123352,
      "rewards/forensics_sinkhorn_reward": 0.7881763577461243,
      "rewards/forensics_strict_edge_reward": 0.06941370666027069,
      "rewards/forensics_tiou_reward": 0.6108965277671814,
      "step": 1177
    },
    {
      "completion_length": 25.1875,
      "epoch": 3.6144393241167436,
      "grad_norm": 7.590672016143799,
      "kl": 0.4580078125,
      "learning_rate": 9.461538461538461e-08,
      "loss": -0.0556,
      "reward": 0.7992891371250153,
      "reward_std": 0.011433084961026907,
      "reward_std_all": 0.012123082531616092,
      "rewards-max/forensics_sinkhorn_reward": 0.9870987832546234,
      "rewards-max/forensics_strict_edge_reward": 0.5107811987400055,
      "rewards-max/forensics_tiou_reward": 0.9729125201702118,
      "rewards/forensics_sinkhorn_reward": 0.8064468801021576,
      "rewards/forensics_strict_edge_reward": 0.07474647089838982,
      "rewards/forensics_tiou_reward": 0.6148582696914673,
      "step": 1178
    },
    {
      "completion_length": 25.0,
      "epoch": 3.6175115207373274,
      "grad_norm": 9.583656311035156,
      "kl": 0.4814453125,
      "learning_rate": 9.384615384615384e-08,
      "loss": 0.106,
      "reward": 0.7894335091114044,
      "reward_std": 0.031296381726861,
      "reward_std_all": 0.028377966955304146,
      "rewards-max/forensics_sinkhorn_reward": 0.9932767450809479,
      "rewards-max/forensics_strict_edge_reward": 0.7528800368309021,
      "rewards-max/forensics_tiou_reward": 0.9887517690658569,
      "rewards/forensics_sinkhorn_reward": 0.8308866024017334,
      "rewards/forensics_strict_edge_reward": 0.13015012070536613,
      "rewards/forensics_tiou_reward": 0.7110938429832458,
      "step": 1179
    },
    {
      "completion_length": 24.625,
      "epoch": 3.6205837173579107,
      "grad_norm": 8.00652027130127,
      "kl": 0.453125,
      "learning_rate": 9.307692307692306e-08,
      "loss": -0.0181,
      "reward": 0.7864709794521332,
      "reward_std": 0.030472940095933154,
      "reward_std_all": 0.028690066887065768,
      "rewards-max/forensics_sinkhorn_reward": 0.9941783249378204,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9899362325668335,
      "rewards/forensics_sinkhorn_reward": 0.8253160715103149,
      "rewards/forensics_strict_edge_reward": 0.0795719064772129,
      "rewards/forensics_tiou_reward": 0.6942245960235596,
      "step": 1180
    },
    {
      "completion_length": 25.84375,
      "epoch": 3.6236559139784945,
      "grad_norm": 13.077445030212402,
      "kl": 0.73828125,
      "learning_rate": 9.23076923076923e-08,
      "loss": -0.0011,
      "reward": 0.783194363117218,
      "reward_std": 0.02567214542068541,
      "reward_std_all": 0.025951406452804804,
      "rewards-max/forensics_sinkhorn_reward": 0.9962179660797119,
      "rewards-max/forensics_strict_edge_reward": 0.6206092536449432,
      "rewards-max/forensics_tiou_reward": 0.9951118528842926,
      "rewards/forensics_sinkhorn_reward": 0.8018752932548523,
      "rewards/forensics_strict_edge_reward": 0.11223294958472252,
      "rewards/forensics_tiou_reward": 0.6904558837413788,
      "step": 1181
    },
    {
      "completion_length": 24.15625,
      "epoch": 3.6267281105990783,
      "grad_norm": 9.052806854248047,
      "kl": 0.5390625,
      "learning_rate": 9.153846153846153e-08,
      "loss": -0.0332,
      "reward": 0.7969667017459869,
      "reward_std": 0.01648610131815076,
      "reward_std_all": 0.016055951360613108,
      "rewards-max/forensics_sinkhorn_reward": 0.9945550560951233,
      "rewards-max/forensics_strict_edge_reward": 0.615243673324585,
      "rewards-max/forensics_tiou_reward": 0.9894348382949829,
      "rewards/forensics_sinkhorn_reward": 0.8013487160205841,
      "rewards/forensics_strict_edge_reward": 0.09275487437844276,
      "rewards/forensics_tiou_reward": 0.6367948353290558,
      "step": 1182
    },
    {
      "completion_length": 25.0,
      "epoch": 3.629800307219662,
      "grad_norm": 12.42894458770752,
      "kl": 0.4130859375,
      "learning_rate": 9.076923076923076e-08,
      "loss": -0.0217,
      "reward": 0.7846917510032654,
      "reward_std": 0.026795043144375086,
      "reward_std_all": 0.026469238102436066,
      "rewards-max/forensics_sinkhorn_reward": 0.9916374981403351,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9845593273639679,
      "rewards/forensics_sinkhorn_reward": 0.8084726631641388,
      "rewards/forensics_strict_edge_reward": 0.14961763843894005,
      "rewards/forensics_tiou_reward": 0.648549497127533,
      "step": 1183
    },
    {
      "completion_length": 28.359375,
      "epoch": 3.632872503840246,
      "grad_norm": 10.30914306640625,
      "kl": 0.3681640625,
      "learning_rate": 9e-08,
      "loss": 0.0309,
      "reward": 0.784561425447464,
      "reward_std": 0.02700707595795393,
      "reward_std_all": 0.026720083318650723,
      "rewards-max/forensics_sinkhorn_reward": 0.9898011088371277,
      "rewards-max/forensics_strict_edge_reward": 0.3942515701055527,
      "rewards-max/forensics_tiou_reward": 0.9755394458770752,
      "rewards/forensics_sinkhorn_reward": 0.8149219453334808,
      "rewards/forensics_strict_edge_reward": 0.04498006030917168,
      "rewards/forensics_tiou_reward": 0.6854023933410645,
      "step": 1184
    },
    {
      "completion_length": 27.484375,
      "epoch": 3.6359447004608296,
      "grad_norm": 13.692744255065918,
      "kl": 0.482421875,
      "learning_rate": 8.923076923076923e-08,
      "loss": 0.0329,
      "reward": 0.7411127388477325,
      "reward_std": 0.06307038478553295,
      "reward_std_all": 0.06356553733348846,
      "rewards-max/forensics_sinkhorn_reward": 0.9852146506309509,
      "rewards-max/forensics_strict_edge_reward": 0.5559177100658417,
      "rewards-max/forensics_tiou_reward": 0.9685728847980499,
      "rewards/forensics_sinkhorn_reward": 0.7246686816215515,
      "rewards/forensics_strict_edge_reward": 0.057235077023506165,
      "rewards/forensics_tiou_reward": 0.526065319776535,
      "step": 1185
    },
    {
      "completion_length": 25.453125,
      "epoch": 3.6390168970814134,
      "grad_norm": 7.64868688583374,
      "kl": 0.4912109375,
      "learning_rate": 8.846153846153847e-08,
      "loss": -0.055,
      "reward": 0.7849090397357941,
      "reward_std": 0.025388329289853573,
      "reward_std_all": 0.025533578358590603,
      "rewards-max/forensics_sinkhorn_reward": 0.9954430758953094,
      "rewards-max/forensics_strict_edge_reward": 0.6101458668708801,
      "rewards-max/forensics_tiou_reward": 0.9928705990314484,
      "rewards/forensics_sinkhorn_reward": 0.7992655634880066,
      "rewards/forensics_strict_edge_reward": 0.0782730020582676,
      "rewards/forensics_tiou_reward": 0.6492582559585571,
      "step": 1186
    },
    {
      "completion_length": 25.9375,
      "epoch": 3.6420890937019967,
      "grad_norm": 7.802980899810791,
      "kl": 0.4912109375,
      "learning_rate": 8.769230769230769e-08,
      "loss": -0.0351,
      "reward": 0.8020831942558289,
      "reward_std": 0.008072700933553278,
      "reward_std_all": 0.008854974410496652,
      "rewards-max/forensics_sinkhorn_reward": 0.9849373698234558,
      "rewards-max/forensics_strict_edge_reward": 0.6602591574192047,
      "rewards-max/forensics_tiou_reward": 0.9681052267551422,
      "rewards/forensics_sinkhorn_reward": 0.8463473320007324,
      "rewards/forensics_strict_edge_reward": 0.1106630265712738,
      "rewards/forensics_tiou_reward": 0.6823500692844391,
      "step": 1187
    },
    {
      "completion_length": 28.546875,
      "epoch": 3.6451612903225805,
      "grad_norm": 7.771912097930908,
      "kl": 0.4306640625,
      "learning_rate": 8.692307692307692e-08,
      "loss": -0.0642,
      "reward": 0.7608178853988647,
      "reward_std": 0.019690541550517082,
      "reward_std_all": 0.034333779476583004,
      "rewards-max/forensics_sinkhorn_reward": 0.9894097149372101,
      "rewards-max/forensics_strict_edge_reward": 0.6095658242702484,
      "rewards-max/forensics_tiou_reward": 0.9779844880104065,
      "rewards/forensics_sinkhorn_reward": 0.7454997599124908,
      "rewards/forensics_strict_edge_reward": 0.05947329103946686,
      "rewards/forensics_tiou_reward": 0.637147456407547,
      "step": 1188
    },
    {
      "completion_length": 24.0,
      "epoch": 3.6482334869431643,
      "grad_norm": 9.036861419677734,
      "kl": 0.5048828125,
      "learning_rate": 8.615384615384615e-08,
      "loss": 0.0318,
      "reward": 0.7741157412528992,
      "reward_std": 0.04070706479251385,
      "reward_std_all": 0.04063025675714016,
      "rewards-max/forensics_sinkhorn_reward": 0.989083468914032,
      "rewards-max/forensics_strict_edge_reward": 0.5192448049783707,
      "rewards-max/forensics_tiou_reward": 0.9774311780929565,
      "rewards/forensics_sinkhorn_reward": 0.7838776409626007,
      "rewards/forensics_strict_edge_reward": 0.07592050731182098,
      "rewards/forensics_tiou_reward": 0.684573620557785,
      "step": 1189
    },
    {
      "completion_length": 22.484375,
      "epoch": 3.651305683563748,
      "grad_norm": 9.689045906066895,
      "kl": 0.572265625,
      "learning_rate": 8.538461538461537e-08,
      "loss": 0.0876,
      "reward": 0.7830123603343964,
      "reward_std": 0.03760697692632675,
      "reward_std_all": 0.034545065835118294,
      "rewards-max/forensics_sinkhorn_reward": 0.9861361980438232,
      "rewards-max/forensics_strict_edge_reward": 0.5484400391578674,
      "rewards-max/forensics_tiou_reward": 0.9709779620170593,
      "rewards/forensics_sinkhorn_reward": 0.7847537994384766,
      "rewards/forensics_strict_edge_reward": 0.07812873274087906,
      "rewards/forensics_tiou_reward": 0.5946074724197388,
      "step": 1190
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.654377880184332,
      "grad_norm": 11.060285568237305,
      "kl": 0.705078125,
      "learning_rate": 8.461538461538461e-08,
      "loss": 0.0006,
      "reward": 0.8035075068473816,
      "reward_std": 0.0063470202730968595,
      "reward_std_all": 0.007190754171460867,
      "rewards-max/forensics_sinkhorn_reward": 0.9953447580337524,
      "rewards-max/forensics_strict_edge_reward": 0.6198612302541733,
      "rewards-max/forensics_tiou_reward": 0.9918777644634247,
      "rewards/forensics_sinkhorn_reward": 0.8676250278949738,
      "rewards/forensics_strict_edge_reward": 0.09855925850570202,
      "rewards/forensics_tiou_reward": 0.7160958051681519,
      "step": 1191
    },
    {
      "completion_length": 22.4375,
      "epoch": 3.6574500768049156,
      "grad_norm": 9.123038291931152,
      "kl": 0.712890625,
      "learning_rate": 8.384615384615384e-08,
      "loss": 0.0113,
      "reward": 0.783039927482605,
      "reward_std": 0.02754002995789051,
      "reward_std_all": 0.027369407936930656,
      "rewards-max/forensics_sinkhorn_reward": 0.9948974847793579,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9885277152061462,
      "rewards/forensics_sinkhorn_reward": 0.7747781872749329,
      "rewards/forensics_strict_edge_reward": 0.12292080745100975,
      "rewards/forensics_tiou_reward": 0.6362981200218201,
      "step": 1192
    },
    {
      "completion_length": 26.421875,
