{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 956,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 23.046875,
      "epoch": 0.0041841004184100415,
      "grad_norm": 15.792669296264648,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": 0.1774,
      "reward": 0.4692053645849228,
      "reward_std": 0.2604151964187622,
      "reward_std_all": 0.2927922159433365,
      "rewards-max/forensics_sinkhorn_reward": 0.8494397103786469,
      "rewards-max/forensics_strict_edge_reward": 0.03246131680498365,
      "rewards-max/forensics_tiou_reward": 0.6463126540184021,
      "rewards/forensics_sinkhorn_reward": 0.3575640618801117,
      "rewards/forensics_strict_edge_reward": 0.001620107808321336,
      "rewards/forensics_tiou_reward": 0.14521798491477966,
      "step": 1
    },
    {
      "completion_length": 23.90625,
      "epoch": 0.008368200836820083,
      "grad_norm": 6.257342338562012,
      "kl": 0.00040340423583984375,
      "learning_rate": 9.989539748953975e-07,
      "loss": 0.0964,
      "reward": 0.542992427945137,
      "reward_std": 0.21707497537136078,
      "reward_std_all": 0.23201030492782593,
      "rewards-max/forensics_sinkhorn_reward": 0.9493957757949829,
      "rewards-max/forensics_strict_edge_reward": 0.006090680719353259,
      "rewards-max/forensics_tiou_reward": 0.8869291543960571,
      "rewards/forensics_sinkhorn_reward": 0.4457518011331558,
      "rewards/forensics_strict_edge_reward": 0.0004389965397422202,
      "rewards/forensics_tiou_reward": 0.21324709802865982,
      "step": 2
    },
    {
      "completion_length": 23.28125,
      "epoch": 0.012552301255230125,
      "grad_norm": 6.531919956207275,
      "kl": 0.0020599365234375,
      "learning_rate": 9.97907949790795e-07,
      "loss": 0.0918,
      "reward": 0.5241373181343079,
      "reward_std": 0.24059882760047913,
      "reward_std_all": 0.24917282164096832,
      "rewards-max/forensics_sinkhorn_reward": 0.8129183948040009,
      "rewards-max/forensics_strict_edge_reward": 6.814503603891353e-05,
      "rewards-max/forensics_tiou_reward": 0.5758000910282135,
      "rewards/forensics_sinkhorn_reward": 0.4059979021549225,
      "rewards/forensics_strict_edge_reward": 3.204721366500962e-06,
      "rewards/forensics_tiou_reward": 0.12423179298639297,
      "step": 3
    },
    {
      "completion_length": 23.234375,
      "epoch": 0.016736401673640166,
      "grad_norm": 6.28652811050415,
      "kl": 0.003124237060546875,
      "learning_rate": 9.968619246861925e-07,
      "loss": 0.0928,
      "reward": 0.5453928709030151,
      "reward_std": 0.25078868865966797,
      "reward_std_all": 0.24535268545150757,
      "rewards-max/forensics_sinkhorn_reward": 0.9463425576686859,
      "rewards-max/forensics_strict_edge_reward": 0.04894637409597635,
      "rewards-max/forensics_tiou_reward": 0.8847882449626923,
      "rewards/forensics_sinkhorn_reward": 0.43622174859046936,
      "rewards/forensics_strict_edge_reward": 0.002043078769929707,
      "rewards/forensics_tiou_reward": 0.18237587809562683,
      "step": 4
    },
    {
      "completion_length": 21.765625,
      "epoch": 0.02092050209205021,
      "grad_norm": 7.644857883453369,
      "kl": 0.013153076171875,
      "learning_rate": 9.9581589958159e-07,
      "loss": 0.2475,
      "reward": 0.566672682762146,
      "reward_std": 0.1793363019824028,
      "reward_std_all": 0.20209543406963348,
      "rewards-max/forensics_sinkhorn_reward": 0.8430197536945343,
      "rewards-max/forensics_strict_edge_reward": 0.0017297947249375056,
      "rewards-max/forensics_tiou_reward": 0.6192428320646286,
      "rewards/forensics_sinkhorn_reward": 0.45368359982967377,
      "rewards/forensics_strict_edge_reward": 0.00014969322921309924,
      "rewards/forensics_tiou_reward": 0.163836307823658,
      "step": 5
    },
    {
      "completion_length": 22.21875,
      "epoch": 0.02510460251046025,
      "grad_norm": 5.499014854431152,
      "kl": 0.01837158203125,
      "learning_rate": 9.947698744769873e-07,
      "loss": 0.2236,
      "reward": 0.5724351406097412,
      "reward_std": 0.16364839673042297,
      "reward_std_all": 0.18643251061439514,
      "rewards-max/forensics_sinkhorn_reward": 0.97243931889534,
      "rewards-max/forensics_strict_edge_reward": 0.3585696288209874,
      "rewards-max/forensics_tiou_reward": 0.9534751176834106,
      "rewards/forensics_sinkhorn_reward": 0.4563644826412201,
      "rewards/forensics_strict_edge_reward": 0.015041870020468195,
      "rewards/forensics_tiou_reward": 0.15701786801218987,
      "step": 6
    },
    {
      "completion_length": 22.15625,
      "epoch": 0.029288702928870293,
      "grad_norm": 3.8824219703674316,
      "kl": 0.025634765625,
      "learning_rate": 9.937238493723848e-07,
      "loss": 0.1831,
      "reward": 0.5864841938018799,
      "reward_std": 0.1357056014239788,
      "reward_std_all": 0.16907904297113419,
      "rewards-max/forensics_sinkhorn_reward": 0.921358734369278,
      "rewards-max/forensics_strict_edge_reward": 0.08550463803112507,
      "rewards-max/forensics_tiou_reward": 0.8395148515701294,
      "rewards/forensics_sinkhorn_reward": 0.4731210172176361,
      "rewards/forensics_strict_edge_reward": 0.0035917062778025866,
      "rewards/forensics_tiou_reward": 0.19966398179531097,
      "step": 7
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.03347280334728033,
      "grad_norm": 5.519156455993652,
      "kl": 0.0340576171875,
      "learning_rate": 9.926778242677823e-07,
      "loss": 0.1571,
      "reward": 0.5725448727607727,
      "reward_std": 0.18244238942861557,
      "reward_std_all": 0.20083454251289368,
      "rewards-max/forensics_sinkhorn_reward": 0.9023396372795105,
      "rewards-max/forensics_strict_edge_reward": 0.20876271557062864,
      "rewards-max/forensics_tiou_reward": 0.7929216027259827,
      "rewards/forensics_sinkhorn_reward": 0.46315230429172516,
      "rewards/forensics_strict_edge_reward": 0.009152861370239407,
      "rewards/forensics_tiou_reward": 0.21216679364442825,
      "step": 8
    },
    {
      "completion_length": 22.78125,
      "epoch": 0.03765690376569038,
      "grad_norm": 4.270610809326172,
      "kl": 0.0377197265625,
      "learning_rate": 9.916317991631798e-07,
      "loss": 0.1135,
      "reward": 0.594406247138977,
      "reward_std": 0.15163327008485794,
      "reward_std_all": 0.17794296145439148,
      "rewards-max/forensics_sinkhorn_reward": 0.9640792310237885,
      "rewards-max/forensics_strict_edge_reward": 0.41030948533443734,
      "rewards-max/forensics_tiou_reward": 0.9143810272216797,
      "rewards/forensics_sinkhorn_reward": 0.4936779588460922,
      "rewards/forensics_strict_edge_reward": 0.02706366499478463,
      "rewards/forensics_tiou_reward": 0.22284918278455734,
      "step": 9
    },
    {
      "completion_length": 23.5625,
      "epoch": 0.04184100418410042,
      "grad_norm": 5.209514617919922,
      "kl": 0.119873046875,
      "learning_rate": 9.905857740585774e-07,
      "loss": 0.1626,
      "reward": 0.5413075685501099,
      "reward_std": 0.14990021288394928,
      "reward_std_all": 0.19169225543737411,
      "rewards-max/forensics_sinkhorn_reward": 0.8744174540042877,
      "rewards-max/forensics_strict_edge_reward": 3.6458853365672894e-05,
      "rewards-max/forensics_tiou_reward": 0.7511946856975555,
      "rewards/forensics_sinkhorn_reward": 0.42438124120235443,
      "rewards/forensics_strict_edge_reward": 1.5195449866722832e-06,
      "rewards/forensics_tiou_reward": 0.1593071147799492,
      "step": 10
    },
    {
      "completion_length": 24.140625,
      "epoch": 0.04602510460251046,
      "grad_norm": 5.461408615112305,
      "kl": 0.14111328125,
      "learning_rate": 9.895397489539749e-07,
      "loss": 0.1031,
      "reward": 0.5852597653865814,
      "reward_std": 0.1534106582403183,
      "reward_std_all": 0.18192614614963531,
      "rewards-max/forensics_sinkhorn_reward": 0.9364532828330994,
      "rewards-max/forensics_strict_edge_reward": 0.22369655221700668,
      "rewards-max/forensics_tiou_reward": 0.8529622852802277,
      "rewards/forensics_sinkhorn_reward": 0.46202149987220764,
      "rewards/forensics_strict_edge_reward": 0.009890091139823198,
      "rewards/forensics_tiou_reward": 0.1961829960346222,
      "step": 11
    },
    {
      "completion_length": 21.625,
      "epoch": 0.0502092050209205,
      "grad_norm": 4.911070346832275,
      "kl": 0.1904296875,
      "learning_rate": 9.884937238493724e-07,
      "loss": 0.2508,
      "reward": 0.5986931324005127,
      "reward_std": 0.1699618324637413,
      "reward_std_all": 0.18390138447284698,
      "rewards-max/forensics_sinkhorn_reward": 0.8746095895767212,
      "rewards-max/forensics_strict_edge_reward": 0.032727541169151664,
      "rewards-max/forensics_tiou_reward": 0.7061965465545654,
      "rewards/forensics_sinkhorn_reward": 0.49398644268512726,
      "rewards/forensics_strict_edge_reward": 0.0016088418442450347,
      "rewards/forensics_tiou_reward": 0.2020401805639267,
      "step": 12
    },
    {
      "completion_length": 22.1875,
      "epoch": 0.05439330543933055,
      "grad_norm": 4.50308895111084,
      "kl": 0.17041015625,
      "learning_rate": 9.8744769874477e-07,
      "loss": 0.1313,
      "reward": 0.6666364669799805,
      "reward_std": 0.1597231775522232,
      "reward_std_all": 0.15035968273878098,
      "rewards-max/forensics_sinkhorn_reward": 0.9423770010471344,
      "rewards-max/forensics_strict_edge_reward": 0.2174275739089353,
      "rewards-max/forensics_tiou_reward": 0.8808007836341858,
      "rewards/forensics_sinkhorn_reward": 0.5535479187965393,
      "rewards/forensics_strict_edge_reward": 0.010554582173426752,
      "rewards/forensics_tiou_reward": 0.2912745773792267,
      "step": 13
    },
    {
      "completion_length": 22.828125,
      "epoch": 0.058577405857740586,
      "grad_norm": 3.991748094558716,
      "kl": 0.15576171875,
      "learning_rate": 9.864016736401674e-07,
      "loss": 0.2166,
      "reward": 0.608587384223938,
      "reward_std": 0.17733772099018097,
      "reward_std_all": 0.18607552349567413,
      "rewards-max/forensics_sinkhorn_reward": 0.8880496323108673,
      "rewards-max/forensics_strict_edge_reward": 0.02328993820742653,
      "rewards-max/forensics_tiou_reward": 0.7118096649646759,
      "rewards/forensics_sinkhorn_reward": 0.5010628402233124,
      "rewards/forensics_strict_edge_reward": 0.0009868546688664992,
      "rewards/forensics_tiou_reward": 0.21443387866020203,
      "step": 14
    },
    {
      "completion_length": 20.6875,
      "epoch": 0.06276150627615062,
      "grad_norm": 5.442530632019043,
      "kl": 0.17626953125,
      "learning_rate": 9.85355648535565e-07,
      "loss": 0.0942,
      "reward": 0.6796411871910095,
      "reward_std": 0.0992271788418293,
      "reward_std_all": 0.11457009613513947,
      "rewards-max/forensics_sinkhorn_reward": 0.8667139410972595,
      "rewards-max/forensics_strict_edge_reward": 0.12753352348227054,
      "rewards-max/forensics_tiou_reward": 0.7051013559103012,
      "rewards/forensics_sinkhorn_reward": 0.5747380256652832,
      "rewards/forensics_strict_edge_reward": 0.005356984361242212,
      "rewards/forensics_tiou_reward": 0.21917849779129028,
      "step": 15
    },
    {
      "completion_length": 22.484375,
      "epoch": 0.06694560669456066,
      "grad_norm": 4.611914157867432,
      "kl": 0.197265625,
      "learning_rate": 9.843096234309622e-07,
      "loss": 0.1628,
      "reward": 0.6422728300094604,
      "reward_std": 0.14277870953083038,
      "reward_std_all": 0.1486809030175209,
      "rewards-max/forensics_sinkhorn_reward": 0.9198012351989746,
      "rewards-max/forensics_strict_edge_reward": 0.0047596043441444635,
      "rewards-max/forensics_tiou_reward": 0.8048739433288574,
      "rewards/forensics_sinkhorn_reward": 0.5368130803108215,
      "rewards/forensics_strict_edge_reward": 0.00020101920381421223,
      "rewards/forensics_tiou_reward": 0.23528224974870682,
      "step": 16
    },
    {
      "completion_length": 19.390625,
      "epoch": 0.07112970711297072,
      "grad_norm": 5.545915603637695,
      "kl": 0.21728515625,
      "learning_rate": 9.832635983263597e-07,
      "loss": 0.2514,
      "reward": 0.6430337429046631,
      "reward_std": 0.139505997300148,
      "reward_std_all": 0.15164706110954285,
      "rewards-max/forensics_sinkhorn_reward": 0.9876781404018402,
      "rewards-max/forensics_strict_edge_reward": 0.6168098747730255,
      "rewards-max/forensics_tiou_reward": 0.9961282014846802,
      "rewards/forensics_sinkhorn_reward": 0.5723722279071808,
      "rewards/forensics_strict_edge_reward": 0.04200271517038345,
      "rewards/forensics_tiou_reward": 0.2813178822398186,
      "step": 17
    },
    {
      "completion_length": 20.6875,
      "epoch": 0.07531380753138076,
      "grad_norm": 4.211256504058838,
      "kl": 0.2255859375,
      "learning_rate": 9.822175732217572e-07,
      "loss": 0.1063,
      "reward": 0.6852478981018066,
      "reward_std": 0.09323486685752869,
      "reward_std_all": 0.10120473057031631,
      "rewards-max/forensics_sinkhorn_reward": 0.9203271865844727,
      "rewards-max/forensics_strict_edge_reward": 0.0013165008858777583,
      "rewards-max/forensics_tiou_reward": 0.8116961121559143,
      "rewards/forensics_sinkhorn_reward": 0.5860735476016998,
      "rewards/forensics_strict_edge_reward": 5.5413151130778715e-05,
      "rewards/forensics_tiou_reward": 0.22235696762800217,
      "step": 18
    },
    {
      "completion_length": 21.28125,
      "epoch": 0.0794979079497908,
      "grad_norm": 5.285574913024902,
      "kl": 0.23291015625,
      "learning_rate": 9.811715481171548e-07,
      "loss": 0.249,
      "reward": 0.6028056442737579,
      "reward_std": 0.1305978149175644,
      "reward_std_all": 0.1574201062321663,
      "rewards-max/forensics_sinkhorn_reward": 0.8988212943077087,
      "rewards-max/forensics_strict_edge_reward": 0.019176081754267216,
      "rewards-max/forensics_tiou_reward": 0.7601620256900787,
      "rewards/forensics_sinkhorn_reward": 0.49780499935150146,
      "rewards/forensics_strict_edge_reward": 0.0008064719731919467,
      "rewards/forensics_tiou_reward": 0.16221964359283447,
      "step": 19
    },
    {
      "completion_length": 21.3125,
      "epoch": 0.08368200836820083,
      "grad_norm": 4.2882161140441895,
      "kl": 0.2294921875,
      "learning_rate": 9.801255230125523e-07,
      "loss": 0.0655,
      "reward": 0.6595905125141144,
      "reward_std": 0.124163419008255,
      "reward_std_all": 0.12975098192691803,
      "rewards-max/forensics_sinkhorn_reward": 0.9493655562400818,
      "rewards-max/forensics_strict_edge_reward": 0.0456080031581223,
      "rewards-max/forensics_tiou_reward": 0.8927965760231018,
      "rewards/forensics_sinkhorn_reward": 0.549843817949295,
      "rewards/forensics_strict_edge_reward": 0.002080210193526,
      "rewards/forensics_tiou_reward": 0.23696640133857727,
      "step": 20
    },
    {
      "completion_length": 20.6875,
      "epoch": 0.08786610878661087,
      "grad_norm": 4.027358055114746,
      "kl": 0.23388671875,
      "learning_rate": 9.790794979079498e-07,
      "loss": 0.1292,
      "reward": 0.6545178890228271,
      "reward_std": 0.10709446668624878,
      "reward_std_all": 0.12602145224809647,
      "rewards-max/forensics_sinkhorn_reward": 0.913782924413681,
      "rewards-max/forensics_strict_edge_reward": 0.01801540981978178,
      "rewards-max/forensics_tiou_reward": 0.8112528622150421,
      "rewards/forensics_sinkhorn_reward": 0.5463005602359772,
      "rewards/forensics_strict_edge_reward": 0.0009181998902931809,
      "rewards/forensics_tiou_reward": 0.2207762375473976,
      "step": 21
    },
    {
      "completion_length": 20.609375,
      "epoch": 0.09205020920502092,
      "grad_norm": 5.084755897521973,
      "kl": 0.244140625,
      "learning_rate": 9.780334728033473e-07,
      "loss": 0.1358,
      "reward": 0.6804028749465942,
      "reward_std": 0.11533960700035095,
      "reward_std_all": 0.11645488440990448,
      "rewards-max/forensics_sinkhorn_reward": 0.9320398271083832,
      "rewards-max/forensics_strict_edge_reward": 0.04898987588239834,
      "rewards-max/forensics_tiou_reward": 0.8302538990974426,
      "rewards/forensics_sinkhorn_reward": 0.5959420204162598,
      "rewards/forensics_strict_edge_reward": 0.0027312603924656287,
      "rewards/forensics_tiou_reward": 0.25685106962919235,
      "step": 22
    },
    {
      "completion_length": 21.140625,
      "epoch": 0.09623430962343096,
      "grad_norm": 4.563536643981934,
      "kl": 0.2197265625,
      "learning_rate": 9.769874476987448e-07,
      "loss": 0.1647,
      "reward": 0.6874531507492065,
      "reward_std": 0.10845623165369034,
      "reward_std_all": 0.10965507850050926,
      "rewards-max/forensics_sinkhorn_reward": 0.8742459416389465,
      "rewards-max/forensics_strict_edge_reward": 0.019455070316325873,
      "rewards-max/forensics_tiou_reward": 0.7169546484947205,
      "rewards/forensics_sinkhorn_reward": 0.5737697184085846,
      "rewards/forensics_strict_edge_reward": 0.0016472498409711989,
      "rewards/forensics_tiou_reward": 0.20477645099163055,
      "step": 23
    },
    {
      "completion_length": 20.1875,
      "epoch": 0.100418410041841,
      "grad_norm": 4.832355976104736,
      "kl": 0.4560546875,
      "learning_rate": 9.759414225941423e-07,
      "loss": 0.1235,
      "reward": 0.7060138881206512,
      "reward_std": 0.10626495629549026,
      "reward_std_all": 0.10276904702186584,
      "rewards-max/forensics_sinkhorn_reward": 0.9225222766399384,
      "rewards-max/forensics_strict_edge_reward": 0.1517574079334736,
      "rewards-max/forensics_tiou_reward": 0.8442344069480896,
      "rewards/forensics_sinkhorn_reward": 0.6420442461967468,
      "rewards/forensics_strict_edge_reward": 0.0065268296748399734,
      "rewards/forensics_tiou_reward": 0.3382207006216049,
      "step": 24
    },
    {
      "completion_length": 20.96875,
      "epoch": 0.10460251046025104,
      "grad_norm": 4.834485054016113,
      "kl": 0.19775390625,
      "learning_rate": 9.748953974895398e-07,
      "loss": 0.0231,
      "reward": 0.6747651696205139,
      "reward_std": 0.10663226246833801,
      "reward_std_all": 0.1130276545882225,