      "epoch": 3.6605222734254994,
      "grad_norm": 10.27086067199707,
      "kl": 0.4423828125,
      "learning_rate": 8.307692307692307e-08,
      "loss": -0.0165,
      "reward": 0.7952747642993927,
      "reward_std": 0.01811068458482623,
      "reward_std_all": 0.018061564303934574,
      "rewards-max/forensics_sinkhorn_reward": 0.9859582483768463,
      "rewards-max/forensics_strict_edge_reward": 0.5469937175512314,
      "rewards-max/forensics_tiou_reward": 0.9702242016792297,
      "rewards/forensics_sinkhorn_reward": 0.7924093306064606,
      "rewards/forensics_strict_edge_reward": 0.05442635528743267,
      "rewards/forensics_tiou_reward": 0.6390120387077332,
      "step": 1193
    },
    {
      "completion_length": 22.453125,
      "epoch": 3.6635944700460827,
      "grad_norm": 7.611258029937744,
      "kl": 0.43359375,
      "learning_rate": 8.23076923076923e-08,
      "loss": -0.1055,
      "reward": 0.7851566672325134,
      "reward_std": 0.023996905947569758,
      "reward_std_all": 0.024394692271016538,
      "rewards-max/forensics_sinkhorn_reward": 0.9926124811172485,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9859050214290619,
      "rewards/forensics_sinkhorn_reward": 0.8214524686336517,
      "rewards/forensics_strict_edge_reward": 0.09097110293805599,
      "rewards/forensics_tiou_reward": 0.6463482975959778,
      "step": 1194
    },
    {
      "completion_length": 22.421875,
      "epoch": 3.6666666666666665,
      "grad_norm": 7.91177225112915,
      "kl": 0.4814453125,
      "learning_rate": 8.153846153846153e-08,
      "loss": 0.0123,
      "reward": 0.7733860909938812,
      "reward_std": 0.057387569919228554,
      "reward_std_all": 0.05147985927760601,
      "rewards-max/forensics_sinkhorn_reward": 0.9948663413524628,
      "rewards-max/forensics_strict_edge_reward": 0.8062050342559814,
      "rewards-max/forensics_tiou_reward": 0.9915492534637451,
      "rewards/forensics_sinkhorn_reward": 0.7770355045795441,
      "rewards/forensics_strict_edge_reward": 0.11440319940447807,
      "rewards/forensics_tiou_reward": 0.6045616865158081,
      "step": 1195
    },
    {
      "completion_length": 24.046875,
      "epoch": 3.6697388632872503,
      "grad_norm": 11.654193878173828,
      "kl": 0.4326171875,
      "learning_rate": 8.076923076923077e-08,
      "loss": 0.0986,
      "reward": 0.7715449929237366,
      "reward_std": 0.055109333246946335,
      "reward_std_all": 0.05037375260144472,
      "rewards-max/forensics_sinkhorn_reward": 0.9853077828884125,
      "rewards-max/forensics_strict_edge_reward": 0.5599738955497742,
      "rewards-max/forensics_tiou_reward": 0.9712858498096466,
      "rewards/forensics_sinkhorn_reward": 0.7624650597572327,
      "rewards/forensics_strict_edge_reward": 0.04563015513122082,
      "rewards/forensics_tiou_reward": 0.577143132686615,
      "step": 1196
    },
    {
      "completion_length": 26.4375,
      "epoch": 3.672811059907834,
      "grad_norm": 9.716193199157715,
      "kl": 0.470703125,
      "learning_rate": 8e-08,
      "loss": -0.0808,
      "reward": 0.7823395133018494,
      "reward_std": 0.03151187486946583,
      "reward_std_all": 0.03072363417595625,
      "rewards-max/forensics_sinkhorn_reward": 0.9868773818016052,
      "rewards-max/forensics_strict_edge_reward": 0.46096503734588623,
      "rewards-max/forensics_tiou_reward": 0.9629982709884644,
      "rewards/forensics_sinkhorn_reward": 0.7831089198589325,
      "rewards/forensics_strict_edge_reward": 0.05751068703830242,
      "rewards/forensics_tiou_reward": 0.6058492064476013,
      "step": 1197
    },
    {
      "completion_length": 24.265625,
      "epoch": 3.675883256528418,
      "grad_norm": 8.67502212524414,
      "kl": 0.4130859375,
      "learning_rate": 7.923076923076923e-08,
      "loss": 0.016,
      "reward": 0.7455922961235046,
      "reward_std": 0.03841435257345438,
      "reward_std_all": 0.052654678001999855,
      "rewards-max/forensics_sinkhorn_reward": 0.984289139509201,
      "rewards-max/forensics_strict_edge_reward": 0.5952748656272888,
      "rewards-max/forensics_tiou_reward": 0.9665425717830658,
      "rewards/forensics_sinkhorn_reward": 0.7457156777381897,
      "rewards/forensics_strict_edge_reward": 0.06071155145764351,
      "rewards/forensics_tiou_reward": 0.5593787431716919,
      "step": 1198
    },
    {
      "completion_length": 25.421875,
      "epoch": 3.6789554531490016,
      "grad_norm": 9.364633560180664,
      "kl": 0.4541015625,
      "learning_rate": 7.846153846153846e-08,
      "loss": -0.0469,
      "reward": 0.7964633703231812,
      "reward_std": 0.01193807041272521,
      "reward_std_all": 0.01287034060806036,
      "rewards-max/forensics_sinkhorn_reward": 0.9894371330738068,
      "rewards-max/forensics_strict_edge_reward": 0.5619910657405853,
      "rewards-max/forensics_tiou_reward": 0.9707334339618683,
      "rewards/forensics_sinkhorn_reward": 0.8451552093029022,
      "rewards/forensics_strict_edge_reward": 0.06644340418279171,
      "rewards/forensics_tiou_reward": 0.7158275246620178,
      "step": 1199
    },
    {
      "completion_length": 22.875,
      "epoch": 3.6820276497695854,
      "grad_norm": 8.111799240112305,
      "kl": 0.49609375,
      "learning_rate": 7.769230769230768e-08,
      "loss": -0.1081,
      "reward": 0.7956512868404388,
      "reward_std": 0.013197305554058403,
      "reward_std_all": 0.01409086980856955,
      "rewards-max/forensics_sinkhorn_reward": 0.9814141690731049,
      "rewards-max/forensics_strict_edge_reward": 0.6497856378555298,
      "rewards-max/forensics_tiou_reward": 0.9614735841751099,
      "rewards/forensics_sinkhorn_reward": 0.8291790783405304,
      "rewards/forensics_strict_edge_reward": 0.08459140360355377,
      "rewards/forensics_tiou_reward": 0.686757355928421,
      "step": 1200
    },
    {
      "completion_length": 23.75,
      "epoch": 3.6850998463901687,
      "grad_norm": 8.44949722290039,
      "kl": 0.5029296875,
      "learning_rate": 7.692307692307692e-08,
      "loss": 0.0149,
      "reward": 0.7887855172157288,
      "reward_std": 0.025741524063050747,
      "reward_std_all": 0.024624151177704334,
      "rewards-max/forensics_sinkhorn_reward": 0.9939826130867004,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9918049275875092,
      "rewards/forensics_sinkhorn_reward": 0.8152337074279785,
      "rewards/forensics_strict_edge_reward": 0.08113397285342216,
      "rewards/forensics_tiou_reward": 0.6507208347320557,
      "step": 1201
    },
    {
      "completion_length": 22.609375,
      "epoch": 3.688172043010753,
      "grad_norm": 8.732840538024902,
      "kl": 0.4091796875,
      "learning_rate": 7.615384615384615e-08,
      "loss": 0.0051,
      "reward": 0.789155900478363,
      "reward_std": 0.02903344016522169,
      "reward_std_all": 0.02709783799946308,
      "rewards-max/forensics_sinkhorn_reward": 0.9776592552661896,
      "rewards-max/forensics_strict_edge_reward": 0.5090594589710236,
      "rewards-max/forensics_tiou_reward": 0.9502246975898743,
      "rewards/forensics_sinkhorn_reward": 0.7889891862869263,
      "rewards/forensics_strict_edge_reward": 0.05823967047035694,
      "rewards/forensics_tiou_reward": 0.5883823335170746,
      "step": 1202
    },
    {
      "completion_length": 28.03125,
      "epoch": 3.6912442396313363,
      "grad_norm": 8.551039695739746,
      "kl": 0.56640625,
      "learning_rate": 7.538461538461538e-08,
      "loss": -0.0967,
      "reward": 0.7991835176944733,
      "reward_std": 0.011724979616701603,
      "reward_std_all": 0.011880733072757721,
      "rewards-max/forensics_sinkhorn_reward": 0.9893724322319031,
      "rewards-max/forensics_strict_edge_reward": 0.6799494326114655,
      "rewards-max/forensics_tiou_reward": 0.9792713820934296,
      "rewards/forensics_sinkhorn_reward": 0.7920761406421661,
      "rewards/forensics_strict_edge_reward": 0.1306694969534874,
      "rewards/forensics_tiou_reward": 0.6673693656921387,
      "step": 1203
    },
    {
      "completion_length": 23.953125,
      "epoch": 3.69431643625192,
      "grad_norm": 7.2816009521484375,
      "kl": 0.43359375,
      "learning_rate": 7.46153846153846e-08,
      "loss": 0.0156,
      "reward": 0.7865684628486633,
      "reward_std": 0.024292892310768366,
      "reward_std_all": 0.024710770696401596,
      "rewards-max/forensics_sinkhorn_reward": 0.9828951060771942,
      "rewards-max/forensics_strict_edge_reward": 0.33076252043247223,
      "rewards-max/forensics_tiou_reward": 0.9577116072177887,
      "rewards/forensics_sinkhorn_reward": 0.7667082846164703,
      "rewards/forensics_strict_edge_reward": 0.03403100371360779,
      "rewards/forensics_tiou_reward": 0.5298637598752975,
      "step": 1204
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.697388632872504,
      "grad_norm": 10.060660362243652,
      "kl": 0.7158203125,
      "learning_rate": 7.384615384615384e-08,
      "loss": 0.0347,
      "reward": 0.7994356751441956,
      "reward_std": 0.013978487811982632,
      "reward_std_all": 0.01338900811970234,
      "rewards-max/forensics_sinkhorn_reward": 0.9953569769859314,
      "rewards-max/forensics_strict_edge_reward": 0.8872029185295105,
      "rewards-max/forensics_tiou_reward": 0.9936141967773438,
      "rewards/forensics_sinkhorn_reward": 0.8542818427085876,
      "rewards/forensics_strict_edge_reward": 0.14896167814731598,
      "rewards/forensics_tiou_reward": 0.7340662479400635,
      "step": 1205
    },
    {
      "completion_length": 25.328125,
      "epoch": 3.7004608294930876,
      "grad_norm": 7.606468677520752,
      "kl": 0.689453125,
      "learning_rate": 7.307692307692308e-08,
      "loss": 0.058,
      "reward": 0.7836622595787048,
      "reward_std": 0.039936816319823265,
      "reward_std_all": 0.036305056884884834,
      "rewards-max/forensics_sinkhorn_reward": 0.9892097413539886,
      "rewards-max/forensics_strict_edge_reward": 0.3908846378326416,
      "rewards-max/forensics_tiou_reward": 0.9736469089984894,
      "rewards/forensics_sinkhorn_reward": 0.8246844708919525,
      "rewards/forensics_strict_edge_reward": 0.061486138962209225,
      "rewards/forensics_tiou_reward": 0.650269627571106,
      "step": 1206
    },
    {
      "completion_length": 26.0,
      "epoch": 3.7035330261136714,
      "grad_norm": 7.943129539489746,
      "kl": 0.3662109375,
      "learning_rate": 7.230769230769231e-08,
      "loss": -0.0691,
      "reward": 0.7905043363571167,
      "reward_std": 0.027825184515677392,
      "reward_std_all": 0.02590525324922055,
      "rewards-max/forensics_sinkhorn_reward": 0.9944714307785034,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9907499253749847,
      "rewards/forensics_sinkhorn_reward": 0.8179076611995697,
      "rewards/forensics_strict_edge_reward": 0.12675317376852036,
      "rewards/forensics_tiou_reward": 0.6607395112514496,
      "step": 1207
    },
    {
      "completion_length": 27.125,
      "epoch": 3.7066052227342547,
      "grad_norm": 8.748106002807617,
      "kl": 0.42578125,
      "learning_rate": 7.153846153846154e-08,
      "loss": -0.0559,
      "reward": 0.7621119022369385,
      "reward_std": 0.004217560752294958,
      "reward_std_all": 0.026179422857239842,
      "rewards-max/forensics_sinkhorn_reward": 0.9941329956054688,
      "rewards-max/forensics_strict_edge_reward": 0.685482993721962,
      "rewards-max/forensics_tiou_reward": 0.9899418950080872,
      "rewards/forensics_sinkhorn_reward": 0.8079100251197815,
      "rewards/forensics_strict_edge_reward": 0.11955202743411064,
      "rewards/forensics_tiou_reward": 0.660539299249649,
      "step": 1208
    },
    {
      "completion_length": 25.5,
      "epoch": 3.709677419354839,
      "grad_norm": 7.46025276184082,
      "kl": 0.462890625,
      "learning_rate": 7.076923076923076e-08,
      "loss": 0.0393,
      "reward": 0.7808685898780823,
      "reward_std": 0.026370959356427193,
      "reward_std_all": 0.027091980911791325,
      "rewards-max/forensics_sinkhorn_reward": 0.9923641979694366,
      "rewards-max/forensics_strict_edge_reward": 0.6966203600168228,
      "rewards-max/forensics_tiou_reward": 0.9862292110919952,
      "rewards/forensics_sinkhorn_reward": 0.7753024995326996,
      "rewards/forensics_strict_edge_reward": 0.11709252744913101,
      "rewards/forensics_tiou_reward": 0.6550838947296143,
      "step": 1209
    },
    {
      "completion_length": 26.0625,
      "epoch": 3.7127496159754223,
      "grad_norm": 8.67493724822998,
      "kl": 0.4443359375,
      "learning_rate": 7e-08,
      "loss": 0.0157,
      "reward": 0.778519868850708,
      "reward_std": 0.027378097409382463,
      "reward_std_all": 0.028439295245334506,
      "rewards-max/forensics_sinkhorn_reward": 0.987877607345581,
      "rewards-max/forensics_strict_edge_reward": 0.6250697076320648,
      "rewards-max/forensics_tiou_reward": 0.9746203124523163,
      "rewards/forensics_sinkhorn_reward": 0.7574232518672943,
      "rewards/forensics_strict_edge_reward": 0.06853415071964264,
      "rewards/forensics_tiou_reward": 0.5757879614830017,
      "step": 1210
    },
    {
      "completion_length": 22.546875,
      "epoch": 3.715821812596006,
      "grad_norm": 8.090981483459473,
      "kl": 0.4677734375,
      "learning_rate": 6.923076923076923e-08,
      "loss": -0.0848,
      "reward": 0.7988250851631165,
      "reward_std": 0.01421809894964099,
      "reward_std_all": 0.013907862361520529,
      "rewards-max/forensics_sinkhorn_reward": 0.9960869550704956,
      "rewards-max/forensics_strict_edge_reward": 0.6575055122375488,
      "rewards-max/forensics_tiou_reward": 0.9948238134384155,
      "rewards/forensics_sinkhorn_reward": 0.8465840518474579,
      "rewards/forensics_strict_edge_reward": 0.09169064462184906,
      "rewards/forensics_tiou_reward": 0.7049231231212616,
      "step": 1211
    },
    {
      "completion_length": 28.015625,
      "epoch": 3.71889400921659,
      "grad_norm": 6.475098609924316,
      "kl": 0.482421875,
      "learning_rate": 6.846153846153846e-08,
      "loss": -0.0125,
      "reward": 0.7934310734272003,
      "reward_std": 0.02199253672733903,
      "reward_std_all": 0.021115380339324474,
      "rewards-max/forensics_sinkhorn_reward": 0.9848334193229675,
      "rewards-max/forensics_strict_edge_reward": 0.6483443379402161,
      "rewards-max/forensics_tiou_reward": 0.9676065742969513,
      "rewards/forensics_sinkhorn_reward": 0.8172029554843903,
      "rewards/forensics_strict_edge_reward": 0.056088488548994064,
      "rewards/forensics_tiou_reward": 0.6491985023021698,
      "step": 1212
    },
    {
      "completion_length": 26.21875,
      "epoch": 3.7219662058371736,
      "grad_norm": 7.38319206237793,
      "kl": 0.4755859375,
      "learning_rate": 6.769230769230769e-08,
      "loss": -0.0139,
      "reward": 0.7811302244663239,
      "reward_std": 0.023954524658620358,
      "reward_std_all": 0.02523748856037855,
      "rewards-max/forensics_sinkhorn_reward": 0.9935097694396973,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9885905683040619,
      "rewards/forensics_sinkhorn_reward": 0.8266451358795166,
      "rewards/forensics_strict_edge_reward": 0.10373582690954208,
      "rewards/forensics_tiou_reward": 0.6549795567989349,
      "step": 1213
    },
    {
      "completion_length": 25.234375,
      "epoch": 3.7250384024577574,
      "grad_norm": 7.507135391235352,
      "kl": 0.4091796875,
      "learning_rate": 6.692307692307691e-08,
      "loss": -0.0297,
      "reward": 0.7709713280200958,
      "reward_std": 0.011807004455477,
      "reward_std_all": 0.02559872856363654,
      "rewards-max/forensics_sinkhorn_reward": 0.9958577752113342,
      "rewards-max/forensics_strict_edge_reward": 0.5612265467643738,
      "rewards-max/forensics_tiou_reward": 0.9948664903640747,
      "rewards/forensics_sinkhorn_reward": 0.7905269265174866,
      "rewards/forensics_strict_edge_reward": 0.07625898718833923,
      "rewards/forensics_tiou_reward": 0.5823879987001419,
      "step": 1214
    },
    {
      "completion_length": 24.828125,
      "epoch": 3.7281105990783407,
      "grad_norm": 7.60023307800293,
      "kl": 0.3583984375,
      "learning_rate": 6.615384615384615e-08,
      "loss": 0.0538,
      "reward": 0.797936886548996,
      "reward_std": 0.016188030713237822,
      "reward_std_all": 0.015446062665432692,
      "rewards-max/forensics_sinkhorn_reward": 0.9948537349700928,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9914529621601105,
      "rewards/forensics_sinkhorn_reward": 0.8567191362380981,
      "rewards/forensics_strict_edge_reward": 0.12354189157485962,
      "rewards/forensics_tiou_reward": 0.7230786383152008,
      "step": 1215
    },
    {
      "completion_length": 22.734375,
      "epoch": 3.731182795698925,
      "grad_norm": 7.2207794189453125,
      "kl": 0.486328125,
      "learning_rate": 6.538461538461538e-08,
      "loss": 0.0155,
      "reward": 0.7833257615566254,
      "reward_std": 0.04126307589467615,
      "reward_std_all": 0.037094816914759576,
      "rewards-max/forensics_sinkhorn_reward": 0.9945197105407715,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.9906553626060486,
      "rewards/forensics_sinkhorn_reward": 0.8231113851070404,
      "rewards/forensics_strict_edge_reward": 0.14903520792722702,
      "rewards/forensics_tiou_reward": 0.6612846255302429,
      "step": 1216
    },
    {
      "completion_length": 28.453125,
      "epoch": 3.7342549923195083,
      "grad_norm": 9.560161590576172,
      "kl": 0.3701171875,
      "learning_rate": 6.461538461538462e-08,
      "loss": 0.0085,
      "reward": 0.7563789486885071,
      "reward_std": 0.024329184787347913,
      "reward_std_all": 0.03697087662294507,
      "rewards-max/forensics_sinkhorn_reward": 0.9904236197471619,
      "rewards-max/forensics_strict_edge_reward": 0.5972513854503632,
      "rewards-max/forensics_tiou_reward": 0.9781650006771088,
      "rewards/forensics_sinkhorn_reward": 0.7676907479763031,
      "rewards/forensics_strict_edge_reward": 0.07616507820785046,
      "rewards/forensics_tiou_reward": 0.6033869087696075,
      "step": 1217
    },
    {
      "completion_length": 23.765625,
      "epoch": 3.737327188940092,
      "grad_norm": 11.796320915222168,
      "kl": 0.5048828125,
      "learning_rate": 6.384615384615385e-08,
      "loss": -0.0793,
      "reward": 0.796200305223465,
      "reward_std": 0.01906145690008998,
      "reward_std_all": 0.01809245301410556,
      "rewards-max/forensics_sinkhorn_reward": 0.9864270389080048,
      "rewards-max/forensics_strict_edge_reward": 0.5311143696308136,
      "rewards-max/forensics_tiou_reward": 0.9699999392032623,
      "rewards/forensics_sinkhorn_reward": 0.8802094757556915,
      "rewards/forensics_strict_edge_reward": 0.0839620940387249,
      "rewards/forensics_tiou_reward": 0.7384104132652283,
      "step": 1218
    },
    {
      "completion_length": 23.765625,
      "epoch": 3.740399385560676,
      "grad_norm": 7.198083400726318,
      "kl": 0.416015625,
      "learning_rate": 6.307692307692307e-08,
      "loss": -0.0628,
      "reward": 0.786143571138382,
      "reward_std": 0.016811595298349857,
      "reward_std_all": 0.018992583034560084,
      "rewards-max/forensics_sinkhorn_reward": 0.9802584052085876,
      "rewards-max/forensics_strict_edge_reward": 0.7178284525871277,
      "rewards-max/forensics_tiou_reward": 0.9603300094604492,
      "rewards/forensics_sinkhorn_reward": 0.7897323668003082,
      "rewards/forensics_strict_edge_reward": 0.0917203351855278,
      "rewards/forensics_tiou_reward": 0.600421667098999,
      "step": 1219
    },
    {
      "completion_length": 22.390625,
      "epoch": 3.7434715821812596,
      "grad_norm": 8.935758590698242,
      "kl": 0.3642578125,
      "learning_rate": 6.23076923076923e-08,
      "loss": -0.0093,
      "reward": 0.7892237603664398,
      "reward_std": 0.01515389559790492,
      "reward_std_all": 0.017136825248599052,
      "rewards-max/forensics_sinkhorn_reward": 0.9809115529060364,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9589508771896362,
      "rewards/forensics_sinkhorn_reward": 0.8036531209945679,
      "rewards/forensics_strict_edge_reward": 0.06913571059703827,
      "rewards/forensics_tiou_reward": 0.6312476992607117,
      "step": 1220
    },
    {
      "completion_length": 27.0,
      "epoch": 3.7465437788018434,
      "grad_norm": 9.135346412658691,
      "kl": 0.4267578125,
      "learning_rate": 6.153846153846154e-08,
      "loss": 0.0416,
      "reward": 0.7770334482192993,
      "reward_std": 0.03442533686757088,
      "reward_std_all": 0.033875368535518646,
      "rewards-max/forensics_sinkhorn_reward": 0.9847799837589264,
      "rewards-max/forensics_strict_edge_reward": 0.5984183251857758,
      "rewards-max/forensics_tiou_reward": 0.9658978283405304,
      "rewards/forensics_sinkhorn_reward": 0.7902141213417053,
      "rewards/forensics_strict_edge_reward": 0.08385561406612396,
      "rewards/forensics_tiou_reward": 0.6367964148521423,
      "step": 1221
    },
    {
      "completion_length": 22.5625,
      "epoch": 3.749615975422427,
      "grad_norm": 7.5983195304870605,
      "kl": 0.4501953125,
      "learning_rate": 6.076923076923077e-08,
      "loss": -0.0646,
      "reward": 0.7840430438518524,
      "reward_std": 0.02189692948013544,
      "reward_std_all": 0.02348380722105503,
      "rewards-max/forensics_sinkhorn_reward": 0.9863352477550507,
      "rewards-max/forensics_strict_edge_reward": 0.4169531762599945,
      "rewards-max/forensics_tiou_reward": 0.9704962968826294,
      "rewards/forensics_sinkhorn_reward": 0.7712957561016083,
      "rewards/forensics_strict_edge_reward": 0.04636244289577007,
      "rewards/forensics_tiou_reward": 0.6158427000045776,
      "step": 1222
    },
    {
      "completion_length": 24.359375,
      "epoch": 3.752688172043011,
      "grad_norm": 9.274062156677246,
      "kl": 0.400390625,
      "learning_rate": 6e-08,
      "loss": 0.0785,
      "reward": 0.7739333808422089,
      "reward_std": 0.04483056627213955,
      "reward_std_all": 0.04190354701131582,
      "rewards-max/forensics_sinkhorn_reward": 0.9947956204414368,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9920263886451721,
      "rewards/forensics_sinkhorn_reward": 0.7609679400920868,