      "rewards-max/forensics_sinkhorn_reward": 0.8894841372966766,
      "rewards-max/forensics_strict_edge_reward": 0.0021568600823229644,
      "rewards-max/forensics_tiou_reward": 0.7027697861194611,
      "rewards/forensics_sinkhorn_reward": 0.5754060745239258,
      "rewards/forensics_strict_edge_reward": 9.391848516315804e-05,
      "rewards/forensics_tiou_reward": 0.22803795337677002,
      "step": 25
    },
    {
      "completion_length": 20.609375,
      "epoch": 0.1087866108786611,
      "grad_norm": 4.45652437210083,
      "kl": 0.197265625,
      "learning_rate": 9.738493723849371e-07,
      "loss": 0.0596,
      "reward": 0.707344651222229,
      "reward_std": 0.06784206070005894,
      "reward_std_all": 0.08278313279151917,
      "rewards-max/forensics_sinkhorn_reward": 0.923992782831192,
      "rewards-max/forensics_strict_edge_reward": 0.017017672769725323,
      "rewards-max/forensics_tiou_reward": 0.8270621001720428,
      "rewards/forensics_sinkhorn_reward": 0.6109670400619507,
      "rewards/forensics_strict_edge_reward": 0.0016334940737579018,
      "rewards/forensics_tiou_reward": 0.2781525254249573,
      "step": 26
    },
    {
      "completion_length": 20.65625,
      "epoch": 0.11297071129707113,
      "grad_norm": 4.287718772888184,
      "kl": 0.197265625,
      "learning_rate": 9.728033472803347e-07,
      "loss": 0.013,
      "reward": 0.6966844499111176,
      "reward_std": 0.09510937333106995,
      "reward_std_all": 0.09932287037372589,
      "rewards-max/forensics_sinkhorn_reward": 0.8986561596393585,
      "rewards-max/forensics_strict_edge_reward": 0.016183795294637093,
      "rewards-max/forensics_tiou_reward": 0.7514901459217072,
      "rewards/forensics_sinkhorn_reward": 0.585329681634903,
      "rewards/forensics_strict_edge_reward": 0.0015831197277975662,
      "rewards/forensics_tiou_reward": 0.25879475474357605,
      "step": 27
    },
    {
      "completion_length": 20.625,
      "epoch": 0.11715481171548117,
      "grad_norm": 3.9522714614868164,
      "kl": 0.23974609375,
      "learning_rate": 9.717573221757322e-07,
      "loss": 0.1672,
      "reward": 0.7140634059906006,
      "reward_std": 0.08362230658531189,
      "reward_std_all": 0.08887658268213272,
      "rewards-max/forensics_sinkhorn_reward": 0.9709955751895905,
      "rewards-max/forensics_strict_edge_reward": 0.3124852478504181,
      "rewards-max/forensics_tiou_reward": 0.9439889788627625,
      "rewards/forensics_sinkhorn_reward": 0.6353549957275391,
      "rewards/forensics_strict_edge_reward": 0.015753655694425106,
      "rewards/forensics_tiou_reward": 0.30534714460372925,
      "step": 28
    },
    {
      "completion_length": 21.0,
      "epoch": 0.12133891213389121,
      "grad_norm": 4.833073139190674,
      "kl": 0.29052734375,
      "learning_rate": 9.707112970711297e-07,
      "loss": 0.1963,
      "reward": 0.6778136491775513,
      "reward_std": 0.09749481081962585,
      "reward_std_all": 0.10771588981151581,
      "rewards-max/forensics_sinkhorn_reward": 0.8765552043914795,
      "rewards-max/forensics_strict_edge_reward": 0.1514560483628884,
      "rewards-max/forensics_tiou_reward": 0.7429804801940918,
      "rewards/forensics_sinkhorn_reward": 0.5562231540679932,
      "rewards/forensics_strict_edge_reward": 0.006367751062498428,
      "rewards/forensics_tiou_reward": 0.22961296886205673,
      "step": 29
    },
    {
      "completion_length": 21.453125,
      "epoch": 0.12552301255230125,
      "grad_norm": 4.572509765625,
      "kl": 0.20751953125,
      "learning_rate": 9.696652719665272e-07,
      "loss": 0.2116,
      "reward": 0.6483758091926575,
      "reward_std": 0.12735845521092415,
      "reward_std_all": 0.1383885145187378,
      "rewards-max/forensics_sinkhorn_reward": 0.9369700253009796,
      "rewards-max/forensics_strict_edge_reward": 0.3969262231839821,
      "rewards-max/forensics_tiou_reward": 0.8571477830410004,
      "rewards/forensics_sinkhorn_reward": 0.5568287372589111,
      "rewards/forensics_strict_edge_reward": 0.01690467142907437,
      "rewards/forensics_tiou_reward": 0.2470608577132225,
      "step": 30
    },
    {
      "completion_length": 20.0625,
      "epoch": 0.1297071129707113,
      "grad_norm": 4.509507179260254,
      "kl": 0.21533203125,
      "learning_rate": 9.686192468619247e-07,
      "loss": 0.3162,
      "reward": 0.6628539860248566,
      "reward_std": 0.09784890711307526,
      "reward_std_all": 0.11828882619738579,
      "rewards-max/forensics_sinkhorn_reward": 0.9736836552619934,
      "rewards-max/forensics_strict_edge_reward": 0.2555675841867924,
      "rewards-max/forensics_tiou_reward": 0.958007425069809,
      "rewards/forensics_sinkhorn_reward": 0.5813169479370117,
      "rewards/forensics_strict_edge_reward": 0.011933977948501706,
      "rewards/forensics_tiou_reward": 0.25458874553442,
      "step": 31
    },
    {
      "completion_length": 22.734375,
      "epoch": 0.13389121338912133,
      "grad_norm": 4.275415420532227,
      "kl": 0.19921875,
      "learning_rate": 9.675732217573222e-07,
      "loss": 0.1735,
      "reward": 0.6333380937576294,
      "reward_std": 0.14213446900248528,
      "reward_std_all": 0.14837748184800148,
      "rewards-max/forensics_sinkhorn_reward": 0.8813243508338928,
      "rewards-max/forensics_strict_edge_reward": 0.003889945743139833,
      "rewards-max/forensics_tiou_reward": 0.699639618396759,
      "rewards/forensics_sinkhorn_reward": 0.5360657423734665,
      "rewards/forensics_strict_edge_reward": 0.00018216629905509762,
      "rewards/forensics_tiou_reward": 0.24160334467887878,
      "step": 32
    },
    {
      "completion_length": 20.15625,
      "epoch": 0.13807531380753138,
      "grad_norm": 3.678189277648926,
      "kl": 0.1982421875,
      "learning_rate": 9.665271966527197e-07,
      "loss": 0.0124,
      "reward": 0.6620001494884491,
      "reward_std": 0.12426695227622986,
      "reward_std_all": 0.127024844288826,
      "rewards-max/forensics_sinkhorn_reward": 0.9076316058635712,
      "rewards-max/forensics_strict_edge_reward": 0.09839049796573818,
      "rewards-max/forensics_tiou_reward": 0.7934170663356781,
      "rewards/forensics_sinkhorn_reward": 0.554381251335144,
      "rewards/forensics_strict_edge_reward": 0.00664074523592717,
      "rewards/forensics_tiou_reward": 0.2274969071149826,
      "step": 33
    },
    {
      "completion_length": 21.15625,
      "epoch": 0.14225941422594143,
      "grad_norm": 5.326334476470947,
      "kl": 0.23974609375,
      "learning_rate": 9.654811715481172e-07,
      "loss": 0.1788,
      "reward": 0.6003984808921814,
      "reward_std": 0.147904172539711,
      "reward_std_all": 0.17151565849781036,
      "rewards-max/forensics_sinkhorn_reward": 0.9347912073135376,
      "rewards-max/forensics_strict_edge_reward": 0.003954987390898168,
      "rewards-max/forensics_tiou_reward": 0.8333602845668793,
      "rewards/forensics_sinkhorn_reward": 0.5146069377660751,
      "rewards/forensics_strict_edge_reward": 0.0001985383642022498,
      "rewards/forensics_tiou_reward": 0.2276962250471115,
      "step": 34
    },
    {
      "completion_length": 20.0,
      "epoch": 0.14644351464435146,
      "grad_norm": 4.59984016418457,
      "kl": 0.21142578125,
      "learning_rate": 9.644351464435145e-07,
      "loss": 0.1828,
      "reward": 0.6252763271331787,
      "reward_std": 0.11433953791856766,
      "reward_std_all": 0.14519579708576202,
      "rewards-max/forensics_sinkhorn_reward": 0.9405063986778259,
      "rewards-max/forensics_strict_edge_reward": 0.12066118977963924,
      "rewards-max/forensics_tiou_reward": 0.8527171909809113,
      "rewards/forensics_sinkhorn_reward": 0.5375770628452301,
      "rewards/forensics_strict_edge_reward": 0.005230255424976349,
      "rewards/forensics_tiou_reward": 0.23283924907445908,
      "step": 35
    },
    {
      "completion_length": 19.984375,
      "epoch": 0.1506276150627615,
      "grad_norm": 4.343069076538086,
      "kl": 0.22021484375,
      "learning_rate": 9.63389121338912e-07,
      "loss": 0.1416,
      "reward": 0.6934757232666016,
      "reward_std": 0.11853928118944168,
      "reward_std_all": 0.11607768386602402,
      "rewards-max/forensics_sinkhorn_reward": 0.9027147591114044,
      "rewards-max/forensics_strict_edge_reward": 0.10536623187363148,
      "rewards-max/forensics_tiou_reward": 0.7759073078632355,
      "rewards/forensics_sinkhorn_reward": 0.6066514551639557,
      "rewards/forensics_strict_edge_reward": 0.004576865932904184,
      "rewards/forensics_tiou_reward": 0.2800520360469818,
      "step": 36
    },
    {
      "completion_length": 20.671875,
      "epoch": 0.15481171548117154,
      "grad_norm": 4.359643936157227,
      "kl": 0.232421875,
      "learning_rate": 9.623430962343096e-07,
      "loss": 0.1736,
      "reward": 0.6820591688156128,
      "reward_std": 0.14937610179185867,
      "reward_std_all": 0.1392119824886322,
      "rewards-max/forensics_sinkhorn_reward": 0.9626912176609039,
      "rewards-max/forensics_strict_edge_reward": 0.23576344549655914,
      "rewards-max/forensics_tiou_reward": 0.9220856130123138,
      "rewards/forensics_sinkhorn_reward": 0.6067031621932983,
      "rewards/forensics_strict_edge_reward": 0.015588616952300072,
      "rewards/forensics_tiou_reward": 0.3051804453134537,
      "step": 37
    },
    {
      "completion_length": 19.671875,
      "epoch": 0.1589958158995816,
      "grad_norm": 5.454073905944824,
      "kl": 0.2431640625,
      "learning_rate": 9.61297071129707e-07,
      "loss": 0.2211,
      "reward": 0.6326093077659607,
      "reward_std": 0.1883159950375557,
      "reward_std_all": 0.18240567296743393,
      "rewards-max/forensics_sinkhorn_reward": 0.9520781934261322,
      "rewards-max/forensics_strict_edge_reward": 0.024154260754585266,
      "rewards-max/forensics_tiou_reward": 0.8789899051189423,
      "rewards/forensics_sinkhorn_reward": 0.5494737923145294,
      "rewards/forensics_strict_edge_reward": 0.0012622844078578055,
      "rewards/forensics_tiou_reward": 0.24769359081983566,
      "step": 38
    },
    {
      "completion_length": 21.296875,
      "epoch": 0.16317991631799164,
      "grad_norm": 4.2329277992248535,
      "kl": 0.22119140625,
      "learning_rate": 9.602510460251046e-07,
      "loss": 0.1445,
      "reward": 0.6534326374530792,
      "reward_std": 0.11468060687184334,
      "reward_std_all": 0.1297196000814438,
      "rewards-max/forensics_sinkhorn_reward": 0.9552069008350372,
      "rewards-max/forensics_strict_edge_reward": 0.08560299407690763,
      "rewards-max/forensics_tiou_reward": 0.900328516960144,
      "rewards/forensics_sinkhorn_reward": 0.5774612426757812,
      "rewards/forensics_strict_edge_reward": 0.004521023540291935,
      "rewards/forensics_tiou_reward": 0.27893099188804626,
      "step": 39
    },
    {
      "completion_length": 19.984375,
      "epoch": 0.16736401673640167,
      "grad_norm": 5.188544750213623,
      "kl": 0.2421875,
      "learning_rate": 9.592050209205021e-07,
      "loss": 0.1954,
      "reward": 0.6756670773029327,
      "reward_std": 0.12596242129802704,
      "reward_std_all": 0.12539863213896751,
      "rewards-max/forensics_sinkhorn_reward": 0.8816106915473938,
      "rewards-max/forensics_strict_edge_reward": 0.00157204206334427,
      "rewards-max/forensics_tiou_reward": 0.7144743204116821,
      "rewards/forensics_sinkhorn_reward": 0.5682973265647888,
      "rewards/forensics_strict_edge_reward": 8.114692900562659e-05,
      "rewards/forensics_tiou_reward": 0.22246994078159332,
      "step": 40
    },
    {
      "completion_length": 22.25,
      "epoch": 0.17154811715481172,
      "grad_norm": 4.207794189453125,
      "kl": 0.2744140625,
      "learning_rate": 9.581589958158996e-07,
      "loss": 0.0644,
      "reward": 0.6028891801834106,
      "reward_std": 0.14295969903469086,
      "reward_std_all": 0.1674063578248024,
      "rewards-max/forensics_sinkhorn_reward": 0.887496143579483,
      "rewards-max/forensics_strict_edge_reward": 0.27338163246167824,
      "rewards-max/forensics_tiou_reward": 0.7251083850860596,
      "rewards/forensics_sinkhorn_reward": 0.5020026862621307,
      "rewards/forensics_strict_edge_reward": 0.011971099746006075,
      "rewards/forensics_tiou_reward": 0.1788572035729885,
      "step": 41
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.17573221757322174,
      "grad_norm": 5.211370468139648,
      "kl": 0.25537109375,
      "learning_rate": 9.571129707112971e-07,
      "loss": 0.147,
      "reward": 0.6365690231323242,
      "reward_std": 0.12681538239121437,
      "reward_std_all": 0.14109590649604797,
      "rewards-max/forensics_sinkhorn_reward": 0.8519799113273621,
      "rewards-max/forensics_strict_edge_reward": 0.0025790897925617173,
      "rewards-max/forensics_tiou_reward": 0.6862936317920685,
      "rewards/forensics_sinkhorn_reward": 0.5249125808477402,
      "rewards/forensics_strict_edge_reward": 0.0001614118828001665,
      "rewards/forensics_tiou_reward": 0.17632800340652466,
      "step": 42
    },
    {
      "completion_length": 20.125,
      "epoch": 0.1799163179916318,
      "grad_norm": 5.289707660675049,
      "kl": 0.24365234375,
      "learning_rate": 9.560669456066946e-07,
      "loss": -0.0344,
      "reward": 0.6596375107765198,
      "reward_std": 0.14368535578250885,
      "reward_std_all": 0.1412343829870224,
      "rewards-max/forensics_sinkhorn_reward": 0.9615763127803802,
      "rewards-max/forensics_strict_edge_reward": 0.2263733632862568,
      "rewards-max/forensics_tiou_reward": 0.924135148525238,
      "rewards/forensics_sinkhorn_reward": 0.565334677696228,
      "rewards/forensics_strict_edge_reward": 0.011905692284926772,
      "rewards/forensics_tiou_reward": 0.22317411750555038,
      "step": 43
    },
    {
      "completion_length": 20.203125,
      "epoch": 0.18410041841004185,
      "grad_norm": 4.410024642944336,
      "kl": 0.23486328125,
      "learning_rate": 9.55020920502092e-07,
      "loss": 0.043,
      "reward": 0.6797577440738678,
      "reward_std": 0.11416104808449745,
      "reward_std_all": 0.1151411347091198,
      "rewards-max/forensics_sinkhorn_reward": 0.9667483270168304,
      "rewards-max/forensics_strict_edge_reward": 0.1793062537908554,
      "rewards-max/forensics_tiou_reward": 0.9414298832416534,
      "rewards/forensics_sinkhorn_reward": 0.5846109986305237,
      "rewards/forensics_strict_edge_reward": 0.009644224075600505,
      "rewards/forensics_tiou_reward": 0.24799740314483643,
      "step": 44
    },
    {
      "completion_length": 20.53125,
      "epoch": 0.18828451882845187,
      "grad_norm": 3.8779349327087402,
      "kl": 0.22802734375,
      "learning_rate": 9.539748953974895e-07,
      "loss": 0.1482,
      "reward": 0.6879991292953491,
      "reward_std": 0.125468909740448,
      "reward_std_all": 0.12095476686954498,
      "rewards-max/forensics_sinkhorn_reward": 0.8750619292259216,
      "rewards-max/forensics_strict_edge_reward": 0.011319699813611805,
      "rewards-max/forensics_tiou_reward": 0.7210803329944611,
      "rewards/forensics_sinkhorn_reward": 0.581699013710022,
      "rewards/forensics_strict_edge_reward": 0.0005602722631010693,
      "rewards/forensics_tiou_reward": 0.2337348833680153,
      "step": 45
    },
    {
      "completion_length": 20.5625,
      "epoch": 0.19246861924686193,
      "grad_norm": 4.53864860534668,
      "kl": 0.21337890625,
      "learning_rate": 9.52928870292887e-07,
      "loss": 0.196,
      "reward": 0.6114706993103027,
      "reward_std": 0.12770162522792816,
      "reward_std_all": 0.1537434086203575,
      "rewards-max/forensics_sinkhorn_reward": 0.9429401755332947,
      "rewards-max/forensics_strict_edge_reward": 0.3002916472032666,
      "rewards-max/forensics_tiou_reward": 0.8740258514881134,
      "rewards/forensics_sinkhorn_reward": 0.5140936076641083,
      "rewards/forensics_strict_edge_reward": 0.013345849700272083,
      "rewards/forensics_tiou_reward": 0.18886880576610565,
      "step": 46
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.19665271966527198,
      "grad_norm": 5.114710807800293,
      "kl": 0.1982421875,
      "learning_rate": 9.518828451882844e-07,
      "loss": 0.1334,
      "reward": 0.6458286046981812,
      "reward_std": 0.11745278164744377,
      "reward_std_all": 0.1347862184047699,
      "rewards-max/forensics_sinkhorn_reward": 0.9749543070793152,
      "rewards-max/forensics_strict_edge_reward": 0.08208499848842621,
      "rewards-max/forensics_tiou_reward": 0.957763671875,
      "rewards/forensics_sinkhorn_reward": 0.5791595876216888,
      "rewards/forensics_strict_edge_reward": 0.003670252044685185,
      "rewards/forensics_tiou_reward": 0.26474547386169434,
      "step": 47
    },
    {
      "completion_length": 22.46875,
      "epoch": 0.200836820083682,
      "grad_norm": 5.634273052215576,
      "kl": 0.2607421875,
      "learning_rate": 9.508368200836819e-07,
      "loss": 0.1594,
      "reward": 0.619413822889328,
      "reward_std": 0.1608053594827652,
      "reward_std_all": 0.16722533106803894,
      "rewards-max/forensics_sinkhorn_reward": 0.863447904586792,
      "rewards-max/forensics_strict_edge_reward": 0.021956381853669882,
      "rewards-max/forensics_tiou_reward": 0.722783625125885,
      "rewards/forensics_sinkhorn_reward": 0.5123143494129181,
      "rewards/forensics_strict_edge_reward": 0.0010464201623108238,
      "rewards/forensics_tiou_reward": 0.18665818870067596,
      "step": 48
    },
    {
      "completion_length": 20.765625,
      "epoch": 0.20502092050209206,
      "grad_norm": 4.7134599685668945,
      "kl": 0.22607421875,
      "learning_rate": 9.497907949790794e-07,
      "loss": 0.2171,
      "reward": 0.6533156931400299,
      "reward_std": 0.10379528254270554,
      "reward_std_all": 0.12343642860651016,
      "rewards-max/forensics_sinkhorn_reward": 0.9675306975841522,
      "rewards-max/forensics_strict_edge_reward": 0.12087441619951278,
      "rewards-max/forensics_tiou_reward": 0.9305368065834045,
      "rewards/forensics_sinkhorn_reward": 0.5517948269844055,
      "rewards/forensics_strict_edge_reward": 0.005067337442596909,
      "rewards/forensics_tiou_reward": 0.20875757932662964,
      "step": 49
    },
    {
      "completion_length": 21.375,
      "epoch": 0.20920502092050208,
      "grad_norm": 5.016998291015625,
      "kl": 0.24560546875,
      "learning_rate": 9.487447698744769e-07,
      "loss": 0.1693,
      "reward": 0.6100812256336212,
      "reward_std": 0.13398199528455734,
      "reward_std_all": 0.15691925585269928,
      "rewards-max/forensics_sinkhorn_reward": 0.8965475559234619,
      "rewards-max/forensics_strict_edge_reward": 0.06772164323047036,
      "rewards-max/forensics_tiou_reward": 0.7645878791809082,
      "rewards/forensics_sinkhorn_reward": 0.5079483389854431,
      "rewards/forensics_strict_edge_reward": 0.002824137760399026,
      "rewards/forensics_tiou_reward": 0.1967230960726738,
      "step": 50
    },
    {
      "completion_length": 21.296875,
      "epoch": 0.21338912133891214,
      "grad_norm": 4.532074451446533,
      "kl": 0.2197265625,
      "learning_rate": 9.476987447698744e-07,
      "loss": 0.2075,
      "reward": 0.6673320233821869,
      "reward_std": 0.12556159496307373,
      "reward_std_all": 0.13058848679065704,
      "rewards-max/forensics_sinkhorn_reward": 0.9682750403881073,
      "rewards-max/forensics_strict_edge_reward": 0.3275787755846977,
      "rewards-max/forensics_tiou_reward": 0.9378870129585266,
      "rewards/forensics_sinkhorn_reward": 0.595645546913147,
      "rewards/forensics_strict_edge_reward": 0.016914385836571455,
      "rewards/forensics_tiou_reward": 0.29697467386722565,
      "step": 51
    },
    {
      "completion_length": 21.171875,
      "epoch": 0.2175732217573222,
      "grad_norm": 5.5137529373168945,
      "kl": 0.2265625,
      "learning_rate": 9.466527196652719e-07,
      "loss": 0.094,
      "reward": 0.6590143144130707,
      "reward_std": 0.11986365169286728,
      "reward_std_all": 0.13024285435676575,
      "rewards-max/forensics_sinkhorn_reward": 0.8629155457019806,
      "rewards-max/forensics_strict_edge_reward": 0.057902876287698746,
      "rewards-max/forensics_tiou_reward": 0.7297886908054352,
      "rewards/forensics_sinkhorn_reward": 0.5601779520511627,
      "rewards/forensics_strict_edge_reward": 0.0031845843186601996,
      "rewards/forensics_tiou_reward": 0.2358589991927147,
      "step": 52
    },
    {
      "completion_length": 23.4375,
      "epoch": 0.2217573221757322,
      "grad_norm": 5.338464736938477,
      "kl": 0.228515625,
      "learning_rate": 9.456066945606695e-07,
      "loss": 0.1126,
      "reward": 0.6723293364048004,
      "reward_std": 0.13212240114808083,
      "reward_std_all": 0.13295139372348785,
      "rewards-max/forensics_sinkhorn_reward": 0.8784101605415344,
      "rewards-max/forensics_strict_edge_reward": 0.03591936103065052,
      "rewards-max/forensics_tiou_reward": 0.7108279466629028,
      "rewards/forensics_sinkhorn_reward": 0.5693820118904114,
      "rewards/forensics_strict_edge_reward": 0.001612938932764818,
      "rewards/forensics_tiou_reward": 0.21229668706655502,
      "step": 53
    },
    {
      "completion_length": 21.796875,
      "epoch": 0.22594142259414227,
      "grad_norm": 5.275298595428467,
      "kl": 0.25244140625,
      "learning_rate": 9.44560669456067e-07,
      "loss": 0.1506,
      "reward": 0.7277972102165222,
      "reward_std": 0.07886917144060135,
      "reward_std_all": 0.0775708444416523,
      "rewards-max/forensics_sinkhorn_reward": 0.980102926492691,
      "rewards-max/forensics_strict_edge_reward": 0.5925651341676712,
      "rewards-max/forensics_tiou_reward": 0.9755702912807465,
      "rewards/forensics_sinkhorn_reward": 0.6251388788223267,
      "rewards/forensics_strict_edge_reward": 0.027954189106822014,
      "rewards/forensics_tiou_reward": 0.3037734627723694,
      "step": 54
    },
    {
      "completion_length": 23.296875,
      "epoch": 0.2301255230125523,
      "grad_norm": 5.206155776977539,
      "kl": 0.2109375,
      "learning_rate": 9.435146443514645e-07,
      "loss": 0.0831,
      "reward": 0.6791204512119293,
      "reward_std": 0.11839919909834862,
      "reward_std_all": 0.11910323426127434,
      "rewards-max/forensics_sinkhorn_reward": 0.8808932602405548,
      "rewards-max/forensics_strict_edge_reward": 0.0010022993083111942,
      "rewards-max/forensics_tiou_reward": 0.7078559696674347,
      "rewards/forensics_sinkhorn_reward": 0.5626018643379211,
      "rewards/forensics_strict_edge_reward": 4.664097104978282e-05,
      "rewards/forensics_tiou_reward": 0.2407548874616623,
      "step": 55
    },
    {
      "completion_length": 22.5625,
      "epoch": 0.23430962343096234,
      "grad_norm": 5.206369876861572,
      "kl": 0.2294921875,
      "learning_rate": 9.424686192468618e-07,
      "loss": 0.1482,
      "reward": 0.5993455648422241,
      "reward_std": 0.13271857053041458,
      "reward_std_all": 0.1619618982076645,
      "rewards-max/forensics_sinkhorn_reward": 0.8546276986598969,
      "rewards-max/forensics_strict_edge_reward": 0.01367313388891489,
      "rewards-max/forensics_tiou_reward": 0.665997639298439,
      "rewards/forensics_sinkhorn_reward": 0.49246346950531006,
      "rewards/forensics_strict_edge_reward": 0.0007607259482256268,
      "rewards/forensics_tiou_reward": 0.1982407085597515,
      "step": 56
    },
    {
      "completion_length": 20.421875,
      "epoch": 0.2384937238493724,
      "grad_norm": 4.750775337219238,
      "kl": 0.279296875,
      "learning_rate": 9.414225941422593e-07,
      "loss": 0.1659,
      "reward": 0.6659110188484192,
      "reward_std": 0.0862639732658863,
      "reward_std_all": 0.11333784833550453,
      "rewards-max/forensics_sinkhorn_reward": 0.9591612815856934,
      "rewards-max/forensics_strict_edge_reward": 0.2431606389582157,
      "rewards-max/forensics_tiou_reward": 0.9219000339508057,
      "rewards/forensics_sinkhorn_reward": 0.5892727673053741,
      "rewards/forensics_strict_edge_reward": 0.010166685096919537,
      "rewards/forensics_tiou_reward": 0.2978736162185669,
      "step": 57
    },
    {
      "completion_length": 23.515625,
      "epoch": 0.24267782426778242,
      "grad_norm": 4.673685550689697,
      "kl": 0.19580078125,
      "learning_rate": 9.403765690376568e-07,
      "loss": 0.1644,
      "reward": 0.6731643378734589,
      "reward_std": 0.09651222825050354,
      "reward_std_all": 0.11152392998337746,
      "rewards-max/forensics_sinkhorn_reward": 0.9300321638584137,
      "rewards-max/forensics_strict_edge_reward": 0.1368244830518961,
      "rewards-max/forensics_tiou_reward": 0.8481764197349548,
      "rewards/forensics_sinkhorn_reward": 0.5822968482971191,
      "rewards/forensics_strict_edge_reward": 0.006129974091891199,
      "rewards/forensics_tiou_reward": 0.26804229617118835,
      "step": 58
    },
    {
      "completion_length": 20.140625,
      "epoch": 0.24686192468619247,
      "grad_norm": 4.632630825042725,
      "kl": 0.208984375,
      "learning_rate": 9.393305439330543e-07,
      "loss": 0.1129,
      "reward": 0.7301389276981354,
      "reward_std": 0.0815749354660511,
      "reward_std_all": 0.07990891486406326,
      "rewards-max/forensics_sinkhorn_reward": 0.9761570990085602,
      "rewards-max/forensics_strict_edge_reward": 0.3990534320473671,
      "rewards-max/forensics_tiou_reward": 0.9586207866668701,
      "rewards/forensics_sinkhorn_reward": 0.6300190687179565,
      "rewards/forensics_strict_edge_reward": 0.019151471555233,
      "rewards/forensics_tiou_reward": 0.31051546335220337,
      "step": 59
    },
    {
      "completion_length": 22.25,
      "epoch": 0.2510460251046025,
      "grad_norm": 4.024607181549072,
      "kl": 0.2216796875,
      "learning_rate": 9.382845188284518e-07,
      "loss": 0.0637,
      "reward": 0.6828815340995789,
      "reward_std": 0.10919731855392456,
      "reward_std_all": 0.11855620890855789,
      "rewards-max/forensics_sinkhorn_reward": 0.9502929747104645,
      "rewards-max/forensics_strict_edge_reward": 0.0018421387576381676,
      "rewards-max/forensics_tiou_reward": 0.870246410369873,
      "rewards/forensics_sinkhorn_reward": 0.5742321908473969,
      "rewards/forensics_strict_edge_reward": 0.00015846392821572408,
      "rewards/forensics_tiou_reward": 0.2825285792350769,
      "step": 60
    },
    {
      "completion_length": 21.078125,
      "epoch": 0.25523012552301255,
      "grad_norm": 4.762994766235352,
      "kl": 0.21923828125,
      "learning_rate": 9.372384937238493e-07,
      "loss": 0.2247,
      "reward": 0.7182239890098572,
      "reward_std": 0.09097680076956749,
      "reward_std_all": 0.08890450745820999,
      "rewards-max/forensics_sinkhorn_reward": 0.9579527974128723,
      "rewards-max/forensics_strict_edge_reward": 0.2375368531793356,
      "rewards-max/forensics_tiou_reward": 0.9210996925830841,
      "rewards/forensics_sinkhorn_reward": 0.6321642398834229,
      "rewards/forensics_strict_edge_reward": 0.018251163884997368,
      "rewards/forensics_tiou_reward": 0.31780625879764557,
      "step": 61
    },
    {
      "completion_length": 21.1875,
      "epoch": 0.2594142259414226,
      "grad_norm": 5.080166339874268,
      "kl": 0.30419921875,
      "learning_rate": 9.361924686192469e-07,
      "loss": 0.1349,
      "reward": 0.5979130864143372,
      "reward_std": 0.16044993698596954,
      "reward_std_all": 0.178270623087883,
      "rewards-max/forensics_sinkhorn_reward": 0.8981431424617767,
      "rewards-max/forensics_strict_edge_reward": 0.006977253477089107,
      "rewards-max/forensics_tiou_reward": 0.7796753942966461,
      "rewards/forensics_sinkhorn_reward": 0.4785263240337372,
      "rewards/forensics_strict_edge_reward": 0.00030675571906613186,
      "rewards/forensics_tiou_reward": 0.16056979447603226,
      "step": 62
    },
    {
      "completion_length": 20.375,
      "epoch": 0.26359832635983266,
      "grad_norm": 5.027713298797607,
      "kl": 0.22412109375,
      "learning_rate": 9.351464435146444e-07,
      "loss": 0.1933,
      "reward": 0.6784601509571075,
      "reward_std": 0.09679830074310303,
      "reward_std_all": 0.10780464112758636,
      "rewards-max/forensics_sinkhorn_reward": 0.8975608944892883,
      "rewards-max/forensics_strict_edge_reward": 0.014576543122529984,
      "rewards-max/forensics_tiou_reward": 0.7539146840572357,
      "rewards/forensics_sinkhorn_reward": 0.581847220659256,
      "rewards/forensics_strict_edge_reward": 0.0009154685249086469,
      "rewards/forensics_tiou_reward": 0.24767781794071198,
      "step": 63
    },
    {
      "completion_length": 21.125,
      "epoch": 0.26778242677824265,
      "grad_norm": 5.745501518249512,
      "kl": 0.2265625,
      "learning_rate": 9.341004184100419e-07,
      "loss": 0.1163,
      "reward": 0.6584150493144989,
      "reward_std": 0.15884008258581161,
      "reward_std_all": 0.15649863332509995,
      "rewards-max/forensics_sinkhorn_reward": 0.9268203675746918,
      "rewards-max/forensics_strict_edge_reward": 0.21729913674604617,
      "rewards-max/forensics_tiou_reward": 0.8248943090438843,
      "rewards/forensics_sinkhorn_reward": 0.5683234632015228,
      "rewards/forensics_strict_edge_reward": 0.010724020985265703,
      "rewards/forensics_tiou_reward": 0.2837757170200348,
      "step": 64
    },
    {
      "completion_length": 20.640625,
      "epoch": 0.2719665271966527,
      "grad_norm": 4.454521179199219,
      "kl": 0.24365234375,
      "learning_rate": 9.330543933054393e-07,
      "loss": 0.1176,
      "reward": 0.6689749360084534,
      "reward_std": 0.1526164561510086,
      "reward_std_all": 0.14506475254893303,
      "rewards-max/forensics_sinkhorn_reward": 0.8634263575077057,
      "rewards-max/forensics_strict_edge_reward": 0.01775200664997101,
      "rewards-max/forensics_tiou_reward": 0.6980620324611664,
      "rewards/forensics_sinkhorn_reward": 0.5603163838386536,
      "rewards/forensics_strict_edge_reward": 0.001269014537683688,
      "rewards/forensics_tiou_reward": 0.22811435163021088,
      "step": 65
    },
    {
      "completion_length": 21.015625,
      "epoch": 0.27615062761506276,
      "grad_norm": 5.719748497009277,
      "kl": 0.2158203125,
      "learning_rate": 9.320083682008367e-07,
      "loss": 0.2641,
      "reward": 0.6344664692878723,
      "reward_std": 0.14240466058254242,
      "reward_std_all": 0.15141789615154266,
      "rewards-max/forensics_sinkhorn_reward": 0.9524237215518951,
      "rewards-max/forensics_strict_edge_reward": 0.06781813129782677,
      "rewards-max/forensics_tiou_reward": 0.8968428671360016,
      "rewards/forensics_sinkhorn_reward": 0.5390256345272064,
      "rewards/forensics_strict_edge_reward": 0.004505821503698826,
      "rewards/forensics_tiou_reward": 0.241160586476326,
      "step": 66
    },
    {
      "completion_length": 21.53125,
      "epoch": 0.2803347280334728,
      "grad_norm": 5.051743984222412,
      "kl": 0.22705078125,
      "learning_rate": 9.309623430962342e-07,
      "loss": 0.2597,
      "reward": 0.6766430735588074,
      "reward_std": 0.09313830733299255,
      "reward_std_all": 0.10572448745369911,
      "rewards-max/forensics_sinkhorn_reward": 0.9719986319541931,
      "rewards-max/forensics_strict_edge_reward": 0.5368446409702301,
      "rewards-max/forensics_tiou_reward": 0.9570841193199158,
      "rewards/forensics_sinkhorn_reward": 0.5901807844638824,
      "rewards/forensics_strict_edge_reward": 0.028182176873087883,
      "rewards/forensics_tiou_reward": 0.281740665435791,
      "step": 67
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.28451882845188287,
      "grad_norm": 4.979437351226807,
      "kl": 0.25830078125,
      "learning_rate": 9.299163179916317e-07,
      "loss": 0.1796,
      "reward": 0.6267316937446594,
      "reward_std": 0.10599355772137642,
      "reward_std_all": 0.13725901395082474,
      "rewards-max/forensics_sinkhorn_reward": 0.9204050600528717,
      "rewards-max/forensics_strict_edge_reward": 0.1966366252527223,
      "rewards-max/forensics_tiou_reward": 0.8553414344787598,
      "rewards/forensics_sinkhorn_reward": 0.5390684008598328,
      "rewards/forensics_strict_edge_reward": 0.013860910022174266,
      "rewards/forensics_tiou_reward": 0.24042723327875137,
      "step": 68
    },
    {
      "completion_length": 21.484375,
      "epoch": 0.28870292887029286,
      "grad_norm": 4.094674587249756,
      "kl": 0.18603515625,
      "learning_rate": 9.288702928870292e-07,
      "loss": 0.1434,
      "reward": 0.7198528349399567,
      "reward_std": 0.07315649464726448,
      "reward_std_all": 0.08010029047727585,
      "rewards-max/forensics_sinkhorn_reward": 0.9360203742980957,
      "rewards-max/forensics_strict_edge_reward": 0.07134917099028826,
      "rewards-max/forensics_tiou_reward": 0.8681119680404663,
      "rewards/forensics_sinkhorn_reward": 0.6439867913722992,
      "rewards/forensics_strict_edge_reward": 0.004145982791669667,
      "rewards/forensics_tiou_reward": 0.3228811025619507,
      "step": 69
    },
    {
      "completion_length": 21.859375,
      "epoch": 0.2928870292887029,
      "grad_norm": 3.987492084503174,
      "kl": 0.20947265625,
      "learning_rate": 9.278242677824268e-07,
      "loss": 0.1657,
      "reward": 0.6496109962463379,
      "reward_std": 0.11306225880980492,
      "reward_std_all": 0.12909665703773499,
      "rewards-max/forensics_sinkhorn_reward": 0.9544736444950104,
      "rewards-max/forensics_strict_edge_reward": 0.06720857508480549,
      "rewards-max/forensics_tiou_reward": 0.9154762625694275,
      "rewards/forensics_sinkhorn_reward": 0.553300142288208,
      "rewards/forensics_strict_edge_reward": 0.0033422360429540277,
      "rewards/forensics_tiou_reward": 0.2213367372751236,
      "step": 70
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.29707112970711297,
      "grad_norm": 5.4326043128967285,
      "kl": 0.185546875,
      "learning_rate": 9.267782426778243e-07,
      "loss": 0.1527,
      "reward": 0.682671457529068,
      "reward_std": 0.09636973589658737,
      "reward_std_all": 0.10728464648127556,
      "rewards-max/forensics_sinkhorn_reward": 0.871088981628418,
      "rewards-max/forensics_strict_edge_reward": 0.061446369625627995,
      "rewards-max/forensics_tiou_reward": 0.7651019096374512,
      "rewards/forensics_sinkhorn_reward": 0.5882611572742462,
      "rewards/forensics_strict_edge_reward": 0.00360623057349585,
      "rewards/forensics_tiou_reward": 0.2518386170268059,
      "step": 71
    },
    {
      "completion_length": 21.609375,
      "epoch": 0.301255230125523,
      "grad_norm": 3.7089383602142334,
      "kl": 0.1748046875,
      "learning_rate": 9.257322175732218e-07,
      "loss": 0.1309,
      "reward": 0.6964947283267975,
      "reward_std": 0.07343300245702267,
      "reward_std_all": 0.09088043123483658,
      "rewards-max/forensics_sinkhorn_reward": 0.949508935213089,
      "rewards-max/forensics_strict_edge_reward": 0.03999222535640001,
      "rewards-max/forensics_tiou_reward": 0.8753885924816132,
      "rewards/forensics_sinkhorn_reward": 0.6122103035449982,
      "rewards/forensics_strict_edge_reward": 0.0019446330843493342,
      "rewards/forensics_tiou_reward": 0.30385030806064606,
      "step": 72
    },
    {
      "completion_length": 20.5625,
      "epoch": 0.3054393305439331,
      "grad_norm": 4.109133243560791,
      "kl": 0.20166015625,
      "learning_rate": 9.246861924686192e-07,
      "loss": 0.1541,
      "reward": 0.6282556653022766,
      "reward_std": 0.11263743042945862,
      "reward_std_all": 0.13625090941786766,
      "rewards-max/forensics_sinkhorn_reward": 0.8976479768753052,
      "rewards-max/forensics_strict_edge_reward": 0.050262245727935806,
      "rewards-max/forensics_tiou_reward": 0.7530065476894379,
      "rewards/forensics_sinkhorn_reward": 0.530043214559555,
      "rewards/forensics_strict_edge_reward": 0.0027267261502856854,
      "rewards/forensics_tiou_reward": 0.20726972818374634,
      "step": 73
    },
    {
      "completion_length": 22.125,
      "epoch": 0.30962343096234307,
      "grad_norm": 4.816046714782715,
      "kl": 0.2021484375,
      "learning_rate": 9.236401673640167e-07,
      "loss": 0.0451,
      "reward": 0.6714005470275879,
      "reward_std": 0.1158982366323471,
      "reward_std_all": 0.12107151374220848,
      "rewards-max/forensics_sinkhorn_reward": 0.9746144413948059,
      "rewards-max/forensics_strict_edge_reward": 0.5842172801494598,
      "rewards-max/forensics_tiou_reward": 0.958534449338913,
      "rewards/forensics_sinkhorn_reward": 0.5790830850601196,