      "rewards/forensics_strict_edge_reward": 0.11929181218147278,
      "rewards/forensics_tiou_reward": 0.5964071750640869,
      "step": 1223
    },
    {
      "completion_length": 23.578125,
      "epoch": 3.7557603686635943,
      "grad_norm": 8.295845031738281,
      "kl": 0.3837890625,
      "learning_rate": 5.923076923076923e-08,
      "loss": 0.0714,
      "reward": 0.7828924655914307,
      "reward_std": 0.042564354836940765,
      "reward_std_all": 0.038277642801404,
      "rewards-max/forensics_sinkhorn_reward": 0.9963209629058838,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9955156743526459,
      "rewards/forensics_sinkhorn_reward": 0.8277632594108582,
      "rewards/forensics_strict_edge_reward": 0.1606532782316208,
      "rewards/forensics_tiou_reward": 0.6809883415699005,
      "step": 1224
    },
    {
      "completion_length": 25.0625,
      "epoch": 3.758832565284178,
      "grad_norm": 9.983345985412598,
      "kl": 0.486328125,
      "learning_rate": 5.8461538461538455e-08,
      "loss": 0.0644,
      "reward": 0.7363189160823822,
      "reward_std": 0.050023261457681656,
      "reward_std_all": 0.05894077010452747,
      "rewards-max/forensics_sinkhorn_reward": 0.9977524876594543,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9993407428264618,
      "rewards/forensics_sinkhorn_reward": 0.7560495436191559,
      "rewards/forensics_strict_edge_reward": 0.12789233773946762,
      "rewards/forensics_tiou_reward": 0.6550677716732025,
      "step": 1225
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.761904761904762,
      "grad_norm": 8.149961471557617,
      "kl": 0.5419921875,
      "learning_rate": 5.7692307692307695e-08,
      "loss": -0.0467,
      "reward": 0.7961968779563904,
      "reward_std": 0.009839461708907038,
      "reward_std_all": 0.011514621321111917,
      "rewards-max/forensics_sinkhorn_reward": 0.9902666211128235,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9789577126502991,
      "rewards/forensics_sinkhorn_reward": 0.8347530364990234,
      "rewards/forensics_strict_edge_reward": 0.08891911804676056,
      "rewards/forensics_tiou_reward": 0.6462025046348572,
      "step": 1226
    },
    {
      "completion_length": 26.140625,
      "epoch": 3.7649769585253456,
      "grad_norm": 9.43168830871582,
      "kl": 0.490234375,
      "learning_rate": 5.692307692307692e-08,
      "loss": 0.0869,
      "reward": 0.7780584394931793,
      "reward_std": 0.03822644520550966,
      "reward_std_all": 0.03632968943566084,
      "rewards-max/forensics_sinkhorn_reward": 0.9893490076065063,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9793521761894226,
      "rewards/forensics_sinkhorn_reward": 0.8102033138275146,
      "rewards/forensics_strict_edge_reward": 0.0971156507730484,
      "rewards/forensics_tiou_reward": 0.6377863883972168,
      "step": 1227
    },
    {
      "completion_length": 25.359375,
      "epoch": 3.7680491551459294,
      "grad_norm": 8.24642276763916,
      "kl": 0.3779296875,
      "learning_rate": 5.615384615384615e-08,
      "loss": -0.0148,
      "reward": 0.7741650342941284,
      "reward_std": 0.0027044479502364993,
      "reward_std_all": 0.019291477277874947,
      "rewards-max/forensics_sinkhorn_reward": 0.9938775897026062,
      "rewards-max/forensics_strict_edge_reward": 0.5107362568378448,
      "rewards-max/forensics_tiou_reward": 0.9858244955539703,
      "rewards/forensics_sinkhorn_reward": 0.8178476393222809,
      "rewards/forensics_strict_edge_reward": 0.09256041795015335,
      "rewards/forensics_tiou_reward": 0.6691676378250122,
      "step": 1228
    },
    {
      "completion_length": 26.296875,
      "epoch": 3.771121351766513,
      "grad_norm": 9.95655632019043,
      "kl": 0.369140625,
      "learning_rate": 5.538461538461538e-08,
      "loss": 0.0063,
      "reward": 0.7925842106342316,
      "reward_std": 0.02135884016752243,
      "reward_std_all": 0.020735513418912888,
      "rewards-max/forensics_sinkhorn_reward": 0.9800558686256409,
      "rewards-max/forensics_strict_edge_reward": 0.6088831722736359,
      "rewards-max/forensics_tiou_reward": 0.9571420848369598,
      "rewards/forensics_sinkhorn_reward": 0.7999258041381836,
      "rewards/forensics_strict_edge_reward": 0.07904151361435652,
      "rewards/forensics_tiou_reward": 0.6090584993362427,
      "step": 1229
    },
    {
      "completion_length": 22.671875,
      "epoch": 3.774193548387097,
      "grad_norm": 10.45288372039795,
      "kl": 0.3984375,
      "learning_rate": 5.461538461538461e-08,
      "loss": -0.0289,
      "reward": 0.7878194153308868,
      "reward_std": 0.022038275375962257,
      "reward_std_all": 0.0224462253972888,
      "rewards-max/forensics_sinkhorn_reward": 0.9900052547454834,
      "rewards-max/forensics_strict_edge_reward": 0.7674418687820435,
      "rewards-max/forensics_tiou_reward": 0.9795551300048828,
      "rewards/forensics_sinkhorn_reward": 0.8020815849304199,
      "rewards/forensics_strict_edge_reward": 0.06814635545015335,
      "rewards/forensics_tiou_reward": 0.622392863035202,
      "step": 1230
    },
    {
      "completion_length": 28.421875,
      "epoch": 3.7772657450076803,
      "grad_norm": 21.047733306884766,
      "kl": 1.15625,
      "learning_rate": 5.384615384615385e-08,
      "loss": 0.1624,
      "reward": 0.7963716089725494,
      "reward_std": 0.01571217505261302,
      "reward_std_all": 0.015989442355930805,
      "rewards-max/forensics_sinkhorn_reward": 0.9918481111526489,
      "rewards-max/forensics_strict_edge_reward": 0.3375380411744118,
      "rewards-max/forensics_tiou_reward": 0.9835707247257233,
      "rewards/forensics_sinkhorn_reward": 0.7439163029193878,
      "rewards/forensics_strict_edge_reward": 0.05257073789834976,
      "rewards/forensics_tiou_reward": 0.6253985166549683,
      "step": 1231
    },
    {
      "completion_length": 22.625,
      "epoch": 3.780337941628264,
      "grad_norm": 10.665775299072266,
      "kl": 0.416015625,
      "learning_rate": 5.307692307692308e-08,
      "loss": 0.0151,
      "reward": 0.7900541424751282,
      "reward_std": 0.024004101753234863,
      "reward_std_all": 0.023031814955174923,
      "rewards-max/forensics_sinkhorn_reward": 0.9725640118122101,
      "rewards-max/forensics_strict_edge_reward": 0.5042839050292969,
      "rewards-max/forensics_tiou_reward": 0.938349574804306,
      "rewards/forensics_sinkhorn_reward": 0.7832517623901367,
      "rewards/forensics_strict_edge_reward": 0.06296109035611153,
      "rewards/forensics_tiou_reward": 0.54482302069664,
      "step": 1232
    },
    {
      "completion_length": 26.890625,
      "epoch": 3.783410138248848,
      "grad_norm": 7.381103515625,
      "kl": 0.587890625,
      "learning_rate": 5.2307692307692304e-08,
      "loss": -0.0055,
      "reward": 0.7826534509658813,
      "reward_std": 0.037030214443802834,
      "reward_std_all": 0.034225926734507084,
      "rewards-max/forensics_sinkhorn_reward": 0.9889337122440338,
      "rewards-max/forensics_strict_edge_reward": 0.5101855397224426,
      "rewards-max/forensics_tiou_reward": 0.9740275740623474,
      "rewards/forensics_sinkhorn_reward": 0.8025584518909454,
      "rewards/forensics_strict_edge_reward": 0.0979146659374237,
      "rewards/forensics_tiou_reward": 0.6611928939819336,
      "step": 1233
    },
    {
      "completion_length": 22.171875,
      "epoch": 3.7864823348694316,
      "grad_norm": 9.791524887084961,
      "kl": 0.51171875,
      "learning_rate": 5.153846153846154e-08,
      "loss": -0.0364,
      "reward": 0.7837867140769958,
      "reward_std": 0.03475348884239793,
      "reward_std_all": 0.03263899032026529,
      "rewards-max/forensics_sinkhorn_reward": 0.9961372911930084,
      "rewards-max/forensics_strict_edge_reward": 0.7506848573684692,
      "rewards-max/forensics_tiou_reward": 0.9944119155406952,
      "rewards/forensics_sinkhorn_reward": 0.8026801943778992,
      "rewards/forensics_strict_edge_reward": 0.07852134481072426,
      "rewards/forensics_tiou_reward": 0.6010577082633972,
      "step": 1234
    },
    {
      "completion_length": 25.78125,
      "epoch": 3.7895545314900154,
      "grad_norm": 9.102734565734863,
      "kl": 0.486328125,
      "learning_rate": 5.0769230769230764e-08,
      "loss": -0.0489,
      "reward": 0.757262647151947,
      "reward_std": 0.04609774053096771,
      "reward_std_all": 0.04712739586830139,
      "rewards-max/forensics_sinkhorn_reward": 0.9892920255661011,
      "rewards-max/forensics_strict_edge_reward": 0.5548095405101776,
      "rewards-max/forensics_tiou_reward": 0.9785680770874023,
      "rewards/forensics_sinkhorn_reward": 0.7770867943763733,
      "rewards/forensics_strict_edge_reward": 0.06739555299282074,
      "rewards/forensics_tiou_reward": 0.612655520439148,
      "step": 1235
    },
    {
      "completion_length": 25.59375,
      "epoch": 3.792626728110599,
      "grad_norm": 8.561155319213867,
      "kl": 0.529296875,
      "learning_rate": 5e-08,
      "loss": 0.0113,
      "reward": 0.778479129076004,
      "reward_std": 0.0463616109918803,
      "reward_std_all": 0.04223682824522257,
      "rewards-max/forensics_sinkhorn_reward": 0.9819446802139282,
      "rewards-max/forensics_strict_edge_reward": 0.6005697548389435,
      "rewards-max/forensics_tiou_reward": 0.9613420069217682,
      "rewards/forensics_sinkhorn_reward": 0.7994493246078491,
      "rewards/forensics_strict_edge_reward": 0.07632721401751041,
      "rewards/forensics_tiou_reward": 0.6615841686725616,
      "step": 1236
    },
    {
      "completion_length": 23.0,
      "epoch": 3.795698924731183,
      "grad_norm": 10.056875228881836,
      "kl": 0.3896484375,
      "learning_rate": 4.923076923076923e-08,
      "loss": -0.0183,
      "reward": 0.7997697591781616,
      "reward_std": 0.011857666075229645,
      "reward_std_all": 0.011998048285022378,
      "rewards-max/forensics_sinkhorn_reward": 0.9901491701602936,
      "rewards-max/forensics_strict_edge_reward": 0.5984183251857758,
      "rewards-max/forensics_tiou_reward": 0.9775896966457367,
      "rewards/forensics_sinkhorn_reward": 0.8309224545955658,
      "rewards/forensics_strict_edge_reward": 0.06120079755783081,
      "rewards/forensics_tiou_reward": 0.6803955733776093,
      "step": 1237
    },
    {
      "completion_length": 25.796875,
      "epoch": 3.7987711213517663,
      "grad_norm": 6.462705135345459,
      "kl": 0.505859375,
      "learning_rate": 4.846153846153846e-08,
      "loss": -0.0072,
      "reward": 0.7989087104797363,
      "reward_std": 0.010494537418708205,
      "reward_std_all": 0.011271702591329813,
      "rewards-max/forensics_sinkhorn_reward": 0.9924434423446655,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.986604630947113,
      "rewards/forensics_sinkhorn_reward": 0.8344148993492126,
      "rewards/forensics_strict_edge_reward": 0.0971156656742096,
      "rewards/forensics_tiou_reward": 0.6906367242336273,
      "step": 1238
    },
    {
      "completion_length": 27.296875,
      "epoch": 3.80184331797235,
      "grad_norm": 7.838397979736328,
      "kl": 0.4755859375,
      "learning_rate": 4.769230769230769e-08,
      "loss": 0.0061,
      "reward": 0.7971764504909515,