      "rewards/forensics_strict_edge_reward": 0.03785602655261755,
      "rewards/forensics_tiou_reward": 0.3286350816488266,
      "step": 74
    },
    {
      "completion_length": 20.078125,
      "epoch": 0.3138075313807531,
      "grad_norm": 7.104576587677002,
      "kl": 0.19677734375,
      "learning_rate": 9.225941422594142e-07,
      "loss": 0.096,
      "reward": 0.6697810292243958,
      "reward_std": 0.11866747960448265,
      "reward_std_all": 0.12342453747987747,
      "rewards-max/forensics_sinkhorn_reward": 0.9765456318855286,
      "rewards-max/forensics_strict_edge_reward": 0.3001514635980129,
      "rewards-max/forensics_tiou_reward": 0.9541797935962677,
      "rewards/forensics_sinkhorn_reward": 0.5827585458755493,
      "rewards/forensics_strict_edge_reward": 0.01870706002227962,
      "rewards/forensics_tiou_reward": 0.2940242439508438,
      "step": 75
    },
    {
      "completion_length": 22.796875,
      "epoch": 0.3179916317991632,
      "grad_norm": 6.126336097717285,
      "kl": 0.23828125,
      "learning_rate": 9.215481171548116e-07,
      "loss": 0.1029,
      "reward": 0.7059115767478943,
      "reward_std": 0.1406630575656891,
      "reward_std_all": 0.12751509249210358,
      "rewards-max/forensics_sinkhorn_reward": 0.9180253744125366,
      "rewards-max/forensics_strict_edge_reward": 0.035715402103960514,
      "rewards-max/forensics_tiou_reward": 0.7966075837612152,
      "rewards/forensics_sinkhorn_reward": 0.6376602053642273,
      "rewards/forensics_strict_edge_reward": 0.0030385271966224536,
      "rewards/forensics_tiou_reward": 0.34593553841114044,
      "step": 76
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.32217573221757323,
      "grad_norm": 4.837143421173096,
      "kl": 0.21533203125,
      "learning_rate": 9.205020920502091e-07,
      "loss": 0.1364,
      "reward": 0.7271449863910675,
      "reward_std": 0.07706426084041595,
      "reward_std_all": 0.07648396119475365,
      "rewards-max/forensics_sinkhorn_reward": 0.9367554783821106,
      "rewards-max/forensics_strict_edge_reward": 0.1614157957956195,
      "rewards-max/forensics_tiou_reward": 0.8576315343379974,
      "rewards/forensics_sinkhorn_reward": 0.6373910903930664,
      "rewards/forensics_strict_edge_reward": 0.01220685406588018,
      "rewards/forensics_tiou_reward": 0.3139156401157379,
      "step": 77
    },
    {
      "completion_length": 22.5625,
      "epoch": 0.3263598326359833,
      "grad_norm": 5.1419219970703125,
      "kl": 0.2138671875,
      "learning_rate": 9.194560669456066e-07,
      "loss": 0.1435,
      "reward": 0.6448645889759064,
      "reward_std": 0.13849198073148727,
      "reward_std_all": 0.14245998114347458,
      "rewards-max/forensics_sinkhorn_reward": 0.8552348911762238,
      "rewards-max/forensics_strict_edge_reward": 0.38484400510787964,
      "rewards-max/forensics_tiou_reward": 0.7002162933349609,
      "rewards/forensics_sinkhorn_reward": 0.5224000215530396,
      "rewards/forensics_strict_edge_reward": 0.016041044611483812,
      "rewards/forensics_tiou_reward": 0.2373434081673622,
      "step": 78
    },
    {
      "completion_length": 22.484375,
      "epoch": 0.3305439330543933,
      "grad_norm": 5.5252509117126465,
      "kl": 0.22265625,
      "learning_rate": 9.184100418410042e-07,
      "loss": 0.1575,
      "reward": 0.6536580920219421,
      "reward_std": 0.13512200862169266,
      "reward_std_all": 0.14241930097341537,
      "rewards-max/forensics_sinkhorn_reward": 0.9564319550991058,
      "rewards-max/forensics_strict_edge_reward": 0.05682708881795406,
      "rewards-max/forensics_tiou_reward": 0.9129817485809326,
      "rewards/forensics_sinkhorn_reward": 0.5583759248256683,
      "rewards/forensics_strict_edge_reward": 0.004664639418479055,
      "rewards/forensics_tiou_reward": 0.2746967226266861,
      "step": 79
    },
    {
      "completion_length": 21.40625,
      "epoch": 0.33472803347280333,
      "grad_norm": 4.397517681121826,
      "kl": 0.20361328125,
      "learning_rate": 9.173640167364017e-07,
      "loss": 0.1225,
      "reward": 0.7290787994861603,
      "reward_std": 0.06922095268964767,
      "reward_std_all": 0.07350296899676323,
      "rewards-max/forensics_sinkhorn_reward": 0.9139866828918457,
      "rewards-max/forensics_strict_edge_reward": 0.11836010962724686,
      "rewards-max/forensics_tiou_reward": 0.8050559759140015,
      "rewards/forensics_sinkhorn_reward": 0.6429221332073212,
      "rewards/forensics_strict_edge_reward": 0.006072340067476034,
      "rewards/forensics_tiou_reward": 0.37603859603405,
      "step": 80
    },
    {
      "completion_length": 21.859375,
      "epoch": 0.3389121338912134,
      "grad_norm": 5.312783718109131,
      "kl": 0.2236328125,
      "learning_rate": 9.163179916317991e-07,
      "loss": 0.2156,
      "reward": 0.6751405298709869,
      "reward_std": 0.12607135623693466,
      "reward_std_all": 0.12457410991191864,
      "rewards-max/forensics_sinkhorn_reward": 0.9580439627170563,
      "rewards-max/forensics_strict_edge_reward": 0.19404058903455734,
      "rewards-max/forensics_tiou_reward": 0.9264831840991974,
      "rewards/forensics_sinkhorn_reward": 0.5720988512039185,
      "rewards/forensics_strict_edge_reward": 0.00823864876292646,
      "rewards/forensics_tiou_reward": 0.2743699699640274,
      "step": 81
    },
    {
      "completion_length": 21.90625,
      "epoch": 0.34309623430962344,
      "grad_norm": 5.064513206481934,
      "kl": 0.22216796875,
      "learning_rate": 9.152719665271966e-07,
      "loss": 0.0919,
      "reward": 0.6665081679821014,
      "reward_std": 0.07678267173469067,
      "reward_std_all": 0.10270706191658974,
      "rewards-max/forensics_sinkhorn_reward": 0.9590848982334137,
      "rewards-max/forensics_strict_edge_reward": 0.5413877367973328,
      "rewards-max/forensics_tiou_reward": 0.9190903604030609,
      "rewards/forensics_sinkhorn_reward": 0.5412905663251877,
      "rewards/forensics_strict_edge_reward": 0.025340155698359013,
      "rewards/forensics_tiou_reward": 0.294353187084198,
      "step": 82
    },
    {
      "completion_length": 20.953125,
      "epoch": 0.3472803347280335,
      "grad_norm": 4.709937572479248,
      "kl": 0.24267578125,
      "learning_rate": 9.142259414225941e-07,
      "loss": 0.1463,
      "reward": 0.6757222712039948,
      "reward_std": 0.11173951625823975,
      "reward_std_all": 0.11673816666007042,
      "rewards-max/forensics_sinkhorn_reward": 0.9601193070411682,
      "rewards-max/forensics_strict_edge_reward": 0.29384535178542137,
      "rewards-max/forensics_tiou_reward": 0.9207263886928558,
      "rewards/forensics_sinkhorn_reward": 0.5859001576900482,
      "rewards/forensics_strict_edge_reward": 0.022812772309407592,
      "rewards/forensics_tiou_reward": 0.29970644414424896,
      "step": 83
    },
    {
      "completion_length": 21.328125,
      "epoch": 0.3514644351464435,
      "grad_norm": 4.994503498077393,
      "kl": 0.23681640625,
      "learning_rate": 9.131799163179916e-07,
      "loss": 0.1707,
      "reward": 0.662704735994339,
      "reward_std": 0.10977324843406677,
      "reward_std_all": 0.1244874931871891,
      "rewards-max/forensics_sinkhorn_reward": 0.9750463962554932,
      "rewards-max/forensics_strict_edge_reward": 0.3478900271002203,
      "rewards-max/forensics_tiou_reward": 0.9537670314311981,
      "rewards/forensics_sinkhorn_reward": 0.5734789967536926,
      "rewards/forensics_strict_edge_reward": 0.023812207829905674,
      "rewards/forensics_tiou_reward": 0.28271549940109253,
      "step": 84
    },
    {
      "completion_length": 20.921875,
      "epoch": 0.35564853556485354,
      "grad_norm": 3.8024604320526123,
      "kl": 0.22216796875,
      "learning_rate": 9.121338912133891e-07,
      "loss": 0.0089,
      "reward": 0.6943402290344238,
      "reward_std": 0.08006172440946102,
      "reward_std_all": 0.09086372330784798,
      "rewards-max/forensics_sinkhorn_reward": 0.9721626937389374,
      "rewards-max/forensics_strict_edge_reward": 0.45472436398267746,
      "rewards-max/forensics_tiou_reward": 0.9487468004226685,
      "rewards/forensics_sinkhorn_reward": 0.6250142753124237,
      "rewards/forensics_strict_edge_reward": 0.019619297003373504,
      "rewards/forensics_tiou_reward": 0.3181489109992981,
      "step": 85
    },
    {
      "completion_length": 20.078125,
      "epoch": 0.3598326359832636,
      "grad_norm": 5.170497417449951,
      "kl": 0.27734375,
      "learning_rate": 9.110878661087865e-07,
      "loss": 0.084,
      "reward": 0.6407007575035095,
      "reward_std": 0.12770840153098106,
      "reward_std_all": 0.13962115347385406,
      "rewards-max/forensics_sinkhorn_reward": 0.9599659144878387,
      "rewards-max/forensics_strict_edge_reward": 0.08019497059285641,
      "rewards-max/forensics_tiou_reward": 0.920789897441864,
      "rewards/forensics_sinkhorn_reward": 0.5496941804885864,
      "rewards/forensics_strict_edge_reward": 0.0037413601530715823,
      "rewards/forensics_tiou_reward": 0.23838825523853302,
      "step": 86
    },
    {
      "completion_length": 21.65625,
      "epoch": 0.36401673640167365,
      "grad_norm": 5.050574779510498,
      "kl": 0.25390625,
      "learning_rate": 9.10041841004184e-07,
      "loss": 0.0259,
      "reward": 0.6372406482696533,
      "reward_std": 0.10545627772808075,
      "reward_std_all": 0.13739033043384552,
      "rewards-max/forensics_sinkhorn_reward": 0.9342870116233826,
      "rewards-max/forensics_strict_edge_reward": 0.02080150879919529,
      "rewards-max/forensics_tiou_reward": 0.8524895906448364,
      "rewards/forensics_sinkhorn_reward": 0.5480083525180817,
      "rewards/forensics_strict_edge_reward": 0.0012030673096887767,
      "rewards/forensics_tiou_reward": 0.28524988889694214,
      "step": 87
    },
    {
      "completion_length": 19.828125,
      "epoch": 0.3682008368200837,
      "grad_norm": 4.4591169357299805,
      "kl": 0.22900390625,
      "learning_rate": 9.089958158995816e-07,
      "loss": 0.0224,
      "reward": 0.6653301119804382,
      "reward_std": 0.08731289207935333,
      "reward_std_all": 0.10881058499217033,
      "rewards-max/forensics_sinkhorn_reward": 0.9555825293064117,
      "rewards-max/forensics_strict_edge_reward": 0.26956089213490486,
      "rewards-max/forensics_tiou_reward": 0.9096428453922272,
      "rewards/forensics_sinkhorn_reward": 0.5540394484996796,
      "rewards/forensics_strict_edge_reward": 0.013455796055495739,
      "rewards/forensics_tiou_reward": 0.23113801330327988,
      "step": 88
    },
    {
      "completion_length": 19.3125,
      "epoch": 0.3723849372384937,
      "grad_norm": 4.6533355712890625,
      "kl": 0.2666015625,
      "learning_rate": 9.07949790794979e-07,
      "loss": 0.0488,
      "reward": 0.6710956394672394,
      "reward_std": 0.15951555222272873,
      "reward_std_all": 0.14913024753332138,
      "rewards-max/forensics_sinkhorn_reward": 0.9428488612174988,
      "rewards-max/forensics_strict_edge_reward": 0.06692725513130426,
      "rewards-max/forensics_tiou_reward": 0.860750824213028,
      "rewards/forensics_sinkhorn_reward": 0.5844845175743103,
      "rewards/forensics_strict_edge_reward": 0.0029556637164205313,
      "rewards/forensics_tiou_reward": 0.24498775601387024,
      "step": 89
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.37656903765690375,
      "grad_norm": 4.949152946472168,
      "kl": 0.251953125,
      "learning_rate": 9.069037656903765e-07,
      "loss": 0.0369,
      "reward": 0.7337252497673035,
      "reward_std": 0.10828414559364319,
      "reward_std_all": 0.09761282801628113,
      "rewards-max/forensics_sinkhorn_reward": 0.8902495205402374,
      "rewards-max/forensics_strict_edge_reward": 0.034768542613164755,
      "rewards-max/forensics_tiou_reward": 0.762821763753891,
      "rewards/forensics_sinkhorn_reward": 0.6190416216850281,
      "rewards/forensics_strict_edge_reward": 0.0022575766056434077,
      "rewards/forensics_tiou_reward": 0.29667001962661743,
      "step": 90
    },
    {
      "completion_length": 21.03125,
      "epoch": 0.3807531380753138,
      "grad_norm": 4.59456729888916,
      "kl": 0.25927734375,
      "learning_rate": 9.05857740585774e-07,
      "loss": 0.2097,
      "reward": 0.7010719776153564,
      "reward_std": 0.09787792153656483,
      "reward_std_all": 0.10153741016983986,
      "rewards-max/forensics_sinkhorn_reward": 0.9220196306705475,
      "rewards-max/forensics_strict_edge_reward": 0.05740719285677187,
      "rewards-max/forensics_tiou_reward": 0.8405013084411621,
      "rewards/forensics_sinkhorn_reward": 0.6153140664100647,
      "rewards/forensics_strict_edge_reward": 0.0027091851052318816,
      "rewards/forensics_tiou_reward": 0.28421033173799515,
      "step": 91
    },
    {
      "completion_length": 21.296875,
      "epoch": 0.38493723849372385,
      "grad_norm": 4.726018905639648,
      "kl": 0.2470703125,
      "learning_rate": 9.048117154811715e-07,
      "loss": 0.0714,
      "reward": 0.685474306344986,
      "reward_std": 0.12287593632936478,
      "reward_std_all": 0.12266214191913605,
      "rewards-max/forensics_sinkhorn_reward": 0.8639401793479919,
      "rewards-max/forensics_strict_edge_reward": 0.029945429880172014,
      "rewards-max/forensics_tiou_reward": 0.742328554391861,
      "rewards/forensics_sinkhorn_reward": 0.5838521718978882,
      "rewards/forensics_strict_edge_reward": 0.0014292424893938005,
      "rewards/forensics_tiou_reward": 0.26613757014274597,
      "step": 92
    },
    {
      "completion_length": 20.21875,
      "epoch": 0.3891213389121339,
      "grad_norm": 5.325535297393799,
      "kl": 0.23583984375,
      "learning_rate": 9.03765690376569e-07,
      "loss": 0.0925,
      "reward": 0.7181781530380249,
      "reward_std": 0.07665946707129478,
      "reward_std_all": 0.08264621533453465,
      "rewards-max/forensics_sinkhorn_reward": 0.8981764912605286,
      "rewards-max/forensics_strict_edge_reward": 0.03157530818134546,
      "rewards-max/forensics_tiou_reward": 0.7828298211097717,
      "rewards/forensics_sinkhorn_reward": 0.6578004062175751,
      "rewards/forensics_strict_edge_reward": 0.0015454341773875058,
      "rewards/forensics_tiou_reward": 0.34747229516506195,
      "step": 93
    },
    {
      "completion_length": 21.421875,
      "epoch": 0.39330543933054396,
      "grad_norm": 4.550325870513916,
      "kl": 0.3125,
      "learning_rate": 9.027196652719665e-07,
      "loss": 0.1179,
      "reward": 0.6848328411579132,
      "reward_std": 0.0671102274209261,
      "reward_std_all": 0.09038657695055008,
      "rewards-max/forensics_sinkhorn_reward": 0.8508898317813873,
      "rewards-max/forensics_strict_edge_reward": 0.006149426455522189,
      "rewards-max/forensics_tiou_reward": 0.6039246916770935,
      "rewards/forensics_sinkhorn_reward": 0.5615029036998749,
      "rewards/forensics_strict_edge_reward": 0.0002605431752726872,
      "rewards/forensics_tiou_reward": 0.22016696631908417,
      "step": 94
    },
    {
      "completion_length": 20.90625,
      "epoch": 0.39748953974895396,
      "grad_norm": 4.319746494293213,
      "kl": 0.23681640625,
      "learning_rate": 9.01673640167364e-07,
      "loss": 0.1648,
      "reward": 0.7129985392093658,
      "reward_std": 0.07861711829900742,
      "reward_std_all": 0.08325027488172054,
      "rewards-max/forensics_sinkhorn_reward": 0.8059324026107788,
      "rewards-max/forensics_strict_edge_reward": 0.0007428072658512974,
      "rewards-max/forensics_tiou_reward": 0.5395344495773315,
      "rewards/forensics_sinkhorn_reward": 0.5954874455928802,
      "rewards/forensics_strict_edge_reward": 3.196786110493122e-05,
      "rewards/forensics_tiou_reward": 0.2638322412967682,
      "step": 95
    },
    {
      "completion_length": 21.484375,
      "epoch": 0.401673640167364,
      "grad_norm": 5.0072197914123535,
      "kl": 0.2626953125,
      "learning_rate": 9.006276150627615e-07,
      "loss": 0.2318,
      "reward": 0.6394062042236328,
      "reward_std": 0.1505112275481224,
      "reward_std_all": 0.15898612141609192,
      "rewards-max/forensics_sinkhorn_reward": 0.9533196985721588,
      "rewards-max/forensics_strict_edge_reward": 0.3373861415311694,
      "rewards-max/forensics_tiou_reward": 0.9155187010765076,
      "rewards/forensics_sinkhorn_reward": 0.5593265891075134,
      "rewards/forensics_strict_edge_reward": 0.014560383278876543,
      "rewards/forensics_tiou_reward": 0.2650410979986191,
      "step": 96
    },
    {
      "completion_length": 21.1875,
      "epoch": 0.40585774058577406,
      "grad_norm": 5.295456409454346,
      "kl": 0.23974609375,
      "learning_rate": 8.99581589958159e-07,
      "loss": 0.0836,
      "reward": 0.6627992689609528,
      "reward_std": 0.1477438062429428,
      "reward_std_all": 0.14439785480499268,
      "rewards-max/forensics_sinkhorn_reward": 0.907680332660675,
      "rewards-max/forensics_strict_edge_reward": 0.0020701547618955374,
      "rewards-max/forensics_tiou_reward": 0.7960499227046967,
      "rewards/forensics_sinkhorn_reward": 0.5589244663715363,
      "rewards/forensics_strict_edge_reward": 0.0001637969253351912,
      "rewards/forensics_tiou_reward": 0.27166545391082764,
      "step": 97
    },
    {
      "completion_length": 21.1875,
      "epoch": 0.4100418410041841,
      "grad_norm": 3.9955241680145264,
      "kl": 0.21337890625,
      "learning_rate": 8.985355648535564e-07,
      "loss": 0.2751,
      "reward": 0.6736164093017578,
      "reward_std": 0.09846961870789528,
      "reward_std_all": 0.1122640036046505,
      "rewards-max/forensics_sinkhorn_reward": 0.9498338103294373,
      "rewards-max/forensics_strict_edge_reward": 0.34021952375769615,
      "rewards-max/forensics_tiou_reward": 0.9079370498657227,
      "rewards/forensics_sinkhorn_reward": 0.5552278757095337,
      "rewards/forensics_strict_edge_reward": 0.014848697930574417,
      "rewards/forensics_tiou_reward": 0.22911080718040466,
      "step": 98
    },
    {
      "completion_length": 21.015625,
      "epoch": 0.41422594142259417,
      "grad_norm": 5.476307392120361,
      "kl": 0.25439453125,
      "learning_rate": 8.974895397489539e-07,
      "loss": 0.0844,
      "reward": 0.702102392911911,
      "reward_std": 0.11302658170461655,
      "reward_std_all": 0.11032155528664589,
      "rewards-max/forensics_sinkhorn_reward": 0.9426509141921997,
      "rewards-max/forensics_strict_edge_reward": 0.1466790735721588,
      "rewards-max/forensics_tiou_reward": 0.8827022910118103,
      "rewards/forensics_sinkhorn_reward": 0.6291083991527557,
      "rewards/forensics_strict_edge_reward": 0.008275959175080061,
      "rewards/forensics_tiou_reward": 0.33260226249694824,
      "step": 99
    },
    {
      "completion_length": 21.375,
      "epoch": 0.41841004184100417,
      "grad_norm": 4.973294734954834,
      "kl": 0.2880859375,
      "learning_rate": 8.964435146443514e-07,
      "loss": 0.0799,
      "reward": 0.6532832086086273,
      "reward_std": 0.1266600489616394,
      "reward_std_all": 0.13373124599456787,
      "rewards-max/forensics_sinkhorn_reward": 0.9483251869678497,
      "rewards-max/forensics_strict_edge_reward": 0.0259947394952178,
      "rewards-max/forensics_tiou_reward": 0.8755823373794556,
      "rewards/forensics_sinkhorn_reward": 0.5608988404273987,
      "rewards/forensics_strict_edge_reward": 0.0011233583791181445,
      "rewards/forensics_tiou_reward": 0.2307305932044983,
      "step": 100
    },
    {
      "completion_length": 19.5625,
      "epoch": 0.4225941422594142,
      "grad_norm": 4.6409406661987305,
      "kl": 0.2490234375,
      "learning_rate": 8.953974895397489e-07,
      "loss": 0.0923,
      "reward": 0.6870929300785065,
      "reward_std": 0.0801009014248848,
      "reward_std_all": 0.09594729915261269,
      "rewards-max/forensics_sinkhorn_reward": 0.9186701774597168,
      "rewards-max/forensics_strict_edge_reward": 0.12203247286379337,
      "rewards-max/forensics_tiou_reward": 0.7998349964618683,
      "rewards/forensics_sinkhorn_reward": 0.6038632392883301,
      "rewards/forensics_strict_edge_reward": 0.005969511577859521,
      "rewards/forensics_tiou_reward": 0.27341003715991974,
      "step": 101
    },
    {
      "completion_length": 20.953125,
      "epoch": 0.42677824267782427,
      "grad_norm": 5.030569553375244,
      "kl": 0.232421875,
      "learning_rate": 8.943514644351464e-07,
      "loss": 0.0683,
      "reward": 0.7151829302310944,
      "reward_std": 0.10205952823162079,
      "reward_std_all": 0.09806271269917488,
      "rewards-max/forensics_sinkhorn_reward": 0.8985743224620819,
      "rewards-max/forensics_strict_edge_reward": 0.23671214282512665,
      "rewards-max/forensics_tiou_reward": 0.778734564781189,
      "rewards/forensics_sinkhorn_reward": 0.6232226192951202,
      "rewards/forensics_strict_edge_reward": 0.013058011885732412,
      "rewards/forensics_tiou_reward": 0.31061188876628876,
      "step": 102
    },
    {
      "completion_length": 22.296875,
      "epoch": 0.4309623430962343,
      "grad_norm": 4.114370346069336,
      "kl": 0.2421875,
      "learning_rate": 8.933054393305439e-07,
      "loss": 0.0999,
      "reward": 0.6664131879806519,
      "reward_std": 0.12577899172902107,
      "reward_std_all": 0.13077115640044212,
      "rewards-max/forensics_sinkhorn_reward": 0.9212903082370758,
      "rewards-max/forensics_strict_edge_reward": 0.0634162575006485,
      "rewards-max/forensics_tiou_reward": 0.8411279022693634,
      "rewards/forensics_sinkhorn_reward": 0.5629808306694031,
      "rewards/forensics_strict_edge_reward": 0.0031101814820431173,
      "rewards/forensics_tiou_reward": 0.2563672214746475,
      "step": 103
    },
    {
      "completion_length": 22.515625,
      "epoch": 0.4351464435146444,
      "grad_norm": 4.008955478668213,
      "kl": 0.2939453125,
      "learning_rate": 8.922594142259414e-07,
      "loss": 0.0331,
      "reward": 0.6811788380146027,
      "reward_std": 0.1343187540769577,
      "reward_std_all": 0.1294378712773323,
      "rewards-max/forensics_sinkhorn_reward": 0.9851747453212738,
      "rewards-max/forensics_strict_edge_reward": 0.39885322004556656,
      "rewards-max/forensics_tiou_reward": 0.9893723428249359,
      "rewards/forensics_sinkhorn_reward": 0.5859697461128235,
      "rewards/forensics_strict_edge_reward": 0.03554975590668619,
      "rewards/forensics_tiou_reward": 0.3693679869174957,
      "step": 104
    },
    {
      "completion_length": 20.53125,
      "epoch": 0.4393305439330544,
      "grad_norm": 4.972544193267822,
      "kl": 0.32177734375,
      "learning_rate": 8.91213389121339e-07,
      "loss": 0.1977,
      "reward": 0.7196989357471466,
      "reward_std": 0.054869987070560455,
      "reward_std_all": 0.06923086196184158,
      "rewards-max/forensics_sinkhorn_reward": 0.9037068486213684,
      "rewards-max/forensics_strict_edge_reward": 0.4257437437772751,
      "rewards-max/forensics_tiou_reward": 0.7831189036369324,
      "rewards/forensics_sinkhorn_reward": 0.6503804922103882,
      "rewards/forensics_strict_edge_reward": 0.023370335809886456,
      "rewards/forensics_tiou_reward": 0.3443371057510376,
      "step": 105
    },
    {
      "completion_length": 22.765625,
      "epoch": 0.4435146443514644,
      "grad_norm": 5.118119239807129,
      "kl": 0.23828125,
      "learning_rate": 8.901673640167363e-07,
      "loss": 0.1633,
      "reward": 0.7351769506931305,
      "reward_std": 0.08987073600292206,
      "reward_std_all": 0.08389214053750038,
      "rewards-max/forensics_sinkhorn_reward": 0.9267374873161316,
      "rewards-max/forensics_strict_edge_reward": 0.10686714667826891,
      "rewards-max/forensics_tiou_reward": 0.8516325354576111,
      "rewards/forensics_sinkhorn_reward": 0.6485709846019745,
      "rewards/forensics_strict_edge_reward": 0.008433541981503367,
      "rewards/forensics_tiou_reward": 0.4063778668642044,
      "step": 106
    },
    {
      "completion_length": 21.90625,
      "epoch": 0.4476987447698745,
      "grad_norm": 4.420165538787842,
      "kl": 0.31005859375,
      "learning_rate": 8.891213389121338e-07,
      "loss": 0.1989,
      "reward": 0.7292194366455078,
      "reward_std": 0.07608578726649284,
      "reward_std_all": 0.07647116482257843,
      "rewards-max/forensics_sinkhorn_reward": 0.9561680257320404,
      "rewards-max/forensics_strict_edge_reward": 0.10245247930288315,
      "rewards-max/forensics_tiou_reward": 0.8960353136062622,
      "rewards/forensics_sinkhorn_reward": 0.64542555809021,
      "rewards/forensics_strict_edge_reward": 0.004532908904366195,
      "rewards/forensics_tiou_reward": 0.3821933716535568,
      "step": 107
    },
    {
      "completion_length": 21.84375,
      "epoch": 0.45188284518828453,
      "grad_norm": 4.747746467590332,
      "kl": 0.2802734375,
      "learning_rate": 8.880753138075313e-07,
      "loss": 0.0671,
      "reward": 0.7047756016254425,
      "reward_std": 0.0888243280351162,
      "reward_std_all": 0.09431849047541618,
      "rewards-max/forensics_sinkhorn_reward": 0.9634489119052887,
      "rewards-max/forensics_strict_edge_reward": 0.14756814017891884,
      "rewards-max/forensics_tiou_reward": 0.927169144153595,
      "rewards/forensics_sinkhorn_reward": 0.6106234788894653,
      "rewards/forensics_strict_edge_reward": 0.0067035488318651915,
      "rewards/forensics_tiou_reward": 0.318075031042099,
      "step": 108
    },
    {
      "completion_length": 20.6875,
      "epoch": 0.4560669456066946,
      "grad_norm": 4.729983329772949,
      "kl": 0.322265625,
      "learning_rate": 8.870292887029288e-07,
      "loss": 0.2044,
      "reward": 0.6959962248802185,
      "reward_std": 0.08873014152050018,
      "reward_std_all": 0.09577491134405136,
      "rewards-max/forensics_sinkhorn_reward": 0.9669191539287567,
      "rewards-max/forensics_strict_edge_reward": 0.45875632762908936,
      "rewards-max/forensics_tiou_reward": 0.9460777640342712,
      "rewards/forensics_sinkhorn_reward": 0.6161276698112488,
      "rewards/forensics_strict_edge_reward": 0.025388449430465698,
      "rewards/forensics_tiou_reward": 0.29743921756744385,
      "step": 109
    },
    {
      "completion_length": 20.171875,
      "epoch": 0.4602510460251046,
      "grad_norm": 4.467528820037842,
      "kl": 0.28564453125,
      "learning_rate": 8.859832635983263e-07,
      "loss": -0.0557,
      "reward": 0.7113789319992065,
      "reward_std": 0.06562216207385063,
      "reward_std_all": 0.07841750979423523,
      "rewards-max/forensics_sinkhorn_reward": 0.9802564978599548,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9760205447673798,
      "rewards/forensics_sinkhorn_reward": 0.6345897912979126,
      "rewards/forensics_strict_edge_reward": 0.0507843978703022,
      "rewards/forensics_tiou_reward": 0.36506760120391846,
      "step": 110
    },
    {
      "completion_length": 22.34375,
      "epoch": 0.46443514644351463,
      "grad_norm": 3.9726908206939697,
      "kl": 0.298828125,
      "learning_rate": 8.849372384937238e-07,
      "loss": 0.1473,
      "reward": 0.6692143380641937,
      "reward_std": 0.10413715615868568,
      "reward_std_all": 0.12293605133891106,
      "rewards-max/forensics_sinkhorn_reward": 0.9603296518325806,
      "rewards-max/forensics_strict_edge_reward": 0.27923316322267056,
      "rewards-max/forensics_tiou_reward": 0.9213449954986572,
      "rewards/forensics_sinkhorn_reward": 0.5782937109470367,
      "rewards/forensics_strict_edge_reward": 0.013259541592560709,
      "rewards/forensics_tiou_reward": 0.2950962334871292,
      "step": 111
    },
    {
      "completion_length": 20.515625,
      "epoch": 0.4686192468619247,
      "grad_norm": 5.792511463165283,
      "kl": 0.232421875,
      "learning_rate": 8.838912133891213e-07,
      "loss": 0.0222,
      "reward": 0.6711887419223785,
      "reward_std": 0.09323765337467194,
      "reward_std_all": 0.10896088927984238,
      "rewards-max/forensics_sinkhorn_reward": 0.8308920562267303,
      "rewards-max/forensics_strict_edge_reward": 0.0061388820897008145,
      "rewards-max/forensics_tiou_reward": 0.581878662109375,
      "rewards/forensics_sinkhorn_reward": 0.5500262975692749,
      "rewards/forensics_strict_edge_reward": 0.0002764794208154342,
      "rewards/forensics_tiou_reward": 0.20449750870466232,
      "step": 112
    },
    {
      "completion_length": 20.9375,
      "epoch": 0.47280334728033474,
      "grad_norm": 4.270845890045166,
      "kl": 0.2451171875,
      "learning_rate": 8.828451882845189e-07,
      "loss": 0.1043,
      "reward": 0.7480307817459106,
      "reward_std": 0.058831362053751945,
      "reward_std_all": 0.06112884357571602,
      "rewards-max/forensics_sinkhorn_reward": 0.9306323528289795,
      "rewards-max/forensics_strict_edge_reward": 0.17708471417427063,
      "rewards-max/forensics_tiou_reward": 0.8181626796722412,
      "rewards/forensics_sinkhorn_reward": 0.6807595193386078,
      "rewards/forensics_strict_edge_reward": 0.00822264002636075,
      "rewards/forensics_tiou_reward": 0.36745980381965637,
      "step": 113
    },
    {
      "completion_length": 20.59375,
      "epoch": 0.4769874476987448,
      "grad_norm": 4.835807800292969,
      "kl": 0.2724609375,
      "learning_rate": 8.817991631799164e-07,
      "loss": 0.1577,
      "reward": 0.6635827720165253,
      "reward_std": 0.12018009275197983,
      "reward_std_all": 0.125595360994339,
      "rewards-max/forensics_sinkhorn_reward": 0.9109250009059906,
      "rewards-max/forensics_strict_edge_reward": 0.014832108397968113,
      "rewards-max/forensics_tiou_reward": 0.8078579306602478,
      "rewards/forensics_sinkhorn_reward": 0.5380076467990875,
      "rewards/forensics_strict_edge_reward": 0.0007326194099732675,
      "rewards/forensics_tiou_reward": 0.2228105515241623,
      "step": 114
    },
    {
      "completion_length": 21.25,
      "epoch": 0.4811715481171548,
      "grad_norm": 4.83760929107666,
      "kl": 0.26171875,
      "learning_rate": 8.807531380753138e-07,
      "loss": 0.0524,
      "reward": 0.6196267604827881,
      "reward_std": 0.12467771768569946,
      "reward_std_all": 0.1487387977540493,
      "rewards-max/forensics_sinkhorn_reward": 0.8969154953956604,
      "rewards-max/forensics_strict_edge_reward": 0.007550061913207173,
      "rewards-max/forensics_tiou_reward": 0.7541579306125641,
      "rewards/forensics_sinkhorn_reward": 0.5127525180578232,
      "rewards/forensics_strict_edge_reward": 0.0003390344645595178,
      "rewards/forensics_tiou_reward": 0.2222927138209343,
      "step": 115
    },
    {
      "completion_length": 20.984375,
      "epoch": 0.48535564853556484,
      "grad_norm": 4.627970218658447,
      "kl": 0.33984375,
      "learning_rate": 8.797071129707112e-07,
      "loss": 0.2039,
      "reward": 0.6407594680786133,
      "reward_std": 0.15571652352809906,
      "reward_std_all": 0.1613377183675766,
      "rewards-max/forensics_sinkhorn_reward": 0.9541464447975159,
      "rewards-max/forensics_strict_edge_reward": 0.4128478467464447,
      "rewards-max/forensics_tiou_reward": 0.9113606810569763,
      "rewards/forensics_sinkhorn_reward": 0.5600471496582031,
      "rewards/forensics_strict_edge_reward": 0.023926532827317715,
      "rewards/forensics_tiou_reward": 0.2968806177377701,
      "step": 116
    },
    {
      "completion_length": 20.125,
      "epoch": 0.4895397489539749,
      "grad_norm": 4.301582336425781,
      "kl": 0.25537109375,
      "learning_rate": 8.786610878661087e-07,
      "loss": 0.2186,
      "reward": 0.7156794369220734,
      "reward_std": 0.08133885264396667,
      "reward_std_all": 0.08708785846829414,
      "rewards-max/forensics_sinkhorn_reward": 0.9541366696357727,
      "rewards-max/forensics_strict_edge_reward": 0.27649304270744324,
      "rewards-max/forensics_tiou_reward": 0.9151773750782013,
      "rewards/forensics_sinkhorn_reward": 0.6348585486412048,
      "rewards/forensics_strict_edge_reward": 0.015072049340233207,
      "rewards/forensics_tiou_reward": 0.34108440577983856,
      "step": 117
    },
    {
      "completion_length": 22.28125,
      "epoch": 0.49372384937238495,
      "grad_norm": 4.0744547843933105,
      "kl": 0.24658203125,
      "learning_rate": 8.776150627615062e-07,
      "loss": 0.1182,
      "reward": 0.6943563520908356,
      "reward_std": 0.052454475313425064,
      "reward_std_all": 0.08478982001543045,
      "rewards-max/forensics_sinkhorn_reward": 0.8888270556926727,
      "rewards-max/forensics_strict_edge_reward": 0.04391614522683085,
      "rewards-max/forensics_tiou_reward": 0.760515034198761,
      "rewards/forensics_sinkhorn_reward": 0.5842094123363495,
      "rewards/forensics_strict_edge_reward": 0.0018394534981780453,
      "rewards/forensics_tiou_reward": 0.29148536920547485,
      "step": 118
    },
    {
      "completion_length": 21.390625,
      "epoch": 0.497907949790795,
      "grad_norm": 4.33146858215332,
      "kl": 0.27734375,
      "learning_rate": 8.765690376569037e-07,
      "loss": 0.13,
      "reward": 0.6922522187232971,
      "reward_std": 0.09529590234160423,
      "reward_std_all": 0.10404876247048378,
      "rewards-max/forensics_sinkhorn_reward": 0.9648212492465973,
      "rewards-max/forensics_strict_edge_reward": 0.2340463027358055,
      "rewards-max/forensics_tiou_reward": 0.9284077882766724,
      "rewards/forensics_sinkhorn_reward": 0.6118902564048767,
      "rewards/forensics_strict_edge_reward": 0.00995484774466604,
      "rewards/forensics_tiou_reward": 0.3282892554998398,
      "step": 119
    },
    {
      "completion_length": 20.625,
      "epoch": 0.502092050209205,
      "grad_norm": 4.22263765335083,
      "kl": 0.3251953125,
      "learning_rate": 8.755230125523012e-07,
      "loss": 0.0812,
      "reward": 0.7416580021381378,
      "reward_std": 0.07423117384314537,
      "reward_std_all": 0.07055908814072609,
      "rewards-max/forensics_sinkhorn_reward": 0.973112165927887,
      "rewards-max/forensics_strict_edge_reward": 0.3095986396074295,
      "rewards-max/forensics_tiou_reward": 0.9546601474285126,
      "rewards/forensics_sinkhorn_reward": 0.6792832911014557,
      "rewards/forensics_strict_edge_reward": 0.018581854179501534,
      "rewards/forensics_tiou_reward": 0.41490085422992706,
      "step": 120
    },
    {
      "completion_length": 21.265625,
      "epoch": 0.5062761506276151,
      "grad_norm": 4.422859191894531,
      "kl": 0.2265625,
      "learning_rate": 8.744769874476987e-07,
      "loss": 0.0653,
      "reward": 0.7212788164615631,
      "reward_std": 0.08017028123140335,
      "reward_std_all": 0.08131022565066814,
      "rewards-max/forensics_sinkhorn_reward": 0.928358644247055,
      "rewards-max/forensics_strict_edge_reward": 0.04831257346086204,
      "rewards-max/forensics_tiou_reward": 0.833413302898407,
      "rewards/forensics_sinkhorn_reward": 0.6373350024223328,
      "rewards/forensics_strict_edge_reward": 0.0021762570104328915,
      "rewards/forensics_tiou_reward": 0.38504573702812195,
      "step": 121
    },
    {
      "completion_length": 20.96875,
      "epoch": 0.5104602510460251,
      "grad_norm": 6.072476387023926,
      "kl": 0.4091796875,
      "learning_rate": 8.734309623430963e-07,
      "loss": 0.1676,
      "reward": 0.669939398765564,
      "reward_std": 0.07033338584005833,
      "reward_std_all": 0.102961964905262,
      "rewards-max/forensics_sinkhorn_reward": 0.7987843453884125,
      "rewards-max/forensics_strict_edge_reward": 0.002074034244287759,
      "rewards-max/forensics_tiou_reward": 0.569980651140213,
      "rewards/forensics_sinkhorn_reward": 0.5440293550491333,
      "rewards/forensics_strict_edge_reward": 0.0001209177280543372,
      "rewards/forensics_tiou_reward": 0.19624386727809906,
      "step": 122
    },
    {
      "completion_length": 20.609375,
      "epoch": 0.5146443514644351,
      "grad_norm": 9.317931175231934,
      "kl": 0.201171875,
      "learning_rate": 8.723849372384937e-07,
      "loss": 0.1005,
      "reward": 0.676047682762146,
      "reward_std": 0.12901409715414047,
      "reward_std_all": 0.13090217858552933,
      "rewards-max/forensics_sinkhorn_reward": 0.9287828505039215,
      "rewards-max/forensics_strict_edge_reward": 0.1020270164590329,
      "rewards-max/forensics_tiou_reward": 0.8105596005916595,
      "rewards/forensics_sinkhorn_reward": 0.6000945866107941,
      "rewards/forensics_strict_edge_reward": 0.0042717092437669635,
      "rewards/forensics_tiou_reward": 0.27503231167793274,
      "step": 123
    },
    {
      "completion_length": 19.8125,
      "epoch": 0.5188284518828452,