      "reward_std": 0.010544416960328817,
      "reward_std_all": 0.011750136967748404,
      "rewards-max/forensics_sinkhorn_reward": 0.9825315177440643,
      "rewards-max/forensics_strict_edge_reward": 0.2781495377421379,
      "rewards-max/forensics_tiou_reward": 0.9634258151054382,
      "rewards/forensics_sinkhorn_reward": 0.7872673869132996,
      "rewards/forensics_strict_edge_reward": 0.04131313506513834,
      "rewards/forensics_tiou_reward": 0.5851119160652161,
      "step": 1239
    },
    {
      "completion_length": 21.359375,
      "epoch": 3.804915514592934,
      "grad_norm": 11.477259635925293,
      "kl": 0.560546875,
      "learning_rate": 4.692307692307692e-08,
      "loss": 0.0584,
      "reward": 0.777825117111206,
      "reward_std": 0.03376724012196064,
      "reward_std_all": 0.03324185870587826,
      "rewards-max/forensics_sinkhorn_reward": 0.991807758808136,
      "rewards-max/forensics_strict_edge_reward": 0.5674395263195038,
      "rewards-max/forensics_tiou_reward": 0.9848537445068359,
      "rewards/forensics_sinkhorn_reward": 0.8052213788032532,
      "rewards/forensics_strict_edge_reward": 0.06491761654615402,
      "rewards/forensics_tiou_reward": 0.6085739135742188,
      "step": 1240
    },
    {
      "completion_length": 24.3125,
      "epoch": 3.8079877112135176,
      "grad_norm": 9.077164649963379,
      "kl": 0.4365234375,
      "learning_rate": 4.615384615384615e-08,
      "loss": 0.0346,
      "reward": 0.765352874994278,
      "reward_std": 0.01465047337114811,
      "reward_std_all": 0.02840124163776636,
      "rewards-max/forensics_sinkhorn_reward": 0.9926900565624237,
      "rewards-max/forensics_strict_edge_reward": 0.5391899347305298,
      "rewards-max/forensics_tiou_reward": 0.9848264753818512,
      "rewards/forensics_sinkhorn_reward": 0.7761728763580322,
      "rewards/forensics_strict_edge_reward": 0.06000255234539509,
      "rewards/forensics_tiou_reward": 0.6069802343845367,
      "step": 1241
    },
    {
      "completion_length": 23.1875,
      "epoch": 3.8110599078341014,
      "grad_norm": 7.393943786621094,
      "kl": 0.45703125,
      "learning_rate": 4.538461538461538e-08,
      "loss": -0.0166,
      "reward": 0.7883204817771912,
      "reward_std": 0.005127302138134837,
      "reward_std_all": 0.013331492897123098,
      "rewards-max/forensics_sinkhorn_reward": 0.9849313497543335,
      "rewards-max/forensics_strict_edge_reward": 0.6176230013370514,
      "rewards-max/forensics_tiou_reward": 0.9661193192005157,
      "rewards/forensics_sinkhorn_reward": 0.8089296221733093,
      "rewards/forensics_strict_edge_reward": 0.09950578957796097,
      "rewards/forensics_tiou_reward": 0.6184390485286713,
      "step": 1242
    },
    {
      "completion_length": 24.78125,
      "epoch": 3.814132104454685,
      "grad_norm": 7.758774280548096,
      "kl": 0.3916015625,
      "learning_rate": 4.461538461538461e-08,
      "loss": 0.0686,
      "reward": 0.7994198501110077,
      "reward_std": 0.012539941468276083,
      "reward_std_all": 0.012477305484935641,
      "rewards-max/forensics_sinkhorn_reward": 0.9975062608718872,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8046832978725433,
      "rewards/forensics_strict_edge_reward": 0.11113516986370087,
      "rewards/forensics_tiou_reward": 0.6589961051940918,
      "step": 1243
    },
    {
      "completion_length": 25.703125,
      "epoch": 3.817204301075269,
      "grad_norm": 9.155503273010254,
      "kl": 0.462890625,
      "learning_rate": 4.3846153846153846e-08,
      "loss": -0.0563,
      "reward": 0.7874863743782043,
      "reward_std": 0.028543567284941673,
      "reward_std_all": 0.02703243400901556,
      "rewards-max/forensics_sinkhorn_reward": 0.9936862289905548,
      "rewards-max/forensics_strict_edge_reward": 0.5843779891729355,
      "rewards-max/forensics_tiou_reward": 0.9858677685260773,
      "rewards/forensics_sinkhorn_reward": 0.8075529932975769,
      "rewards/forensics_strict_edge_reward": 0.0669440720230341,
      "rewards/forensics_tiou_reward": 0.650489866733551,
      "step": 1244
    },
    {
      "completion_length": 33.34375,
      "epoch": 3.8202764976958523,
      "grad_norm": 8.02127742767334,
      "kl": 0.396484375,
      "learning_rate": 4.307692307692307e-08,
      "loss": 0.0167,
      "reward": 0.7842919230461121,
      "reward_std": 0.025446977466344833,
      "reward_std_all": 0.025729904882609844,
      "rewards-max/forensics_sinkhorn_reward": 0.9769529104232788,
      "rewards-max/forensics_strict_edge_reward": 0.6193071901798248,
      "rewards-max/forensics_tiou_reward": 0.9407703876495361,
      "rewards/forensics_sinkhorn_reward": 0.7330162525177002,
      "rewards/forensics_strict_edge_reward": 0.08240185678005219,
      "rewards/forensics_tiou_reward": 0.6251368820667267,
      "step": 1245
    },
    {
      "completion_length": 23.296875,
      "epoch": 3.823348694316436,
      "grad_norm": 9.166569709777832,
      "kl": 0.5029296875,
      "learning_rate": 4.230769230769231e-08,
      "loss": -0.0264,
      "reward": 0.7990696728229523,
      "reward_std": 0.013948855455964804,
      "reward_std_all": 0.013508239760994911,
      "rewards-max/forensics_sinkhorn_reward": 0.9975062608718872,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8407327234745026,
      "rewards/forensics_strict_edge_reward": 0.18014436960220337,
      "rewards/forensics_tiou_reward": 0.6608976721763611,
      "step": 1246
    },
    {
      "completion_length": 27.5625,
      "epoch": 3.82642089093702,
      "grad_norm": 8.315454483032227,
      "kl": 0.4462890625,
      "learning_rate": 4.1538461538461534e-08,
      "loss": -0.0531,
      "reward": 0.7937993109226227,
      "reward_std": 0.014178239274770021,
      "reward_std_all": 0.01533144898712635,
      "rewards-max/forensics_sinkhorn_reward": 0.9826032221317291,
      "rewards-max/forensics_strict_edge_reward": 0.5834526717662811,
      "rewards-max/forensics_tiou_reward": 0.9610188901424408,
      "rewards/forensics_sinkhorn_reward": 0.775272011756897,
      "rewards/forensics_strict_edge_reward": 0.0640573538839817,
      "rewards/forensics_tiou_reward": 0.618560254573822,
      "step": 1247
    },
    {
      "completion_length": 27.203125,
      "epoch": 3.8294930875576036,
      "grad_norm": 9.290385246276855,
      "kl": 0.4873046875,
      "learning_rate": 4.076923076923077e-08,
      "loss": -0.0338,
      "reward": 0.7824415266513824,
      "reward_std": 0.02739140670746565,
      "reward_std_all": 0.027552583254873753,
      "rewards-max/forensics_sinkhorn_reward": 0.9890853762626648,
      "rewards-max/forensics_strict_edge_reward": 0.42900510132312775,
      "rewards-max/forensics_tiou_reward": 0.9756375253200531,
      "rewards/forensics_sinkhorn_reward": 0.8077446222305298,
      "rewards/forensics_strict_edge_reward": 0.07624420523643494,
      "rewards/forensics_tiou_reward": 0.6602638065814972,
      "step": 1248
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.8325652841781874,
      "grad_norm": 8.29442024230957,
      "kl": 0.408203125,
      "learning_rate": 4e-08,
      "loss": 0.0673,
      "reward": 0.7645782232284546,
      "reward_std": 0.02055320586077869,
      "reward_std_all": 0.031194807961583138,
      "rewards-max/forensics_sinkhorn_reward": 0.9770208597183228,
      "rewards-max/forensics_strict_edge_reward": 0.6876175105571747,
      "rewards-max/forensics_tiou_reward": 0.9415213465690613,
      "rewards/forensics_sinkhorn_reward": 0.7728178799152374,
      "rewards/forensics_strict_edge_reward": 0.12519191205501556,
      "rewards/forensics_tiou_reward": 0.6281894445419312,
      "step": 1249
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.835637480798771,
      "grad_norm": 7.743716239929199,
      "kl": 0.37890625,
      "learning_rate": 3.923076923076923e-08,
      "loss": -0.0256,
      "reward": 0.7909063696861267,
      "reward_std": 0.017123552970588207,
      "reward_std_all": 0.017961235251277685,
      "rewards-max/forensics_sinkhorn_reward": 0.9876901209354401,
      "rewards-max/forensics_strict_edge_reward": 0.4698307514190674,
      "rewards-max/forensics_tiou_reward": 0.9708677530288696,
      "rewards/forensics_sinkhorn_reward": 0.8120226263999939,
      "rewards/forensics_strict_edge_reward": 0.0775524377822876,
      "rewards/forensics_tiou_reward": 0.6867555975914001,
      "step": 1250
    },
    {
      "completion_length": 23.96875,
      "epoch": 3.838709677419355,
      "grad_norm": 7.667026996612549,
      "kl": 0.4130859375,
      "learning_rate": 3.846153846153846e-08,
      "loss": -0.0354,
      "reward": 0.7893778681755066,
      "reward_std": 0.016574406705331057,
      "reward_std_all": 0.017937145428732038,
      "rewards-max/forensics_sinkhorn_reward": 0.9851695895195007,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9719439744949341,
      "rewards/forensics_sinkhorn_reward": 0.8180490136146545,
      "rewards/forensics_strict_edge_reward": 0.12749784253537655,
      "rewards/forensics_tiou_reward": 0.6496736109256744,
      "step": 1251
    },
    {
      "completion_length": 25.796875,
      "epoch": 3.8417818740399383,
      "grad_norm": 8.272993087768555,
      "kl": 0.4228515625,
      "learning_rate": 3.769230769230769e-08,
      "loss": -0.0014,
      "reward": 0.7766313552856445,
      "reward_std": 0.014293127227574587,
      "reward_std_all": 0.023556102998554707,
      "rewards-max/forensics_sinkhorn_reward": 0.9907663464546204,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9833047688007355,
      "rewards/forensics_sinkhorn_reward": 0.7882486581802368,
      "rewards/forensics_strict_edge_reward": 0.0943823792040348,
      "rewards/forensics_tiou_reward": 0.6451120972633362,
      "step": 1252
    },
    {
      "completion_length": 28.4375,
      "epoch": 3.8448540706605225,
      "grad_norm": 9.489771842956543,
      "kl": 0.6572265625,
      "learning_rate": 3.692307692307692e-08,
      "loss": -0.0029,
      "reward": 0.7912567853927612,
      "reward_std": 0.014034017222002149,
      "reward_std_all": 0.015711473068222404,
      "rewards-max/forensics_sinkhorn_reward": 0.9875870943069458,
      "rewards-max/forensics_strict_edge_reward": 0.4630069434642792,
      "rewards-max/forensics_tiou_reward": 0.9706440567970276,
      "rewards/forensics_sinkhorn_reward": 0.8185910880565643,
      "rewards/forensics_strict_edge_reward": 0.06333359889686108,
      "rewards/forensics_tiou_reward": 0.7004522979259491,
      "step": 1253
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.847926267281106,
      "grad_norm": 10.620059967041016,
      "kl": 0.458984375,
      "learning_rate": 3.6153846153846155e-08,
      "loss": 0.0549,
      "reward": 0.7826105952262878,
      "reward_std": 0.03233077749609947,
      "reward_std_all": 0.030971995554864407,
      "rewards-max/forensics_sinkhorn_reward": 0.9826180934906006,
      "rewards-max/forensics_strict_edge_reward": 0.6793017089366913,
      "rewards-max/forensics_tiou_reward": 0.9641790986061096,
      "rewards/forensics_sinkhorn_reward": 0.7693797647953033,
      "rewards/forensics_strict_edge_reward": 0.07851315289735794,