      "grad_norm": 5.541645526885986,
      "kl": 0.18603515625,
      "learning_rate": 8.713389121338912e-07,
      "loss": 0.057,
      "reward": 0.717061847448349,
      "reward_std": 0.1092708483338356,
      "reward_std_all": 0.10273203998804092,
      "rewards-max/forensics_sinkhorn_reward": 0.9460267424583435,
      "rewards-max/forensics_strict_edge_reward": 0.11795143596827984,
      "rewards-max/forensics_tiou_reward": 0.8846004009246826,
      "rewards/forensics_sinkhorn_reward": 0.6472710072994232,
      "rewards/forensics_strict_edge_reward": 0.007469837786629796,
      "rewards/forensics_tiou_reward": 0.3118562251329422,
      "step": 124
    },
    {
      "completion_length": 20.234375,
      "epoch": 0.5230125523012552,
      "grad_norm": 4.772406101226807,
      "kl": 0.322265625,
      "learning_rate": 8.702928870292887e-07,
      "loss": 0.1847,
      "reward": 0.667356550693512,
      "reward_std": 0.13172370195388794,
      "reward_std_all": 0.13255062326788902,
      "rewards-max/forensics_sinkhorn_reward": 0.9370623528957367,
      "rewards-max/forensics_strict_edge_reward": 0.14881163672544062,
      "rewards-max/forensics_tiou_reward": 0.8513485789299011,
      "rewards/forensics_sinkhorn_reward": 0.6060092151165009,
      "rewards/forensics_strict_edge_reward": 0.0075325425423216075,
      "rewards/forensics_tiou_reward": 0.30809563398361206,
      "step": 125
    },
    {
      "completion_length": 21.59375,
      "epoch": 0.5271966527196653,
      "grad_norm": 4.211160659790039,
      "kl": 0.2705078125,
      "learning_rate": 8.692468619246861e-07,
      "loss": 0.1984,
      "reward": 0.7231673896312714,
      "reward_std": 0.0935569703578949,
      "reward_std_all": 0.0893617644906044,
      "rewards-max/forensics_sinkhorn_reward": 0.9165208339691162,
      "rewards-max/forensics_strict_edge_reward": 0.101654551923275,
      "rewards-max/forensics_tiou_reward": 0.8080623149871826,
      "rewards/forensics_sinkhorn_reward": 0.6441143453121185,
      "rewards/forensics_strict_edge_reward": 0.006359259597957134,
      "rewards/forensics_tiou_reward": 0.34275463223457336,
      "step": 126
    },
    {
      "completion_length": 19.265625,
      "epoch": 0.5313807531380753,
      "grad_norm": 4.909320831298828,
      "kl": 0.20654296875,
      "learning_rate": 8.682008368200836e-07,
      "loss": 0.0744,
      "reward": 0.6852156817913055,
      "reward_std": 0.06167054455727339,
      "reward_std_all": 0.09089447185397148,
      "rewards-max/forensics_sinkhorn_reward": 0.9488076865673065,
      "rewards-max/forensics_strict_edge_reward": 0.09584501478821039,
      "rewards-max/forensics_tiou_reward": 0.9027408361434937,
      "rewards/forensics_sinkhorn_reward": 0.6369556188583374,
      "rewards/forensics_strict_edge_reward": 0.004700066230725497,
      "rewards/forensics_tiou_reward": 0.3426629602909088,
      "step": 127
    },
    {
      "completion_length": 20.234375,
      "epoch": 0.5355648535564853,
      "grad_norm": 5.050086975097656,
      "kl": 0.279296875,
      "learning_rate": 8.671548117154811e-07,
      "loss": 0.2597,
      "reward": 0.7109875082969666,
      "reward_std": 0.09941747784614563,
      "reward_std_all": 0.09727827832102776,
      "rewards-max/forensics_sinkhorn_reward": 0.9783967137336731,
      "rewards-max/forensics_strict_edge_reward": 0.4378620684146881,
      "rewards-max/forensics_tiou_reward": 0.9666975736618042,
      "rewards/forensics_sinkhorn_reward": 0.6553693413734436,
      "rewards/forensics_strict_edge_reward": 0.019288774114102125,
      "rewards/forensics_tiou_reward": 0.3793889731168747,
      "step": 128
    },
    {
      "completion_length": 19.921875,
      "epoch": 0.5397489539748954,
      "grad_norm": 4.329197406768799,
      "kl": 0.28515625,
      "learning_rate": 8.661087866108786e-07,
      "loss": 0.1053,
      "reward": 0.7072948813438416,
      "reward_std": 0.07163585536181927,
      "reward_std_all": 0.0819661058485508,
      "rewards-max/forensics_sinkhorn_reward": 0.9631127715110779,
      "rewards-max/forensics_strict_edge_reward": 0.3720272370846942,
      "rewards-max/forensics_tiou_reward": 0.9287299811840057,
      "rewards/forensics_sinkhorn_reward": 0.6423455476760864,
      "rewards/forensics_strict_edge_reward": 0.024902131750422996,
      "rewards/forensics_tiou_reward": 0.3562691807746887,
      "step": 129
    },
    {
      "completion_length": 20.265625,
      "epoch": 0.5439330543933054,
      "grad_norm": 5.125804901123047,
      "kl": 0.25537109375,
      "learning_rate": 8.650627615062761e-07,
      "loss": 0.107,
      "reward": 0.7331575155258179,
      "reward_std": 0.07575348764657974,
      "reward_std_all": 0.07431180588901043,
      "rewards-max/forensics_sinkhorn_reward": 0.9718763828277588,
      "rewards-max/forensics_strict_edge_reward": 0.31773683428764343,
      "rewards-max/forensics_tiou_reward": 0.9518689513206482,
      "rewards/forensics_sinkhorn_reward": 0.6453386545181274,
      "rewards/forensics_strict_edge_reward": 0.0168149396777153,
      "rewards/forensics_tiou_reward": 0.3315916657447815,
      "step": 130
    },
    {
      "completion_length": 22.890625,
      "epoch": 0.5481171548117155,
      "grad_norm": 4.409186840057373,
      "kl": 0.259765625,
      "learning_rate": 8.640167364016737e-07,
      "loss": 0.139,
      "reward": 0.6333906948566437,
      "reward_std": 0.0921161063015461,
      "reward_std_all": 0.12697359919548035,
      "rewards-max/forensics_sinkhorn_reward": 0.9441240429878235,
      "rewards-max/forensics_strict_edge_reward": 0.26528801023960114,
      "rewards-max/forensics_tiou_reward": 0.849017471075058,
      "rewards/forensics_sinkhorn_reward": 0.5302350223064423,
      "rewards/forensics_strict_edge_reward": 0.011636818497208878,
      "rewards/forensics_tiou_reward": 0.22227338701486588,
      "step": 131
    },
    {
      "completion_length": 20.484375,
      "epoch": 0.5523012552301255,
      "grad_norm": 4.997668266296387,
      "kl": 0.341796875,
      "learning_rate": 8.629707112970711e-07,
      "loss": 0.1134,
      "reward": 0.7089752554893494,
      "reward_std": 0.08795610442757607,
      "reward_std_all": 0.09047738462686539,
      "rewards-max/forensics_sinkhorn_reward": 0.9575922191143036,
      "rewards-max/forensics_strict_edge_reward": 0.45036743208765984,
      "rewards-max/forensics_tiou_reward": 0.8967825770378113,
      "rewards/forensics_sinkhorn_reward": 0.6434885859489441,
      "rewards/forensics_strict_edge_reward": 0.031182002276182175,
      "rewards/forensics_tiou_reward": 0.3712092190980911,
      "step": 132
    },
    {
      "completion_length": 22.546875,
      "epoch": 0.5564853556485355,
      "grad_norm": 4.424825668334961,
      "kl": 0.255859375,
      "learning_rate": 8.619246861924686e-07,
      "loss": -0.0085,
      "reward": 0.692328006029129,
      "reward_std": 0.1287526786327362,
      "reward_std_all": 0.1218889132142067,
      "rewards-max/forensics_sinkhorn_reward": 0.9568300247192383,
      "rewards-max/forensics_strict_edge_reward": 0.19414592161774635,
      "rewards-max/forensics_tiou_reward": 0.9054490029811859,
      "rewards/forensics_sinkhorn_reward": 0.5924190282821655,
      "rewards/forensics_strict_edge_reward": 0.010455102194100618,
      "rewards/forensics_tiou_reward": 0.35871240496635437,
      "step": 133
    },
    {
      "completion_length": 21.640625,
      "epoch": 0.5606694560669456,
      "grad_norm": 3.9349043369293213,
      "kl": 0.27783203125,
      "learning_rate": 8.608786610878661e-07,
      "loss": 0.2198,
      "reward": 0.7334217429161072,
      "reward_std": 0.077472691424191,
      "reward_std_all": 0.07602263055741787,
      "rewards-max/forensics_sinkhorn_reward": 0.9560038447380066,
      "rewards-max/forensics_strict_edge_reward": 0.21768598072230816,
      "rewards-max/forensics_tiou_reward": 0.9006640911102295,
      "rewards/forensics_sinkhorn_reward": 0.6355761587619781,
      "rewards/forensics_strict_edge_reward": 0.010753444279544055,
      "rewards/forensics_tiou_reward": 0.3414278030395508,
      "step": 134
    },
    {
      "completion_length": 21.234375,
      "epoch": 0.5648535564853556,
      "grad_norm": 4.912817478179932,
      "kl": 0.25,
      "learning_rate": 8.598326359832636e-07,
      "loss": 0.0013,
      "reward": 0.7011628746986389,
      "reward_std": 0.08744438737630844,
      "reward_std_all": 0.09282811731100082,
      "rewards-max/forensics_sinkhorn_reward": 0.9756866693496704,
      "rewards-max/forensics_strict_edge_reward": 0.4395187571644783,
      "rewards-max/forensics_tiou_reward": 0.9671949148178101,
      "rewards/forensics_sinkhorn_reward": 0.6074908375740051,
      "rewards/forensics_strict_edge_reward": 0.026553955860435963,
      "rewards/forensics_tiou_reward": 0.3101697564125061,
      "step": 135
    },
    {
      "completion_length": 20.890625,
      "epoch": 0.5690376569037657,
      "grad_norm": 4.3465657234191895,
      "kl": 0.2587890625,
      "learning_rate": 8.58786610878661e-07,
      "loss": 0.1033,
      "reward": 0.6901870667934418,
      "reward_std": 0.09598682075738907,
      "reward_std_all": 0.10200481489300728,
      "rewards-max/forensics_sinkhorn_reward": 0.9332247972488403,
      "rewards-max/forensics_strict_edge_reward": 0.24980023130774498,
      "rewards-max/forensics_tiou_reward": 0.8612598180770874,
      "rewards/forensics_sinkhorn_reward": 0.603622704744339,
      "rewards/forensics_strict_edge_reward": 0.014540011994540691,
      "rewards/forensics_tiou_reward": 0.3499062657356262,
      "step": 136
    },
    {
      "completion_length": 20.5625,
      "epoch": 0.5732217573221757,
      "grad_norm": 3.400831699371338,
      "kl": 0.23193359375,
      "learning_rate": 8.577405857740585e-07,
      "loss": 0.1571,
      "reward": 0.7293992340564728,
      "reward_std": 0.08707755990326405,
      "reward_std_all": 0.08434986136853695,
      "rewards-max/forensics_sinkhorn_reward": 0.9841373264789581,
      "rewards-max/forensics_strict_edge_reward": 0.6901743710041046,
      "rewards-max/forensics_tiou_reward": 0.9821756482124329,
      "rewards/forensics_sinkhorn_reward": 0.6702795028686523,
      "rewards/forensics_strict_edge_reward": 0.03609178029000759,
      "rewards/forensics_tiou_reward": 0.41693611443042755,
      "step": 137
    },
    {
      "completion_length": 21.875,
      "epoch": 0.5774058577405857,
      "grad_norm": 5.0664849281311035,
      "kl": 0.2685546875,
      "learning_rate": 8.56694560669456e-07,
      "loss": 0.0898,
      "reward": 0.7271926403045654,
      "reward_std": 0.08147059567272663,
      "reward_std_all": 0.08112307265400887,
      "rewards-max/forensics_sinkhorn_reward": 0.9443883001804352,
      "rewards-max/forensics_strict_edge_reward": 0.1345172133296728,
      "rewards-max/forensics_tiou_reward": 0.8838167488574982,
      "rewards/forensics_sinkhorn_reward": 0.6563872992992401,
      "rewards/forensics_strict_edge_reward": 0.010154121788218617,
      "rewards/forensics_tiou_reward": 0.41963501274585724,
      "step": 138
    },
    {
      "completion_length": 19.8125,
      "epoch": 0.5815899581589958,
      "grad_norm": 4.825539588928223,
      "kl": 0.24072265625,
      "learning_rate": 8.556485355648536e-07,
      "loss": 0.0852,
      "reward": 0.7020849883556366,
      "reward_std": 0.05490955337882042,
      "reward_std_all": 0.07802115194499493,
      "rewards-max/forensics_sinkhorn_reward": 0.965630441904068,
      "rewards-max/forensics_strict_edge_reward": 0.31668074429035187,
      "rewards-max/forensics_tiou_reward": 0.93083855509758,
      "rewards/forensics_sinkhorn_reward": 0.6470659971237183,
      "rewards/forensics_strict_edge_reward": 0.017911420203745365,
      "rewards/forensics_tiou_reward": 0.37023212015628815,
      "step": 139
    },
    {
      "completion_length": 21.84375,
      "epoch": 0.5857740585774058,
      "grad_norm": 4.664609432220459,
      "kl": 0.23095703125,
      "learning_rate": 8.54602510460251e-07,
      "loss": 0.2055,
      "reward": 0.7150984704494476,
      "reward_std": 0.11240396648645401,
      "reward_std_all": 0.10472878068685532,
      "rewards-max/forensics_sinkhorn_reward": 0.9727418720722198,
      "rewards-max/forensics_strict_edge_reward": 0.43434032052755356,
      "rewards-max/forensics_tiou_reward": 0.9512523412704468,
      "rewards/forensics_sinkhorn_reward": 0.6328481137752533,
      "rewards/forensics_strict_edge_reward": 0.02523111505433917,
      "rewards/forensics_tiou_reward": 0.3943115770816803,
      "step": 140
    },
    {
      "completion_length": 22.6875,
      "epoch": 0.5899581589958159,
      "grad_norm": 4.8877739906311035,
      "kl": 0.30859375,
      "learning_rate": 8.535564853556485e-07,
      "loss": 0.1746,
      "reward": 0.7312796115875244,
      "reward_std": 0.08772524073719978,
      "reward_std_all": 0.08303219079971313,
      "rewards-max/forensics_sinkhorn_reward": 0.9699095487594604,
      "rewards-max/forensics_strict_edge_reward": 0.45305854827165604,
      "rewards-max/forensics_tiou_reward": 0.9455200433731079,
      "rewards/forensics_sinkhorn_reward": 0.6890096664428711,
      "rewards/forensics_strict_edge_reward": 0.028609307017177343,
      "rewards/forensics_tiou_reward": 0.470380961894989,
      "step": 141
    },
    {
      "completion_length": 20.53125,
      "epoch": 0.5941422594142259,
      "grad_norm": 4.47845983505249,
      "kl": 0.2109375,
      "learning_rate": 8.52510460251046e-07,
      "loss": 0.2832,
      "reward": 0.6910119652748108,
      "reward_std": 0.07627564668655396,
      "reward_std_all": 0.09610321372747421,
      "rewards-max/forensics_sinkhorn_reward": 0.9345208704471588,
      "rewards-max/forensics_strict_edge_reward": 0.13762888312339783,
      "rewards-max/forensics_tiou_reward": 0.8500016331672668,
      "rewards/forensics_sinkhorn_reward": 0.6108808219432831,
      "rewards/forensics_strict_edge_reward": 0.011174785904586315,
      "rewards/forensics_tiou_reward": 0.3175252676010132,
      "step": 142
    },
    {
      "completion_length": 23.625,
      "epoch": 0.5983263598326359,
      "grad_norm": 4.8956828117370605,
      "kl": 0.28076171875,
      "learning_rate": 8.514644351464435e-07,
      "loss": 0.1487,
      "reward": 0.6931778788566589,
      "reward_std": 0.08331413939595222,
      "reward_std_all": 0.09189091622829437,
      "rewards-max/forensics_sinkhorn_reward": 0.945340484380722,
      "rewards-max/forensics_strict_edge_reward": 0.0781316477805376,
      "rewards-max/forensics_tiou_reward": 0.8787558972835541,
      "rewards/forensics_sinkhorn_reward": 0.6030635237693787,
      "rewards/forensics_strict_edge_reward": 0.0038150690379552543,
      "rewards/forensics_tiou_reward": 0.34900061786174774,
      "step": 143
    },
    {
      "completion_length": 20.875,
      "epoch": 0.602510460251046,
      "grad_norm": 5.912429332733154,
      "kl": 0.27685546875,
      "learning_rate": 8.50418410041841e-07,
      "loss": 0.1431,
      "reward": 0.7012292444705963,
      "reward_std": 0.05751827731728554,
      "reward_std_all": 0.07842257618904114,
      "rewards-max/forensics_sinkhorn_reward": 0.9358232915401459,
      "rewards-max/forensics_strict_edge_reward": 0.1240017200470902,
      "rewards-max/forensics_tiou_reward": 0.8457496166229248,
      "rewards/forensics_sinkhorn_reward": 0.6127460896968842,
      "rewards/forensics_strict_edge_reward": 0.007275498806848191,
      "rewards/forensics_tiou_reward": 0.31936097145080566,
      "step": 144
    },
    {
      "completion_length": 24.34375,
      "epoch": 0.606694560669456,
      "grad_norm": 7.047746181488037,
      "kl": 0.27734375,
      "learning_rate": 8.493723849372385e-07,
      "loss": 0.0894,
      "reward": 0.6675803363323212,
      "reward_std": 0.11002100259065628,
      "reward_std_all": 0.11824183911085129,
      "rewards-max/forensics_sinkhorn_reward": 0.9288161993026733,
      "rewards-max/forensics_strict_edge_reward": 0.10763757675886154,
      "rewards-max/forensics_tiou_reward": 0.8536576330661774,
      "rewards/forensics_sinkhorn_reward": 0.5547248721122742,
      "rewards/forensics_strict_edge_reward": 0.005097092245705426,
      "rewards/forensics_tiou_reward": 0.2577173039317131,
      "step": 145
    },
    {
      "completion_length": 21.8125,
      "epoch": 0.6108786610878661,
      "grad_norm": 5.953724384307861,
      "kl": 0.22509765625,
      "learning_rate": 8.483263598326359e-07,
      "loss": 0.0964,
      "reward": 0.6804570853710175,
      "reward_std": 0.14547257870435715,
      "reward_std_all": 0.13878250122070312,
      "rewards-max/forensics_sinkhorn_reward": 0.9183357656002045,
      "rewards-max/forensics_strict_edge_reward": 0.15444660326465964,
      "rewards-max/forensics_tiou_reward": 0.8218707740306854,
      "rewards/forensics_sinkhorn_reward": 0.608658492565155,
      "rewards/forensics_strict_edge_reward": 0.006878252694150433,
      "rewards/forensics_tiou_reward": 0.3420115113258362,
      "step": 146
    },
    {
      "completion_length": 21.921875,
      "epoch": 0.6150627615062761,
      "grad_norm": 4.796083450317383,
      "kl": 0.23486328125,
      "learning_rate": 8.472803347280334e-07,
      "loss": 0.0415,
      "reward": 0.7246265411376953,
      "reward_std": 0.0835752859711647,
      "reward_std_all": 0.08642441406846046,
      "rewards-max/forensics_sinkhorn_reward": 0.9375033378601074,
      "rewards-max/forensics_strict_edge_reward": 0.08560299407690763,
      "rewards-max/forensics_tiou_reward": 0.8699356615543365,
      "rewards/forensics_sinkhorn_reward": 0.65392005443573,
      "rewards/forensics_strict_edge_reward": 0.00541738624451682,
      "rewards/forensics_tiou_reward": 0.4072123169898987,
      "step": 147
    },
    {
      "completion_length": 21.09375,
      "epoch": 0.6192468619246861,
      "grad_norm": 4.8301873207092285,
      "kl": 0.2763671875,
      "learning_rate": 8.462343096234308e-07,
      "loss": 0.1024,
      "reward": 0.7007364928722382,
      "reward_std": 0.08489787019789219,
      "reward_std_all": 0.09696268662810326,
      "rewards-max/forensics_sinkhorn_reward": 0.9735887348651886,
      "rewards-max/forensics_strict_edge_reward": 0.6703200340270996,
      "rewards-max/forensics_tiou_reward": 0.9564283490180969,
      "rewards/forensics_sinkhorn_reward": 0.6180476546287537,