      "rewards/forensics_tiou_reward": 0.5812861621379852,
      "step": 1254
    },
    {
      "completion_length": 27.53125,
      "epoch": 3.8509984639016897,
      "grad_norm": 7.551820755004883,
      "kl": 0.4580078125,
      "learning_rate": 3.538461538461538e-08,
      "loss": 0.0292,
      "reward": 0.7636423408985138,
      "reward_std": 0.03465949557721615,
      "reward_std_all": 0.04271678160876036,
      "rewards-max/forensics_sinkhorn_reward": 0.9888536036014557,
      "rewards-max/forensics_strict_edge_reward": 0.6720768958330154,
      "rewards-max/forensics_tiou_reward": 0.9756952226161957,
      "rewards/forensics_sinkhorn_reward": 0.733671247959137,
      "rewards/forensics_strict_edge_reward": 0.06308982521295547,
      "rewards/forensics_tiou_reward": 0.596645712852478,
      "step": 1255
    },
    {
      "completion_length": 22.921875,
      "epoch": 3.8540706605222734,
      "grad_norm": 9.218559265136719,
      "kl": 0.373046875,
      "learning_rate": 3.4615384615384616e-08,
      "loss": -0.1358,
      "reward": 0.7938461899757385,
      "reward_std": 0.02167472871951759,
      "reward_std_all": 0.02080395701341331,
      "rewards-max/forensics_sinkhorn_reward": 0.9946620464324951,
      "rewards-max/forensics_strict_edge_reward": 0.709127277135849,
      "rewards-max/forensics_tiou_reward": 0.9914138317108154,
      "rewards/forensics_sinkhorn_reward": 0.8224487602710724,
      "rewards/forensics_strict_edge_reward": 0.07864217460155487,
      "rewards/forensics_tiou_reward": 0.6702917218208313,
      "step": 1256
    },
    {
      "completion_length": 25.921875,
      "epoch": 3.857142857142857,
      "grad_norm": 7.645125389099121,
      "kl": 0.412109375,
      "learning_rate": 3.384615384615384e-08,
      "loss": -0.0381,
      "reward": 0.7576587796211243,
      "reward_std": 0.04347812198102474,
      "reward_std_all": 0.04492790810763836,
      "rewards-max/forensics_sinkhorn_reward": 0.9889513254165649,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.975479006767273,
      "rewards/forensics_sinkhorn_reward": 0.7685823440551758,
      "rewards/forensics_strict_edge_reward": 0.10627714544534683,
      "rewards/forensics_tiou_reward": 0.618387758731842,
      "step": 1257
    },
    {
      "completion_length": 27.765625,
      "epoch": 3.860215053763441,
      "grad_norm": 7.4307966232299805,
      "kl": 0.564453125,
      "learning_rate": 3.3076923076923076e-08,
      "loss": 0.0051,
      "reward": 0.7745763659477234,
      "reward_std": 0.03100740537047386,
      "reward_std_all": 0.03186853229999542,
      "rewards-max/forensics_sinkhorn_reward": 0.9953236877918243,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9909237027168274,
      "rewards/forensics_sinkhorn_reward": 0.7781278491020203,
      "rewards/forensics_strict_edge_reward": 0.08803950995206833,
      "rewards/forensics_tiou_reward": 0.6210696399211884,
      "step": 1258
    },
    {
      "completion_length": 24.234375,
      "epoch": 3.8632872503840243,
      "grad_norm": 7.191019058227539,
      "kl": 0.3798828125,
      "learning_rate": 3.230769230769231e-08,
      "loss": -0.0703,
      "reward": 0.8051435053348541,
      "reward_std": 0.004053477663546801,
      "reward_std_all": 0.005070410203188658,
      "rewards-max/forensics_sinkhorn_reward": 0.9924172759056091,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9824097156524658,
      "rewards/forensics_sinkhorn_reward": 0.8794364333152771,
      "rewards/forensics_strict_edge_reward": 0.11113644391298294,
      "rewards/forensics_tiou_reward": 0.7710019946098328,
      "step": 1259
    },
    {
      "completion_length": 22.828125,
      "epoch": 3.8663594470046085,
      "grad_norm": 8.164505004882812,
      "kl": 0.39453125,
      "learning_rate": 3.153846153846154e-08,
      "loss": -0.0641,
      "reward": 0.7962862253189087,
      "reward_std": 0.008804182056337595,
      "reward_std_all": 0.0113940155133605,
      "rewards-max/forensics_sinkhorn_reward": 0.9900071024894714,
      "rewards-max/forensics_strict_edge_reward": 0.7490177750587463,
      "rewards-max/forensics_tiou_reward": 0.9776994585990906,
      "rewards/forensics_sinkhorn_reward": 0.7921700477600098,
      "rewards/forensics_strict_edge_reward": 0.06684285961091518,
      "rewards/forensics_tiou_reward": 0.5711273849010468,
      "step": 1260
    },
    {
      "completion_length": 27.484375,
      "epoch": 3.869431643625192,
      "grad_norm": 7.902182102203369,
      "kl": 0.54296875,
      "learning_rate": 3.076923076923077e-08,
      "loss": 0.0184,
      "reward": 0.7700240015983582,
      "reward_std": 0.01363240135833621,
      "reward_std_all": 0.02707228623330593,
      "rewards-max/forensics_sinkhorn_reward": 0.9905374348163605,
      "rewards-max/forensics_strict_edge_reward": 0.7617588341236115,
      "rewards-max/forensics_tiou_reward": 0.9802878797054291,
      "rewards/forensics_sinkhorn_reward": 0.8228224217891693,
      "rewards/forensics_strict_edge_reward": 0.11576089262962341,
      "rewards/forensics_tiou_reward": 0.7179145216941833,
      "step": 1261
    },
    {
      "completion_length": 25.46875,
      "epoch": 3.8725038402457757,
      "grad_norm": 10.263197898864746,
      "kl": 0.5634765625,
      "learning_rate": 3e-08,
      "loss": 0.01,
      "reward": 0.7853931784629822,
      "reward_std": 0.023527728160843253,
      "reward_std_all": 0.02516356250271201,
      "rewards-max/forensics_sinkhorn_reward": 0.9966751933097839,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9967261254787445,
      "rewards/forensics_sinkhorn_reward": 0.8145299255847931,
      "rewards/forensics_strict_edge_reward": 0.18414414674043655,
      "rewards/forensics_tiou_reward": 0.7210302948951721,
      "step": 1262
    },
    {
      "completion_length": 23.203125,
      "epoch": 3.8755760368663594,
      "grad_norm": 7.763352870941162,
      "kl": 0.486328125,
      "learning_rate": 2.9230769230769228e-08,
      "loss": -0.0011,
      "reward": 0.7934602797031403,
      "reward_std": 0.023492758627980947,
      "reward_std_all": 0.022111650556325912,
      "rewards-max/forensics_sinkhorn_reward": 0.988583892583847,
      "rewards-max/forensics_strict_edge_reward": 0.5578442513942719,
      "rewards-max/forensics_tiou_reward": 0.9787276983261108,
      "rewards/forensics_sinkhorn_reward": 0.8451881110668182,
      "rewards/forensics_strict_edge_reward": 0.11159303039312363,
      "rewards/forensics_tiou_reward": 0.6873533129692078,
      "step": 1263
    },
    {
      "completion_length": 26.578125,
      "epoch": 3.878648233486943,
      "grad_norm": 9.151500701904297,
      "kl": 0.533203125,
      "learning_rate": 2.846153846153846e-08,
      "loss": 0.029,
      "reward": 0.7886808812618256,
      "reward_std": 0.02181830955669284,
      "reward_std_all": 0.021964428946375847,
      "rewards-max/forensics_sinkhorn_reward": 0.9913798272609711,
      "rewards-max/forensics_strict_edge_reward": 0.8106843829154968,
      "rewards-max/forensics_tiou_reward": 0.9860953092575073,
      "rewards/forensics_sinkhorn_reward": 0.7957585752010345,
      "rewards/forensics_strict_edge_reward": 0.07802742719650269,
      "rewards/forensics_tiou_reward": 0.7048030495643616,
      "step": 1264
    },
    {
      "completion_length": 26.0625,
      "epoch": 3.881720430107527,
      "grad_norm": 8.84993839263916,
      "kl": 0.40625,
      "learning_rate": 2.769230769230769e-08,
      "loss": 0.0159,
      "reward": 0.8003213703632355,
      "reward_std": 0.011329937377013266,
      "reward_std_all": 0.011727990000508726,
      "rewards-max/forensics_sinkhorn_reward": 0.997003585100174,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9973287284374237,
      "rewards/forensics_sinkhorn_reward": 0.8501895070075989,
      "rewards/forensics_strict_edge_reward": 0.09146744757890701,
      "rewards/forensics_tiou_reward": 0.7097805142402649,
      "step": 1265
    },
    {
      "completion_length": 22.375,
      "epoch": 3.8847926267281108,
      "grad_norm": 11.414834976196289,
      "kl": 0.5849609375,
      "learning_rate": 2.6923076923076925e-08,
      "loss": -0.056,
      "reward": 0.7838806807994843,
      "reward_std": 0.024749007541686296,
      "reward_std_all": 0.025271114660426974,
      "rewards-max/forensics_sinkhorn_reward": 0.9849526286125183,
      "rewards-max/forensics_strict_edge_reward": 0.5431406646966934,
      "rewards-max/forensics_tiou_reward": 0.9651418626308441,
      "rewards/forensics_sinkhorn_reward": 0.7930193543434143,
      "rewards/forensics_strict_edge_reward": 0.06486480124294758,
      "rewards/forensics_tiou_reward": 0.6165990084409714,
      "step": 1266
    },
    {
      "completion_length": 24.75,
      "epoch": 3.8878648233486945,
      "grad_norm": 7.5964274406433105,
      "kl": 0.55078125,
      "learning_rate": 2.6153846153846152e-08,
      "loss": 0.0013,
      "reward": 0.7900874018669128,
      "reward_std": 0.02918765041977167,
      "reward_std_all": 0.026952922344207764,
      "rewards-max/forensics_sinkhorn_reward": 0.9936779141426086,
      "rewards-max/forensics_strict_edge_reward": 0.8979261517524719,
      "rewards-max/forensics_tiou_reward": 0.9890785217285156,
      "rewards/forensics_sinkhorn_reward": 0.8497303128242493,
      "rewards/forensics_strict_edge_reward": 0.08403681591153145,
      "rewards/forensics_tiou_reward": 0.6901721358299255,
      "step": 1267
    },
    {
      "completion_length": 26.984375,
      "epoch": 3.890937019969278,
      "grad_norm": 7.54554557800293,
      "kl": 0.478515625,
      "learning_rate": 2.5384615384615382e-08,
      "loss": -0.0454,
      "reward": 0.800290048122406,
      "reward_std": 0.01187112694606185,
      "reward_std_all": 0.011842655832879245,
      "rewards-max/forensics_sinkhorn_reward": 0.9868377447128296,
      "rewards-max/forensics_strict_edge_reward": 0.6720768958330154,
      "rewards-max/forensics_tiou_reward": 0.9709805846214294,
      "rewards/forensics_sinkhorn_reward": 0.8441157937049866,
      "rewards/forensics_strict_edge_reward": 0.11909019947052002,
      "rewards/forensics_tiou_reward": 0.6851454079151154,
      "step": 1268
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.8940092165898617,
      "grad_norm": 9.368551254272461,
      "kl": 0.4833984375,
      "learning_rate": 2.4615384615384616e-08,
      "loss": 0.0063,
      "reward": 0.7981778085231781,
      "reward_std": 0.011042251717299223,
      "reward_std_all": 0.011758427834138274,
      "rewards-max/forensics_sinkhorn_reward": 0.9899493753910065,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9817504286766052,
      "rewards/forensics_sinkhorn_reward": 0.8050820827484131,
      "rewards/forensics_strict_edge_reward": 0.08762291632592678,
      "rewards/forensics_tiou_reward": 0.6073541045188904,
      "step": 1269
    },
    {
      "completion_length": 23.78125,
      "epoch": 3.8970814132104454,
      "grad_norm": 8.766160011291504,
      "kl": 0.3681640625,
      "learning_rate": 2.3846153846153846e-08,
      "loss": 0.0261,
      "reward": 0.7971271574497223,
      "reward_std": 0.014637061394751072,
      "reward_std_all": 0.01474507013335824,
      "rewards-max/forensics_sinkhorn_reward": 0.9908685386180878,