      "rewards/forensics_strict_edge_reward": 0.04327636770904064,
      "rewards/forensics_tiou_reward": 0.35553619265556335,
      "step": 148
    },
    {
      "completion_length": 23.4375,
      "epoch": 0.6234309623430963,
      "grad_norm": 5.278594493865967,
      "kl": 0.2509765625,
      "learning_rate": 8.451882845188284e-07,
      "loss": 0.0837,
      "reward": 0.7130126357078552,
      "reward_std": 0.06525919586420059,
      "reward_std_all": 0.07747462391853333,
      "rewards-max/forensics_sinkhorn_reward": 0.882945328950882,
      "rewards-max/forensics_strict_edge_reward": 0.009598683565855026,
      "rewards-max/forensics_tiou_reward": 0.721752256155014,
      "rewards/forensics_sinkhorn_reward": 0.6209991574287415,
      "rewards/forensics_strict_edge_reward": 0.0007720786525169387,
      "rewards/forensics_tiou_reward": 0.3231232166290283,
      "step": 149
    },
    {
      "completion_length": 21.3125,
      "epoch": 0.6276150627615062,
      "grad_norm": 5.1915998458862305,
      "kl": 0.27734375,
      "learning_rate": 8.441422594142259e-07,
      "loss": 0.0926,
      "reward": 0.6705094873905182,
      "reward_std": 0.11100856214761734,
      "reward_std_all": 0.1207636184990406,
      "rewards-max/forensics_sinkhorn_reward": 0.9435794055461884,
      "rewards-max/forensics_strict_edge_reward": 0.2708844020962715,
      "rewards-max/forensics_tiou_reward": 0.8953998982906342,
      "rewards/forensics_sinkhorn_reward": 0.5823807418346405,
      "rewards/forensics_strict_edge_reward": 0.013385127298533916,
      "rewards/forensics_tiou_reward": 0.27768757194280624,
      "step": 150
    },
    {
      "completion_length": 20.96875,
      "epoch": 0.6317991631799164,
      "grad_norm": 4.085594654083252,
      "kl": 0.3583984375,
      "learning_rate": 8.430962343096234e-07,
      "loss": 0.0439,
      "reward": 0.7121001183986664,
      "reward_std": 0.09134243428707123,
      "reward_std_all": 0.0956825315952301,
      "rewards-max/forensics_sinkhorn_reward": 0.9695028364658356,
      "rewards-max/forensics_strict_edge_reward": 0.26206959038972855,
      "rewards-max/forensics_tiou_reward": 0.9532748758792877,
      "rewards/forensics_sinkhorn_reward": 0.6185591816902161,
      "rewards/forensics_strict_edge_reward": 0.014314424246549606,
      "rewards/forensics_tiou_reward": 0.3287987858057022,
      "step": 151
    },
    {
      "completion_length": 21.21875,
      "epoch": 0.6359832635983264,
      "grad_norm": 5.0939717292785645,
      "kl": 0.3154296875,
      "learning_rate": 8.420502092050209e-07,
      "loss": 0.0839,
      "reward": 0.6796178221702576,
      "reward_std": 0.13799624517560005,
      "reward_std_all": 0.1326201632618904,
      "rewards-max/forensics_sinkhorn_reward": 0.9748204350471497,
      "rewards-max/forensics_strict_edge_reward": 0.39161255955696106,
      "rewards-max/forensics_tiou_reward": 0.9583913683891296,
      "rewards/forensics_sinkhorn_reward": 0.6159272789955139,
      "rewards/forensics_strict_edge_reward": 0.01911817630752921,
      "rewards/forensics_tiou_reward": 0.3472883999347687,
      "step": 152
    },
    {
      "completion_length": 19.265625,
      "epoch": 0.6401673640167364,
      "grad_norm": 5.810225963592529,
      "kl": 0.3017578125,
      "learning_rate": 8.410041841004184e-07,
      "loss": 0.0275,
      "reward": 0.7295121550559998,
      "reward_std": 0.052511125802993774,
      "reward_std_all": 0.062425559386610985,
      "rewards-max/forensics_sinkhorn_reward": 0.9652379155158997,
      "rewards-max/forensics_strict_edge_reward": 0.34477679431438446,
      "rewards-max/forensics_tiou_reward": 0.9375792741775513,
      "rewards/forensics_sinkhorn_reward": 0.6818097829818726,
      "rewards/forensics_strict_edge_reward": 0.022911169566214085,
      "rewards/forensics_tiou_reward": 0.373970165848732,
      "step": 153
    },
    {
      "completion_length": 20.015625,
      "epoch": 0.6443514644351465,
      "grad_norm": 6.545444011688232,
      "kl": 0.2998046875,
      "learning_rate": 8.399581589958159e-07,
      "loss": 0.0364,
      "reward": 0.7091613113880157,
      "reward_std": 0.09258678928017616,
      "reward_std_all": 0.09342098981142044,
      "rewards-max/forensics_sinkhorn_reward": 0.9630666077136993,
      "rewards-max/forensics_strict_edge_reward": 0.37159113585948944,
      "rewards-max/forensics_tiou_reward": 0.9356544017791748,
      "rewards/forensics_sinkhorn_reward": 0.6383098661899567,
      "rewards/forensics_strict_edge_reward": 0.0173605028539896,
      "rewards/forensics_tiou_reward": 0.33516839146614075,
      "step": 154
    },
    {
      "completion_length": 19.71875,
      "epoch": 0.6485355648535565,
      "grad_norm": 6.260647296905518,
      "kl": 0.24755859375,
      "learning_rate": 8.389121338912134e-07,
      "loss": 0.0723,
      "reward": 0.7058964669704437,
      "reward_std": 0.06416642293334007,
      "reward_std_all": 0.0843755193054676,
      "rewards-max/forensics_sinkhorn_reward": 0.9833571016788483,
      "rewards-max/forensics_strict_edge_reward": 0.5738542377948761,
      "rewards-max/forensics_tiou_reward": 0.9836746156215668,
      "rewards/forensics_sinkhorn_reward": 0.6587307751178741,
      "rewards/forensics_strict_edge_reward": 0.041884804144501686,
      "rewards/forensics_tiou_reward": 0.36983494460582733,
      "step": 155
    },
    {
      "completion_length": 20.5,
      "epoch": 0.6527196652719666,
      "grad_norm": 5.041263103485107,
      "kl": 0.2138671875,
      "learning_rate": 8.378661087866107e-07,
      "loss": 0.1945,
      "reward": 0.6797813773155212,
      "reward_std": 0.08476816117763519,
      "reward_std_all": 0.1022484302520752,
      "rewards-max/forensics_sinkhorn_reward": 0.8979610800743103,
      "rewards-max/forensics_strict_edge_reward": 0.0399471910786815,
      "rewards-max/forensics_tiou_reward": 0.7694072723388672,
      "rewards/forensics_sinkhorn_reward": 0.602169543504715,
      "rewards/forensics_strict_edge_reward": 0.004809596379345749,
      "rewards/forensics_tiou_reward": 0.3340313583612442,
      "step": 156
    },
    {
      "completion_length": 20.8125,
      "epoch": 0.6569037656903766,
      "grad_norm": 5.261538028717041,
      "kl": 0.2666015625,
      "learning_rate": 8.368200836820083e-07,
      "loss": 0.0933,
      "reward": 0.6505643129348755,
      "reward_std": 0.0853204857558012,
      "reward_std_all": 0.11673573404550552,
      "rewards-max/forensics_sinkhorn_reward": 0.941114991903305,
      "rewards-max/forensics_strict_edge_reward": 0.2459249645471573,
      "rewards-max/forensics_tiou_reward": 0.8643920421600342,
      "rewards/forensics_sinkhorn_reward": 0.5642818212509155,
      "rewards/forensics_strict_edge_reward": 0.01500370493158698,
      "rewards/forensics_tiou_reward": 0.2688145190477371,
      "step": 157
    },
    {
      "completion_length": 19.6875,
      "epoch": 0.6610878661087866,
      "grad_norm": 5.810884952545166,
      "kl": 0.23974609375,
      "learning_rate": 8.357740585774058e-07,
      "loss": 0.1535,
      "reward": 0.7566971182823181,
      "reward_std": 0.04439771734178066,
      "reward_std_all": 0.046066926792263985,
      "rewards-max/forensics_sinkhorn_reward": 0.9048029184341431,
      "rewards-max/forensics_strict_edge_reward": 0.338442787528038,
      "rewards-max/forensics_tiou_reward": 0.7913305163383484,
      "rewards/forensics_sinkhorn_reward": 0.658940464258194,
      "rewards/forensics_strict_edge_reward": 0.018011488020420074,
      "rewards/forensics_tiou_reward": 0.3360998183488846,
      "step": 158
    },
    {
      "completion_length": 21.421875,
      "epoch": 0.6652719665271967,
      "grad_norm": 4.002952575683594,
      "kl": 0.2587890625,
      "learning_rate": 8.347280334728033e-07,
      "loss": 0.0141,
      "reward": 0.7306724786758423,
      "reward_std": 0.03020992036908865,
      "reward_std_all": 0.05273215286433697,
      "rewards-max/forensics_sinkhorn_reward": 0.9180255234241486,
      "rewards-max/forensics_strict_edge_reward": 0.24092985625611618,
      "rewards-max/forensics_tiou_reward": 0.8165039122104645,
      "rewards/forensics_sinkhorn_reward": 0.6432512402534485,
      "rewards/forensics_strict_edge_reward": 0.02177770652269828,
      "rewards/forensics_tiou_reward": 0.3412909656763077,
      "step": 159
    },
    {
      "completion_length": 20.90625,
      "epoch": 0.6694560669456067,
      "grad_norm": 5.706257343292236,
      "kl": 0.2822265625,
      "learning_rate": 8.336820083682008e-07,
      "loss": 0.2043,
      "reward": 0.6840424239635468,
      "reward_std": 0.0924098789691925,
      "reward_std_all": 0.10639629885554314,
      "rewards-max/forensics_sinkhorn_reward": 0.9344517588615417,
      "rewards-max/forensics_strict_edge_reward": 0.06643724255263805,
      "rewards-max/forensics_tiou_reward": 0.8731324672698975,
      "rewards/forensics_sinkhorn_reward": 0.5989737808704376,
      "rewards/forensics_strict_edge_reward": 0.004237269284203649,
      "rewards/forensics_tiou_reward": 0.33303147554397583,
      "step": 160
    },
    {
      "completion_length": 19.546875,
      "epoch": 0.6736401673640168,
      "grad_norm": 5.311330318450928,
      "kl": 0.287109375,
      "learning_rate": 8.326359832635983e-07,
      "loss": 0.0093,
      "reward": 0.6962194442749023,
      "reward_std": 0.054925503209233284,
      "reward_std_all": 0.08273976296186447,
      "rewards-max/forensics_sinkhorn_reward": 0.924916684627533,
      "rewards-max/forensics_strict_edge_reward": 0.13188468151929555,
      "rewards-max/forensics_tiou_reward": 0.8256035149097443,
      "rewards/forensics_sinkhorn_reward": 0.6091883480548859,
      "rewards/forensics_strict_edge_reward": 0.005640203457005555,
      "rewards/forensics_tiou_reward": 0.2699959725141525,
      "step": 161
    },
    {
      "completion_length": 20.8125,
      "epoch": 0.6778242677824268,
      "grad_norm": 5.717287063598633,
      "kl": 0.326171875,
      "learning_rate": 8.315899581589958e-07,
      "loss": 0.1234,
      "reward": 0.7400762736797333,
      "reward_std": 0.07287893258035183,
      "reward_std_all": 0.07050961256027222,
      "rewards-max/forensics_sinkhorn_reward": 0.9834147095680237,
      "rewards-max/forensics_strict_edge_reward": 0.5995397567749023,
      "rewards-max/forensics_tiou_reward": 0.9841173589229584,
      "rewards/forensics_sinkhorn_reward": 0.6699238121509552,
      "rewards/forensics_strict_edge_reward": 0.025075549259781837,
      "rewards/forensics_tiou_reward": 0.40637166798114777,
      "step": 162
    },
    {
      "completion_length": 20.46875,
      "epoch": 0.6820083682008368,
      "grad_norm": 6.320300579071045,
      "kl": 0.21728515625,
      "learning_rate": 8.305439330543933e-07,
      "loss": 0.0667,
      "reward": 0.7416859567165375,
      "reward_std": 0.06380464136600494,
      "reward_std_all": 0.0654451958835125,
      "rewards-max/forensics_sinkhorn_reward": 0.9511286914348602,
      "rewards-max/forensics_strict_edge_reward": 0.18410745647270232,
      "rewards-max/forensics_tiou_reward": 0.9078531563282013,
      "rewards/forensics_sinkhorn_reward": 0.6575965583324432,
      "rewards/forensics_strict_edge_reward": 0.009397937552421354,
      "rewards/forensics_tiou_reward": 0.3763948529958725,
      "step": 163
    },
    {
      "completion_length": 19.71875,
      "epoch": 0.6861924686192469,
      "grad_norm": 4.658205986022949,
      "kl": 0.318359375,
      "learning_rate": 8.294979079497908e-07,
      "loss": 0.0771,
      "reward": 0.712444394826889,
      "reward_std": 0.09949420019984245,
      "reward_std_all": 0.09627219289541245,
      "rewards-max/forensics_sinkhorn_reward": 0.9684043824672699,
      "rewards-max/forensics_strict_edge_reward": 0.41768769919872284,
      "rewards-max/forensics_tiou_reward": 0.9491016566753387,
      "rewards/forensics_sinkhorn_reward": 0.6457288861274719,
      "rewards/forensics_strict_edge_reward": 0.021559080109000206,
      "rewards/forensics_tiou_reward": 0.3464227318763733,
      "step": 164
    },
    {
      "completion_length": 21.28125,
      "epoch": 0.6903765690376569,
      "grad_norm": 4.253142833709717,
      "kl": 0.23388671875,
      "learning_rate": 8.284518828451882e-07,
      "loss": 0.1427,
      "reward": 0.7309743762016296,
      "reward_std": 0.08434360660612583,
      "reward_std_all": 0.08062344789505005,
      "rewards-max/forensics_sinkhorn_reward": 0.9790324568748474,
      "rewards-max/forensics_strict_edge_reward": 0.509157819673419,
      "rewards-max/forensics_tiou_reward": 0.9697898924350739,
      "rewards/forensics_sinkhorn_reward": 0.676325649023056,
      "rewards/forensics_strict_edge_reward": 0.02984002116136253,
      "rewards/forensics_tiou_reward": 0.3769441992044449,
      "step": 165
    },
    {
      "completion_length": 22.375,
      "epoch": 0.694560669456067,
      "grad_norm": 4.755173206329346,
      "kl": 0.26220703125,
      "learning_rate": 8.274058577405857e-07,
      "loss": 0.1168,
      "reward": 0.7124826610088348,
      "reward_std": 0.09989385306835175,
      "reward_std_all": 0.09759825840592384,
      "rewards-max/forensics_sinkhorn_reward": 0.9603721499443054,
      "rewards-max/forensics_strict_edge_reward": 0.29184726625680923,
      "rewards-max/forensics_tiou_reward": 0.9264403879642487,
      "rewards/forensics_sinkhorn_reward": 0.6225598752498627,
      "rewards/forensics_strict_edge_reward": 0.021495106630027294,
      "rewards/forensics_tiou_reward": 0.36380621790885925,
      "step": 166
    },
    {
      "completion_length": 20.1875,
      "epoch": 0.698744769874477,
      "grad_norm": 5.792556285858154,
      "kl": 0.24658203125,
      "learning_rate": 8.263598326359832e-07,
      "loss": 0.0752,
      "reward": 0.711310863494873,
      "reward_std": 0.07913228496909142,
      "reward_std_all": 0.0897756963968277,
      "rewards-max/forensics_sinkhorn_reward": 0.9588468968868256,
      "rewards-max/forensics_strict_edge_reward": 0.23616847395896912,
      "rewards-max/forensics_tiou_reward": 0.9231670200824738,
      "rewards/forensics_sinkhorn_reward": 0.652193546295166,
      "rewards/forensics_strict_edge_reward": 0.01515809865668416,
      "rewards/forensics_tiou_reward": 0.37445567548274994,
      "step": 167
    },
    {
      "completion_length": 19.46875,
      "epoch": 0.702928870292887,
      "grad_norm": 4.608563423156738,
      "kl": 0.19677734375,
      "learning_rate": 8.253138075313807e-07,
      "loss": 0.1279,
      "reward": 0.7092698216438293,
      "reward_std": 0.10046786814928055,
      "reward_std_all": 0.09932399354875088,
      "rewards-max/forensics_sinkhorn_reward": 0.9654972851276398,
      "rewards-max/forensics_strict_edge_reward": 0.5706980228424072,
      "rewards-max/forensics_tiou_reward": 0.9359853267669678,
      "rewards/forensics_sinkhorn_reward": 0.6606115698814392,
      "rewards/forensics_strict_edge_reward": 0.03313949517905712,
      "rewards/forensics_tiou_reward": 0.39007987082004547,
      "step": 168
    },
    {
      "completion_length": 19.5625,
      "epoch": 0.7071129707112971,
      "grad_norm": 4.749666690826416,
      "kl": 0.2177734375,
      "learning_rate": 8.242677824267782e-07,
      "loss": -0.0088,
      "reward": 0.7007420063018799,
      "reward_std": 0.09632006287574768,
      "reward_std_all": 0.09923867881298065,
      "rewards-max/forensics_sinkhorn_reward": 0.9776131212711334,
      "rewards-max/forensics_strict_edge_reward": 0.32777850329875946,
      "rewards-max/forensics_tiou_reward": 0.9653380215167999,
      "rewards/forensics_sinkhorn_reward": 0.6220038533210754,
      "rewards/forensics_strict_edge_reward": 0.014414345845580101,
      "rewards/forensics_tiou_reward": 0.30887626111507416,
      "step": 169
    },
    {
      "completion_length": 21.03125,
      "epoch": 0.7112970711297071,
      "grad_norm": 4.82975959777832,
      "kl": 0.23291015625,
      "learning_rate": 8.232217573221757e-07,
      "loss": 0.1357,
      "reward": 0.6858732104301453,
      "reward_std": 0.08914325386285782,
      "reward_std_all": 0.10013720765709877,
      "rewards-max/forensics_sinkhorn_reward": 0.9543515145778656,
      "rewards-max/forensics_strict_edge_reward": 0.21255244687199593,
      "rewards-max/forensics_tiou_reward": 0.9183217585086823,
      "rewards/forensics_sinkhorn_reward": 0.6125137507915497,
      "rewards/forensics_strict_edge_reward": 0.011253198143094778,
      "rewards/forensics_tiou_reward": 0.34243327379226685,
      "step": 170
    },
    {
      "completion_length": 22.421875,
      "epoch": 0.7154811715481172,
      "grad_norm": 5.333897590637207,
      "kl": 0.2802734375,
      "learning_rate": 8.221757322175732e-07,
      "loss": 0.0551,
      "reward": 0.7362884879112244,
      "reward_std": 0.08620581030845642,
      "reward_std_all": 0.08162283524870872,
      "rewards-max/forensics_sinkhorn_reward": 0.897550493478775,
      "rewards-max/forensics_strict_edge_reward": 0.23461434477940202,
      "rewards-max/forensics_tiou_reward": 0.7886795699596405,
      "rewards/forensics_sinkhorn_reward": 0.6584223508834839,
      "rewards/forensics_strict_edge_reward": 0.010587756987661123,
      "rewards/forensics_tiou_reward": 0.36498674750328064,
      "step": 171
    },
    {
      "completion_length": 20.921875,
      "epoch": 0.7196652719665272,
      "grad_norm": 7.522031784057617,
      "kl": 0.265625,
      "learning_rate": 8.211297071129707e-07,
      "loss": -0.0829,
      "reward": 0.7298644185066223,
      "reward_std": 0.06700720638036728,
      "reward_std_all": 0.06892429664731026,
      "rewards-max/forensics_sinkhorn_reward": 0.9507623910903931,
      "rewards-max/forensics_strict_edge_reward": 0.3356340229511261,
      "rewards-max/forensics_tiou_reward": 0.9068817794322968,
      "rewards/forensics_sinkhorn_reward": 0.6315017640590668,
      "rewards/forensics_strict_edge_reward": 0.02636410901322961,
      "rewards/forensics_tiou_reward": 0.34307055175304413,
      "step": 172
    },
    {
      "completion_length": 20.703125,
      "epoch": 0.7238493723849372,
      "grad_norm": 5.5261149406433105,
      "kl": 0.248046875,
      "learning_rate": 8.200836820083682e-07,
      "loss": 0.1556,
      "reward": 0.715053528547287,
      "reward_std": 0.07545671612024307,