      "rewards-max/forensics_strict_edge_reward": 0.6317985653877258,
      "rewards-max/forensics_tiou_reward": 0.9746728837490082,
      "rewards/forensics_sinkhorn_reward": 0.8324514627456665,
      "rewards/forensics_strict_edge_reward": 0.060219901613891125,
      "rewards/forensics_tiou_reward": 0.6961268782615662,
      "step": 1270
    },
    {
      "completion_length": 26.703125,
      "epoch": 3.900153609831029,
      "grad_norm": 8.744043350219727,
      "kl": 0.4052734375,
      "learning_rate": 2.3076923076923076e-08,
      "loss": -0.0669,
      "reward": 0.7491518259048462,
      "reward_std": 0.018943394999951124,
      "reward_std_all": 0.039348334074020386,
      "rewards-max/forensics_sinkhorn_reward": 0.988890677690506,
      "rewards-max/forensics_strict_edge_reward": 0.5548860281705856,
      "rewards-max/forensics_tiou_reward": 0.9757266044616699,
      "rewards/forensics_sinkhorn_reward": 0.7769257128238678,
      "rewards/forensics_strict_edge_reward": 0.10036935657262802,
      "rewards/forensics_tiou_reward": 0.629296749830246,
      "step": 1271
    },
    {
      "completion_length": 26.046875,
      "epoch": 3.903225806451613,
      "grad_norm": 8.269939422607422,
      "kl": 0.638671875,
      "learning_rate": 2.2307692307692306e-08,
      "loss": -0.0405,
      "reward": 0.7703437209129333,
      "reward_std": 0.022531189490109682,
      "reward_std_all": 0.02876087184995413,
      "rewards-max/forensics_sinkhorn_reward": 0.9674332737922668,
      "rewards-max/forensics_strict_edge_reward": 0.37400607019662857,
      "rewards-max/forensics_tiou_reward": 0.9242128133773804,
      "rewards/forensics_sinkhorn_reward": 0.7914668917655945,
      "rewards/forensics_strict_edge_reward": 0.049727984704077244,
      "rewards/forensics_tiou_reward": 0.6310878396034241,
      "step": 1272
    },
    {
      "completion_length": 26.53125,
      "epoch": 3.9062980030721968,
      "grad_norm": 7.4813714027404785,
      "kl": 0.384765625,
      "learning_rate": 2.1538461538461537e-08,
      "loss": -0.0927,
      "reward": 0.8018041551113129,
      "reward_std": 0.00864611309953034,
      "reward_std_all": 0.009298345074057579,
      "rewards-max/forensics_sinkhorn_reward": 0.9934934973716736,
      "rewards-max/forensics_strict_edge_reward": 0.6137524247169495,
      "rewards-max/forensics_tiou_reward": 0.9878499507904053,
      "rewards/forensics_sinkhorn_reward": 0.8570546507835388,
      "rewards/forensics_strict_edge_reward": 0.09783534333109856,
      "rewards/forensics_tiou_reward": 0.7117400765419006,
      "step": 1273
    },
    {
      "completion_length": 26.203125,
      "epoch": 3.9093701996927805,
      "grad_norm": 8.500452041625977,
      "kl": 0.5693359375,
      "learning_rate": 2.0769230769230767e-08,
      "loss": 0.0281,
      "reward": 0.7747678160667419,
      "reward_std": 0.02139795944094658,
      "reward_std_all": 0.028280727099627256,
      "rewards-max/forensics_sinkhorn_reward": 0.9907922744750977,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9814368486404419,
      "rewards/forensics_sinkhorn_reward": 0.8025031685829163,
      "rewards/forensics_strict_edge_reward": 0.08408531174063683,
      "rewards/forensics_tiou_reward": 0.6302829682826996,
      "step": 1274
    },
    {
      "completion_length": 23.4375,
      "epoch": 3.912442396313364,
      "grad_norm": 8.944900512695312,
      "kl": 0.599609375,
      "learning_rate": 2e-08,
      "loss": -0.1054,
      "reward": 0.8060656189918518,
      "reward_std": 0.0023112326161935925,
      "reward_std_all": 0.003540464793331921,
      "rewards-max/forensics_sinkhorn_reward": 0.9938559234142303,
      "rewards-max/forensics_strict_edge_reward": 0.8414600193500519,
      "rewards-max/forensics_tiou_reward": 0.9904785752296448,
      "rewards/forensics_sinkhorn_reward": 0.8490429520606995,
      "rewards/forensics_strict_edge_reward": 0.1461993157863617,
      "rewards/forensics_tiou_reward": 0.7175088226795197,
      "step": 1275
    },
    {
      "completion_length": 26.59375,
      "epoch": 3.9155145929339477,
      "grad_norm": 10.029511451721191,
      "kl": 0.4658203125,
      "learning_rate": 1.923076923076923e-08,
      "loss": -0.0582,
      "reward": 0.7993696630001068,
      "reward_std": 0.007078394293785095,
      "reward_std_all": 0.008848156780004501,
      "rewards-max/forensics_sinkhorn_reward": 0.9979211688041687,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8441502153873444,
      "rewards/forensics_strict_edge_reward": 0.13234422728419304,
      "rewards/forensics_tiou_reward": 0.6748390793800354,
      "step": 1276
    },
    {
      "completion_length": 22.796875,
      "epoch": 3.9185867895545314,
      "grad_norm": 8.043428421020508,
      "kl": 0.4853515625,
      "learning_rate": 1.846153846153846e-08,
      "loss": 0.0174,
      "reward": 0.8016156256198883,
      "reward_std": 0.006000805646181107,
      "reward_std_all": 0.007523415144532919,
      "rewards-max/forensics_sinkhorn_reward": 0.9829566478729248,
      "rewards-max/forensics_strict_edge_reward": 0.6947998404502869,
      "rewards-max/forensics_tiou_reward": 0.9608315527439117,
      "rewards/forensics_sinkhorn_reward": 0.8347029089927673,
      "rewards/forensics_strict_edge_reward": 0.08681780844926834,
      "rewards/forensics_tiou_reward": 0.6498721241950989,
      "step": 1277
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.921658986175115,
      "grad_norm": 10.55621337890625,
      "kl": 0.7890625,
      "learning_rate": 1.769230769230769e-08,
      "loss": 0.0415,
      "reward": 0.7805726826190948,
      "reward_std": 0.029440411366522312,
      "reward_std_all": 0.029381120577454567,
      "rewards-max/forensics_sinkhorn_reward": 0.9913581907749176,
      "rewards-max/forensics_strict_edge_reward": 0.6526533365249634,
      "rewards-max/forensics_tiou_reward": 0.9841774106025696,
      "rewards/forensics_sinkhorn_reward": 0.7663626074790955,
      "rewards/forensics_strict_edge_reward": 0.0860513336956501,
      "rewards/forensics_tiou_reward": 0.6875113844871521,
      "step": 1278
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.924731182795699,
      "grad_norm": 9.087336540222168,
      "kl": 0.4482421875,
      "learning_rate": 1.692307692307692e-08,
      "loss": 0.0355,
      "reward": 0.7769989967346191,
      "reward_std": 0.04426347650587559,
      "reward_std_all": 0.04099462181329727,
      "rewards-max/forensics_sinkhorn_reward": 0.9680412411689758,
      "rewards-max/forensics_strict_edge_reward": 0.4279094636440277,
      "rewards-max/forensics_tiou_reward": 0.9279996156692505,
      "rewards/forensics_sinkhorn_reward": 0.7920268774032593,
      "rewards/forensics_strict_edge_reward": 0.05595143884420395,
      "rewards/forensics_tiou_reward": 0.6269679069519043,
      "step": 1279
    },
    {
      "completion_length": 27.15625,
      "epoch": 3.9278033794162828,
      "grad_norm": 8.557397842407227,
      "kl": 0.6328125,
      "learning_rate": 1.6153846153846155e-08,
      "loss": -0.0099,
      "reward": 0.7995439767837524,
      "reward_std": 0.013774006976746023,
      "reward_std_all": 0.013373378664255142,
      "rewards-max/forensics_sinkhorn_reward": 0.9927177131175995,
      "rewards-max/forensics_strict_edge_reward": 0.7366199493408203,
      "rewards-max/forensics_tiou_reward": 0.9865093529224396,
      "rewards/forensics_sinkhorn_reward": 0.8481844067573547,
      "rewards/forensics_strict_edge_reward": 0.1357147917151451,
      "rewards/forensics_tiou_reward": 0.7407784759998322,
      "step": 1280
    },
    {
      "completion_length": 28.4375,
      "epoch": 3.9308755760368665,
      "grad_norm": 7.787254333496094,
      "kl": 0.44140625,
      "learning_rate": 1.5384615384615385e-08,
      "loss": -0.0752,
      "reward": 0.7972925305366516,
      "reward_std": 0.015655426890589297,
      "reward_std_all": 0.015645474893972278,
      "rewards-max/forensics_sinkhorn_reward": 0.9823940396308899,
      "rewards-max/forensics_strict_edge_reward": 0.3506074771285057,
      "rewards-max/forensics_tiou_reward": 0.9620687365531921,
      "rewards/forensics_sinkhorn_reward": 0.8061332404613495,
      "rewards/forensics_strict_edge_reward": 0.0683746263384819,
      "rewards/forensics_tiou_reward": 0.6904308795928955,
      "step": 1281
    },
    {
      "completion_length": 27.453125,
      "epoch": 3.93394777265745,
      "grad_norm": 7.497644901275635,
      "kl": 0.4140625,
      "learning_rate": 1.4615384615384614e-08,
      "loss": -0.0528,
      "reward": 0.7733540236949921,
      "reward_std": 0.027104862965643406,
      "reward_std_all": 0.02983657270669937,
      "rewards-max/forensics_sinkhorn_reward": 0.9932679533958435,
      "rewards-max/forensics_strict_edge_reward": 0.6000864505767822,
      "rewards-max/forensics_tiou_reward": 0.9870781004428864,
      "rewards/forensics_sinkhorn_reward": 0.8197441101074219,
      "rewards/forensics_strict_edge_reward": 0.12069111876189709,
      "rewards/forensics_tiou_reward": 0.6504602432250977,
      "step": 1282
    },
    {
      "completion_length": 22.421875,
      "epoch": 3.9370199692780337,
      "grad_norm": 10.600467681884766,
      "kl": 0.376953125,
      "learning_rate": 1.3846153846153846e-08,
      "loss": -0.0597,
      "reward": 0.804301381111145,
      "reward_std": 0.005470501899253577,
      "reward_std_all": 0.006281132576987147,
      "rewards-max/forensics_sinkhorn_reward": 0.9896785616874695,
      "rewards-max/forensics_strict_edge_reward": 0.6862294971942902,
      "rewards-max/forensics_tiou_reward": 0.9779139161109924,
      "rewards/forensics_sinkhorn_reward": 0.8848296701908112,
      "rewards/forensics_strict_edge_reward": 0.09980275109410286,
      "rewards/forensics_tiou_reward": 0.7458344101905823,
      "step": 1283
    },
    {
      "completion_length": 26.203125,
      "epoch": 3.9400921658986174,
      "grad_norm": 7.123496055603027,
      "kl": 0.4521484375,
      "learning_rate": 1.3076923076923076e-08,
      "loss": -0.0065,
      "reward": 0.7979549169540405,
      "reward_std": 0.015170181286521256,
      "reward_std_all": 0.01504483015742153,
      "rewards-max/forensics_sinkhorn_reward": 0.9855254888534546,
      "rewards-max/forensics_strict_edge_reward": 0.5339852869510651,
      "rewards-max/forensics_tiou_reward": 0.9670527577400208,
      "rewards/forensics_sinkhorn_reward": 0.8344742655754089,
      "rewards/forensics_strict_edge_reward": 0.0671205185353756,
      "rewards/forensics_tiou_reward": 0.7069805860519409,
      "step": 1284
    },
    {
      "completion_length": 24.765625,
      "epoch": 3.943164362519201,
      "grad_norm": 8.545400619506836,
      "kl": 0.5029296875,
      "learning_rate": 1.2307692307692308e-08,
      "loss": -0.0923,
      "reward": 0.7851107120513916,
      "reward_std": 0.009953144239261746,
      "reward_std_all": 0.01802523760125041,
      "rewards-max/forensics_sinkhorn_reward": 0.978713870048523,
      "rewards-max/forensics_strict_edge_reward": 0.4538624435663223,
      "rewards-max/forensics_tiou_reward": 0.95079705119133,
      "rewards/forensics_sinkhorn_reward": 0.7751296162605286,
      "rewards/forensics_strict_edge_reward": 0.06128069385886192,
      "rewards/forensics_tiou_reward": 0.5767709016799927,
      "step": 1285
    },
    {
      "completion_length": 27.46875,
      "epoch": 3.946236559139785,
      "grad_norm": 8.003400802612305,
      "kl": 0.548828125,
      "learning_rate": 1.1538461538461538e-08,
      "loss": 0.0043,
      "reward": 0.7785293161869049,
      "reward_std": 0.030748853692784905,
      "reward_std_all": 0.031087303534150124,
      "rewards-max/forensics_sinkhorn_reward": 0.9886005818843842,
      "rewards-max/forensics_strict_edge_reward": 0.23991940915584564,
      "rewards-max/forensics_tiou_reward": 0.9692294895648956,
      "rewards/forensics_sinkhorn_reward": 0.7841314673423767,
      "rewards/forensics_strict_edge_reward": 0.022833604365587234,
      "rewards/forensics_tiou_reward": 0.6298829913139343,
      "step": 1286
    },
    {
      "completion_length": 24.421875,
      "epoch": 3.9493087557603688,
      "grad_norm": 9.82791805267334,
      "kl": 0.541015625,
      "learning_rate": 1.0769230769230768e-08,
      "loss": -0.0054,
      "reward": 0.7941255867481232,
      "reward_std": 0.017695305170491338,
      "reward_std_all": 0.017650828929618,
      "rewards-max/forensics_sinkhorn_reward": 0.9954811036586761,
      "rewards-max/forensics_strict_edge_reward": 0.9067722856998444,
      "rewards-max/forensics_tiou_reward": 0.9945850968360901,
      "rewards/forensics_sinkhorn_reward": 0.8351963758468628,
      "rewards/forensics_strict_edge_reward": 0.1180124580860138,
      "rewards/forensics_tiou_reward": 0.7002649903297424,
      "step": 1287
    },
    {
      "completion_length": 27.453125,
      "epoch": 3.9523809523809526,
      "grad_norm": 8.880599021911621,
      "kl": 0.4384765625,
      "learning_rate": 1e-08,
      "loss": 0.0259,
      "reward": 0.7621094882488251,
      "reward_std": 0.05779853696003556,
      "reward_std_all": 0.05408334161620587,
      "rewards-max/forensics_sinkhorn_reward": 0.9899260997772217,
      "rewards-max/forensics_strict_edge_reward": 0.6204420924186707,
      "rewards-max/forensics_tiou_reward": 0.9783972501754761,
      "rewards/forensics_sinkhorn_reward": 0.8180620074272156,
      "rewards/forensics_strict_edge_reward": 0.1157468892633915,
      "rewards/forensics_tiou_reward": 0.7049926817417145,
      "step": 1288
    },
    {
      "completion_length": 23.328125,
      "epoch": 3.955453149001536,
      "grad_norm": 9.027597427368164,
      "kl": 0.3876953125,
      "learning_rate": 9.23076923076923e-09,
      "loss": -0.0241,
      "reward": 0.7690526247024536,
      "reward_std": 0.050565268262289464,
      "reward_std_all": 0.04738411959260702,
      "rewards-max/forensics_sinkhorn_reward": 0.988436371088028,
      "rewards-max/forensics_strict_edge_reward": 0.6939628571271896,
      "rewards-max/forensics_tiou_reward": 0.9783149063587189,
      "rewards/forensics_sinkhorn_reward": 0.7833908498287201,
      "rewards/forensics_strict_edge_reward": 0.08850707113742828,
      "rewards/forensics_tiou_reward": 0.6363421082496643,
      "step": 1289
    },
    {
      "completion_length": 25.5625,
      "epoch": 3.9585253456221197,
      "grad_norm": 8.759471893310547,
      "kl": 0.478515625,
      "learning_rate": 8.46153846153846e-09,
      "loss": 0.0027,
      "reward": 0.7623693943023682,
      "reward_std": 0.04854067414999008,
      "reward_std_all": 0.04812945728190243,
      "rewards-max/forensics_sinkhorn_reward": 0.9843402206897736,
      "rewards-max/forensics_strict_edge_reward": 0.7075755298137665,
      "rewards-max/forensics_tiou_reward": 0.965985506772995,
      "rewards/forensics_sinkhorn_reward": 0.7870269417762756,
      "rewards/forensics_strict_edge_reward": 0.14357977360486984,
      "rewards/forensics_tiou_reward": 0.6401731073856354,
      "step": 1290
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.9615975422427034,
      "grad_norm": 9.659823417663574,
      "kl": 0.6484375,
      "learning_rate": 7.692307692307693e-09,
      "loss": -0.0456,
      "reward": 0.8051297962665558,
      "reward_std": 0.0035253060050308704,
      "reward_std_all": 0.004739053198136389,
      "rewards-max/forensics_sinkhorn_reward": 0.9961218237876892,
      "rewards-max/forensics_strict_edge_reward": 0.8278508186340332,
      "rewards-max/forensics_tiou_reward": 0.9964786171913147,
      "rewards/forensics_sinkhorn_reward": 0.8705360889434814,
      "rewards/forensics_strict_edge_reward": 0.10666880384087563,
      "rewards/forensics_tiou_reward": 0.712197333574295,
      "step": 1291
    },
    {
      "completion_length": 25.203125,
      "epoch": 3.964669738863287,
      "grad_norm": 8.347105979919434,
      "kl": 0.443359375,
      "learning_rate": 6.923076923076923e-09,
      "loss": 0.0063,
      "reward": 0.7940378189086914,
      "reward_std": 0.017042437568306923,
      "reward_std_all": 0.01719076931476593,
      "rewards-max/forensics_sinkhorn_reward": 0.9943536221981049,
      "rewards-max/forensics_strict_edge_reward": 0.4522130638360977,
      "rewards-max/forensics_tiou_reward": 0.9909932613372803,
      "rewards/forensics_sinkhorn_reward": 0.804374098777771,
      "rewards/forensics_strict_edge_reward": 0.05870785564184189,
      "rewards/forensics_tiou_reward": 0.6202638447284698,
      "step": 1292
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.967741935483871,
      "grad_norm": 7.512411117553711,
      "kl": 0.4697265625,
      "learning_rate": 6.153846153846154e-09,
      "loss": 0.1158,
      "reward": 0.7333736717700958,
      "reward_std": 0.0554416049271822,
      "reward_std_all": 0.06222926266491413,
      "rewards-max/forensics_sinkhorn_reward": 0.9948066174983978,
      "rewards-max/forensics_strict_edge_reward": 0.5212600529193878,
      "rewards-max/forensics_tiou_reward": 0.990376204252243,
      "rewards/forensics_sinkhorn_reward": 0.6790778040885925,
      "rewards/forensics_strict_edge_reward": 0.05644797161221504,
      "rewards/forensics_tiou_reward": 0.5029377937316895,
      "step": 1293
    },
    {
      "completion_length": 24.25,
      "epoch": 3.9708141321044548,
      "grad_norm": 9.026246070861816,
      "kl": 0.5400390625,
      "learning_rate": 5.384615384615384e-09,
      "loss": 0.0015,
      "reward": 0.7380117177963257,
      "reward_std": 0.028606954496353865,
      "reward_std_all": 0.04932361375540495,
      "rewards-max/forensics_sinkhorn_reward": 0.9895489811897278,
      "rewards-max/forensics_strict_edge_reward": 0.6884901523590088,
      "rewards-max/forensics_tiou_reward": 0.9779657423496246,
      "rewards/forensics_sinkhorn_reward": 0.7025197148323059,
      "rewards/forensics_strict_edge_reward": 0.060059040784835815,
      "rewards/forensics_tiou_reward": 0.5182258784770966,
      "step": 1294
    },
    {
      "completion_length": 28.265625,
      "epoch": 3.9738863287250386,
      "grad_norm": 6.6942901611328125,
      "kl": 0.5390625,
      "learning_rate": 4.615384615384615e-09,
      "loss": -0.0266,
      "reward": 0.7830274999141693,
      "reward_std": 0.027914217673242092,
      "reward_std_all": 0.02776301745325327,
      "rewards-max/forensics_sinkhorn_reward": 0.9889737665653229,
      "rewards-max/forensics_strict_edge_reward": 0.7104560136795044,
      "rewards-max/forensics_tiou_reward": 0.9721311926841736,
      "rewards/forensics_sinkhorn_reward": 0.8226942718029022,
      "rewards/forensics_strict_edge_reward": 0.09955653175711632,
      "rewards/forensics_tiou_reward": 0.6584025919437408,
      "step": 1295
    },
    {
      "completion_length": 26.859375,
      "epoch": 3.976958525345622,
      "grad_norm": 6.942562103271484,
      "kl": 0.66796875,
      "learning_rate": 3.846153846153846e-09,
      "loss": -0.0633,
      "reward": 0.8062232732772827,
      "reward_std": 0.0014310484984889627,
      "reward_std_all": 0.0030451257480308414,
      "rewards-max/forensics_sinkhorn_reward": 0.9864111542701721,
      "rewards-max/forensics_strict_edge_reward": 0.5120944976806641,
      "rewards-max/forensics_tiou_reward": 0.969757616519928,
      "rewards/forensics_sinkhorn_reward": 0.8791904747486115,
      "rewards/forensics_strict_edge_reward": 0.08677780255675316,
      "rewards/forensics_tiou_reward": 0.7374826371669769,
      "step": 1296
    },
    {
      "completion_length": 29.09375,
      "epoch": 3.980030721966206,
      "grad_norm": 8.16668701171875,
      "kl": 0.46875,
      "learning_rate": 3.076923076923077e-09,
      "loss": -0.039,
      "reward": 0.7685027420520782,
      "reward_std": 0.04579873941838741,
      "reward_std_all": 0.04402184020727873,
      "rewards-max/forensics_sinkhorn_reward": 0.994010329246521,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9892993867397308,
      "rewards/forensics_sinkhorn_reward": 0.7413274049758911,
      "rewards/forensics_strict_edge_reward": 0.11941718682646751,
      "rewards/forensics_tiou_reward": 0.5896673202514648,
      "step": 1297
    },
    {
      "completion_length": 23.40625,
      "epoch": 3.9831029185867894,
      "grad_norm": 8.582514762878418,
      "kl": 0.443359375,
      "learning_rate": 2.3076923076923076e-09,
      "loss": 0.0004,
      "reward": 0.7807306349277496,
      "reward_std": 0.03145267005311325,
      "reward_std_all": 0.030628938460722566,
      "rewards-max/forensics_sinkhorn_reward": 0.9928257465362549,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9851261377334595,
      "rewards/forensics_sinkhorn_reward": 0.8329184055328369,
      "rewards/forensics_strict_edge_reward": 0.12457820773124695,
      "rewards/forensics_tiou_reward": 0.6701093912124634,
      "step": 1298
    },
    {
      "completion_length": 25.4375,
      "epoch": 3.986175115207373,
      "grad_norm": 8.088836669921875,
      "kl": 0.404296875,
      "learning_rate": 1.5384615384615385e-09,
      "loss": -0.1042,
      "reward": 0.7981788814067841,
      "reward_std": 0.013727718440350145,
      "reward_std_all": 0.014228565152734518,
      "rewards-max/forensics_sinkhorn_reward": 0.9854698777198792,
      "rewards-max/forensics_strict_edge_reward": 0.6215263903141022,
      "rewards-max/forensics_tiou_reward": 0.9701238572597504,
      "rewards/forensics_sinkhorn_reward": 0.8373850882053375,
      "rewards/forensics_strict_edge_reward": 0.05186615325510502,
      "rewards/forensics_tiou_reward": 0.6814626753330231,
      "step": 1299
    },
    {
      "completion_length": 26.203125,
      "epoch": 3.989247311827957,
      "grad_norm": 9.95792293548584,
      "kl": 0.4345703125,
      "learning_rate": 7.692307692307692e-10,
      "loss": -0.0633,
      "reward": 0.8033623695373535,
      "reward_std": 0.005200170679017901,
      "reward_std_all": 0.0065794335678219795,
      "rewards-max/forensics_sinkhorn_reward": 0.990763247013092,
      "rewards-max/forensics_strict_edge_reward": 0.5461844205856323,
      "rewards-max/forensics_tiou_reward": 0.977895051240921,
      "rewards/forensics_sinkhorn_reward": 0.8362525105476379,
      "rewards/forensics_strict_edge_reward": 0.09275419637560844,
      "rewards/forensics_tiou_reward": 0.6721398234367371,
      "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
}