      "reward_std_all": 0.08501413837075233,
      "rewards-max/forensics_sinkhorn_reward": 0.9646815359592438,
      "rewards-max/forensics_strict_edge_reward": 0.12867088103666902,
      "rewards-max/forensics_tiou_reward": 0.9257853329181671,
      "rewards/forensics_sinkhorn_reward": 0.6319543421268463,
      "rewards/forensics_strict_edge_reward": 0.016478356148581952,
      "rewards/forensics_tiou_reward": 0.3557286411523819,
      "step": 173
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.7280334728033473,
      "grad_norm": 9.0404052734375,
      "kl": 0.4794921875,
      "learning_rate": 8.190376569037657e-07,
      "loss": 0.0897,
      "reward": 0.7177753448486328,
      "reward_std": 0.10026637837290764,
      "reward_std_all": 0.09576952457427979,
      "rewards-max/forensics_sinkhorn_reward": 0.9680361747741699,
      "rewards-max/forensics_strict_edge_reward": 0.1950456127524376,
      "rewards-max/forensics_tiou_reward": 0.9454415440559387,
      "rewards/forensics_sinkhorn_reward": 0.6153059899806976,
      "rewards/forensics_strict_edge_reward": 0.009480579989030957,
      "rewards/forensics_tiou_reward": 0.3676567077636719,
      "step": 174
    },
    {
      "completion_length": 21.8125,
      "epoch": 0.7322175732217573,
      "grad_norm": 4.963599681854248,
      "kl": 0.4345703125,
      "learning_rate": 8.179916317991632e-07,
      "loss": 0.1678,
      "reward": 0.717424750328064,
      "reward_std": 0.07531416416168213,
      "reward_std_all": 0.08150504902005196,
      "rewards-max/forensics_sinkhorn_reward": 0.9658482372760773,
      "rewards-max/forensics_strict_edge_reward": 0.09375573880970478,
      "rewards-max/forensics_tiou_reward": 0.943931519985199,
      "rewards/forensics_sinkhorn_reward": 0.6116562187671661,
      "rewards/forensics_strict_edge_reward": 0.004544706433080137,
      "rewards/forensics_tiou_reward": 0.33763501048088074,
      "step": 175
    },
    {
      "completion_length": 21.359375,
      "epoch": 0.7364016736401674,
      "grad_norm": 5.820630073547363,
      "kl": 0.271484375,
      "learning_rate": 8.169456066945606e-07,
      "loss": 0.0062,
      "reward": 0.6993172466754913,
      "reward_std": 0.11497990414500237,
      "reward_std_all": 0.11586394719779491,
      "rewards-max/forensics_sinkhorn_reward": 0.9576795101165771,
      "rewards-max/forensics_strict_edge_reward": 0.31356135814530717,
      "rewards-max/forensics_tiou_reward": 0.8913523256778717,
      "rewards/forensics_sinkhorn_reward": 0.6056685149669647,
      "rewards/forensics_strict_edge_reward": 0.013288306519257276,
      "rewards/forensics_tiou_reward": 0.31871330738067627,
      "step": 176
    },
    {
      "completion_length": 21.21875,
      "epoch": 0.7405857740585774,
      "grad_norm": 5.894769191741943,
      "kl": 0.28125,
      "learning_rate": 8.158995815899581e-07,
      "loss": 0.0909,
      "reward": 0.6864084601402283,
      "reward_std": 0.10840115323662758,
      "reward_std_all": 0.11402837932109833,
      "rewards-max/forensics_sinkhorn_reward": 0.9785868525505066,
      "rewards-max/forensics_strict_edge_reward": 0.31765206158161163,
      "rewards-max/forensics_tiou_reward": 0.9676269888877869,
      "rewards/forensics_sinkhorn_reward": 0.6227598786354065,
      "rewards/forensics_strict_edge_reward": 0.01798000419512391,
      "rewards/forensics_tiou_reward": 0.3941192030906677,
      "step": 177
    },
    {
      "completion_length": 24.5,
      "epoch": 0.7447698744769874,
      "grad_norm": 5.760733604431152,
      "kl": 0.20556640625,
      "learning_rate": 8.148535564853556e-07,
      "loss": 0.1081,
      "reward": 0.732400119304657,
      "reward_std": 0.0865699015557766,
      "reward_std_all": 0.08220086060464382,
      "rewards-max/forensics_sinkhorn_reward": 0.9610759615898132,
      "rewards-max/forensics_strict_edge_reward": 0.27234726399183273,
      "rewards-max/forensics_tiou_reward": 0.9300421476364136,
      "rewards/forensics_sinkhorn_reward": 0.6535515189170837,
      "rewards/forensics_strict_edge_reward": 0.020769049413502216,
      "rewards/forensics_tiou_reward": 0.4426453411579132,
      "step": 178
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.7489539748953975,
      "grad_norm": 4.571873664855957,
      "kl": 0.2548828125,
      "learning_rate": 8.138075313807531e-07,
      "loss": 0.2227,
      "reward": 0.6810352206230164,
      "reward_std": 0.11018044129014015,
      "reward_std_all": 0.11512799188494682,
      "rewards-max/forensics_sinkhorn_reward": 0.9742603898048401,
      "rewards-max/forensics_strict_edge_reward": 0.2785482034087181,
      "rewards-max/forensics_tiou_reward": 0.9564823508262634,
      "rewards/forensics_sinkhorn_reward": 0.59070885181427,
      "rewards/forensics_strict_edge_reward": 0.015514379600062966,
      "rewards/forensics_tiou_reward": 0.35286885499954224,
      "step": 179
    },
    {
      "completion_length": 23.359375,
      "epoch": 0.7531380753138075,
      "grad_norm": 5.338188648223877,
      "kl": 0.265625,
      "learning_rate": 8.127615062761506e-07,
      "loss": 0.1596,
      "reward": 0.7127581238746643,
      "reward_std": 0.06994228810071945,
      "reward_std_all": 0.08261599764227867,
      "rewards-max/forensics_sinkhorn_reward": 0.9765154123306274,
      "rewards-max/forensics_strict_edge_reward": 0.42252161353826523,
      "rewards-max/forensics_tiou_reward": 0.9657044112682343,
      "rewards/forensics_sinkhorn_reward": 0.6143083870410919,
      "rewards/forensics_strict_edge_reward": 0.02166398987174034,
      "rewards/forensics_tiou_reward": 0.37963052093982697,
      "step": 180
    },
    {
      "completion_length": 24.140625,
      "epoch": 0.7573221757322176,
      "grad_norm": 6.478147983551025,
      "kl": 0.287109375,
      "learning_rate": 8.117154811715481e-07,
      "loss": 0.2354,
      "reward": 0.6902070939540863,
      "reward_std": 0.13542743027210236,
      "reward_std_all": 0.1281181052327156,
      "rewards-max/forensics_sinkhorn_reward": 0.9609997868537903,
      "rewards-max/forensics_strict_edge_reward": 0.217692282050848,
      "rewards-max/forensics_tiou_reward": 0.9346928298473358,
      "rewards/forensics_sinkhorn_reward": 0.604763925075531,
      "rewards/forensics_strict_edge_reward": 0.01053567131748423,
      "rewards/forensics_tiou_reward": 0.37909799814224243,
      "step": 181
    },
    {
      "completion_length": 22.171875,
      "epoch": 0.7615062761506276,
      "grad_norm": 5.352029323577881,
      "kl": 0.26123046875,
      "learning_rate": 8.106694560669455e-07,
      "loss": 0.1512,
      "reward": 0.7186486423015594,
      "reward_std": 0.07821368239820004,
      "reward_std_all": 0.08181911334395409,
      "rewards-max/forensics_sinkhorn_reward": 0.9414144456386566,
      "rewards-max/forensics_strict_edge_reward": 0.05332063743844628,
      "rewards-max/forensics_tiou_reward": 0.8728806376457214,
      "rewards/forensics_sinkhorn_reward": 0.6329746842384338,
      "rewards/forensics_strict_edge_reward": 0.0034763451549224555,
      "rewards/forensics_tiou_reward": 0.34364189207553864,
      "step": 182
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.7656903765690377,
      "grad_norm": 4.666289806365967,
      "kl": 0.2724609375,
      "learning_rate": 8.096234309623431e-07,
      "loss": 0.1648,
      "reward": 0.7232230007648468,
      "reward_std": 0.09313157573342323,
      "reward_std_all": 0.08910420909523964,
      "rewards-max/forensics_sinkhorn_reward": 0.9059130251407623,
      "rewards-max/forensics_strict_edge_reward": 0.05240807402879,
      "rewards-max/forensics_tiou_reward": 0.771072506904602,
      "rewards/forensics_sinkhorn_reward": 0.6550672352313995,
      "rewards/forensics_strict_edge_reward": 0.0028342632576823235,
      "rewards/forensics_tiou_reward": 0.3615460991859436,
      "step": 183
    },
    {
      "completion_length": 20.890625,
      "epoch": 0.7698744769874477,
      "grad_norm": 4.180379867553711,
      "kl": 0.296875,
      "learning_rate": 8.085774058577406e-07,
      "loss": 0.0377,
      "reward": 0.6699222326278687,
      "reward_std": 0.0556208360940218,
      "reward_std_all": 0.09394245594739914,
      "rewards-max/forensics_sinkhorn_reward": 0.9704256057739258,
      "rewards-max/forensics_strict_edge_reward": 0.20061761140823364,
      "rewards-max/forensics_tiou_reward": 0.9499876797199249,
      "rewards/forensics_sinkhorn_reward": 0.5950228571891785,
      "rewards/forensics_strict_edge_reward": 0.009434736333787441,
      "rewards/forensics_tiou_reward": 0.3171960562467575,
      "step": 184
    },
    {
      "completion_length": 22.6875,
      "epoch": 0.7740585774058577,
      "grad_norm": 6.078247547149658,
      "kl": 0.22314453125,
      "learning_rate": 8.075313807531381e-07,
      "loss": 0.0852,
      "reward": 0.6975635290145874,
      "reward_std": 0.08371119946241379,
      "reward_std_all": 0.09482472017407417,
      "rewards-max/forensics_sinkhorn_reward": 0.866980791091919,
      "rewards-max/forensics_strict_edge_reward": 0.07465028576552868,
      "rewards-max/forensics_tiou_reward": 0.7448551654815674,
      "rewards/forensics_sinkhorn_reward": 0.5711353719234467,
      "rewards/forensics_strict_edge_reward": 0.003807974688243121,
      "rewards/forensics_tiou_reward": 0.27331826090812683,
      "step": 185
    },
    {
      "completion_length": 20.859375,
      "epoch": 0.7782426778242678,
      "grad_norm": 4.911483287811279,
      "kl": 0.23583984375,
      "learning_rate": 8.064853556485355e-07,
      "loss": 0.1226,
      "reward": 0.7413358092308044,
      "reward_std": 0.08023371733725071,
      "reward_std_all": 0.07558745332062244,
      "rewards-max/forensics_sinkhorn_reward": 0.9540651142597198,
      "rewards-max/forensics_strict_edge_reward": 0.24823588877916336,
      "rewards-max/forensics_tiou_reward": 0.9118434488773346,
      "rewards/forensics_sinkhorn_reward": 0.6863599121570587,
      "rewards/forensics_strict_edge_reward": 0.011472431477159262,
      "rewards/forensics_tiou_reward": 0.3978749215602875,
      "step": 186
    },
    {
      "completion_length": 22.78125,
      "epoch": 0.7824267782426778,
      "grad_norm": 4.558420658111572,
      "kl": 0.3125,
      "learning_rate": 8.05439330543933e-07,
      "loss": 0.1044,
      "reward": 0.7397260069847107,
      "reward_std": 0.06641267985105515,
      "reward_std_all": 0.06894383579492569,
      "rewards-max/forensics_sinkhorn_reward": 0.9573806524276733,
      "rewards-max/forensics_strict_edge_reward": 0.15081009082496166,
      "rewards-max/forensics_tiou_reward": 0.9136832654476166,
      "rewards/forensics_sinkhorn_reward": 0.6324379742145538,
      "rewards/forensics_strict_edge_reward": 0.0063161421567201614,
      "rewards/forensics_tiou_reward": 0.3617943078279495,
      "step": 187
    },
    {
      "completion_length": 20.8125,
      "epoch": 0.7866108786610879,
      "grad_norm": 4.654816150665283,
      "kl": 0.27734375,
      "learning_rate": 8.043933054393305e-07,
      "loss": 0.1234,
      "reward": 0.7431464791297913,
      "reward_std": 0.07111221924424171,
      "reward_std_all": 0.0704822801053524,
      "rewards-max/forensics_sinkhorn_reward": 0.9563363790512085,
      "rewards-max/forensics_strict_edge_reward": 0.22791624069213867,
      "rewards-max/forensics_tiou_reward": 0.9064853191375732,
      "rewards/forensics_sinkhorn_reward": 0.678522527217865,
      "rewards/forensics_strict_edge_reward": 0.012026468757539988,
      "rewards/forensics_tiou_reward": 0.41203758120536804,
      "step": 188
    },
    {
      "completion_length": 21.59375,
      "epoch": 0.7907949790794979,
      "grad_norm": 5.3310770988464355,
      "kl": 0.26220703125,
      "learning_rate": 8.03347280334728e-07,
      "loss": -0.0394,
      "reward": 0.6920918822288513,
      "reward_std": 0.02805657684803009,
      "reward_std_all": 0.06813815236091614,
      "rewards-max/forensics_sinkhorn_reward": 0.9434541463851929,
      "rewards-max/forensics_strict_edge_reward": 0.5013622571714222,
      "rewards-max/forensics_tiou_reward": 0.8409468531608582,
      "rewards/forensics_sinkhorn_reward": 0.6068909466266632,
      "rewards/forensics_strict_edge_reward": 0.034732214582618326,
      "rewards/forensics_tiou_reward": 0.29831649363040924,
      "step": 189
    },
    {
      "completion_length": 19.78125,
      "epoch": 0.7949790794979079,
      "grad_norm": 5.114962100982666,
      "kl": 0.25634765625,
      "learning_rate": 8.023012552301254e-07,
      "loss": 0.0464,
      "reward": 0.7469412684440613,
      "reward_std": 0.04131001606583595,
      "reward_std_all": 0.05094727873802185,
      "rewards-max/forensics_sinkhorn_reward": 0.9467133581638336,
      "rewards-max/forensics_strict_edge_reward": 0.04415102303028107,
      "rewards-max/forensics_tiou_reward": 0.871561735868454,
      "rewards/forensics_sinkhorn_reward": 0.721664160490036,
      "rewards/forensics_strict_edge_reward": 0.0026343846693634987,
      "rewards/forensics_tiou_reward": 0.46225830912590027,
      "step": 190
    },
    {
      "completion_length": 21.0625,
      "epoch": 0.799163179916318,
      "grad_norm": 4.957932472229004,
      "kl": 0.2802734375,
      "learning_rate": 8.01255230125523e-07,
      "loss": 0.0034,
      "reward": 0.692973792552948,
      "reward_std": 0.05059903487563133,
      "reward_std_all": 0.0809769481420517,
      "rewards-max/forensics_sinkhorn_reward": 0.9307883083820343,
      "rewards-max/forensics_strict_edge_reward": 0.5266096722334623,
      "rewards-max/forensics_tiou_reward": 0.8697608113288879,
      "rewards/forensics_sinkhorn_reward": 0.6255812346935272,
      "rewards/forensics_strict_edge_reward": 0.04239494679495692,
      "rewards/forensics_tiou_reward": 0.3855508416891098,
      "step": 191
    },
    {
      "completion_length": 20.375,
      "epoch": 0.803347280334728,
      "grad_norm": 7.088929176330566,
      "kl": 0.2626953125,
      "learning_rate": 8.002092050209205e-07,
      "loss": -0.0003,
      "reward": 0.7466039061546326,
      "reward_std": 0.04003085754811764,
      "reward_std_all": 0.047299740836024284,
      "rewards-max/forensics_sinkhorn_reward": 0.9310287535190582,
      "rewards-max/forensics_strict_edge_reward": 0.08645752258598804,
      "rewards-max/forensics_tiou_reward": 0.8497164249420166,
      "rewards/forensics_sinkhorn_reward": 0.6697781682014465,
      "rewards/forensics_strict_edge_reward": 0.0037664477131329477,
      "rewards/forensics_tiou_reward": 0.38307714462280273,
      "step": 192
    },
    {
      "completion_length": 19.84375,
      "epoch": 0.8075313807531381,
      "grad_norm": 5.076760292053223,
      "kl": 0.23974609375,
      "learning_rate": 7.99163179916318e-07,
      "loss": 0.1265,
      "reward": 0.6790121495723724,
      "reward_std": 0.09539597481489182,
      "reward_std_all": 0.10932547226548195,
      "rewards-max/forensics_sinkhorn_reward": 0.9390204846858978,
      "rewards-max/forensics_strict_edge_reward": 0.07489612954668701,
      "rewards-max/forensics_tiou_reward": 0.8684399425983429,
      "rewards/forensics_sinkhorn_reward": 0.6029604077339172,
      "rewards/forensics_strict_edge_reward": 0.008933703824368422,
      "rewards/forensics_tiou_reward": 0.31830114126205444,
      "step": 193
    },
    {
      "completion_length": 20.828125,
      "epoch": 0.8117154811715481,
      "grad_norm": 4.821258068084717,
      "kl": 0.26708984375,
      "learning_rate": 7.981171548117155e-07,
      "loss": 0.1038,
      "reward": 0.7499898374080658,
      "reward_std": 0.03562485985457897,
      "reward_std_all": 0.047785788774490356,
      "rewards-max/forensics_sinkhorn_reward": 0.9025647640228271,
      "rewards-max/forensics_strict_edge_reward": 0.18650605366565287,
      "rewards-max/forensics_tiou_reward": 0.7647497653961182,
      "rewards/forensics_sinkhorn_reward": 0.6881471276283264,
      "rewards/forensics_strict_edge_reward": 0.013928334330557846,
      "rewards/forensics_tiou_reward": 0.44093039631843567,
      "step": 194
    },
    {
      "completion_length": 20.265625,
      "epoch": 0.8158995815899581,
      "grad_norm": 4.867382049560547,
      "kl": 0.271484375,
      "learning_rate": 7.97071129707113e-07,
      "loss": 0.0272,
      "reward": 0.7340047359466553,
      "reward_std": 0.06481895409524441,
      "reward_std_all": 0.06791162304580212,
      "rewards-max/forensics_sinkhorn_reward": 0.9444732367992401,
      "rewards-max/forensics_strict_edge_reward": 0.23802655562758446,
      "rewards-max/forensics_tiou_reward": 0.8958822786808014,
      "rewards/forensics_sinkhorn_reward": 0.6622244715690613,
      "rewards/forensics_strict_edge_reward": 0.019306362490169704,
      "rewards/forensics_tiou_reward": 0.3528270274400711,
      "step": 195
    },
    {
      "completion_length": 20.203125,
      "epoch": 0.8200836820083682,
      "grad_norm": 6.092759132385254,
      "kl": 0.408203125,
      "learning_rate": 7.960251046025104e-07,
      "loss": 0.0939,
      "reward": 0.7105686962604523,
      "reward_std": 0.05358976870775223,
      "reward_std_all": 0.07229812443256378,
      "rewards-max/forensics_sinkhorn_reward": 0.9485949277877808,
      "rewards-max/forensics_strict_edge_reward": 0.2913331761956215,
      "rewards-max/forensics_tiou_reward": 0.9034143686294556,
      "rewards/forensics_sinkhorn_reward": 0.6351572275161743,
      "rewards/forensics_strict_edge_reward": 0.016612429171800613,
      "rewards/forensics_tiou_reward": 0.353889599442482,
      "step": 196
    },
    {
      "completion_length": 20.359375,
      "epoch": 0.8242677824267782,
      "grad_norm": 6.47003173828125,
      "kl": 0.2265625,
      "learning_rate": 7.949790794979079e-07,
      "loss": 0.0179,
      "reward": 0.7213460803031921,
      "reward_std": 0.0787841584533453,
      "reward_std_all": 0.08048686012625694,
      "rewards-max/forensics_sinkhorn_reward": 0.9510525166988373,
      "rewards-max/forensics_strict_edge_reward": 0.06868399027734995,
      "rewards-max/forensics_tiou_reward": 0.9086790680885315,
      "rewards/forensics_sinkhorn_reward": 0.6379052698612213,
      "rewards/forensics_strict_edge_reward": 0.0034741093404591084,
      "rewards/forensics_tiou_reward": 0.3587264269590378,
      "step": 197
    },
    {
      "completion_length": 24.109375,
      "epoch": 0.8284518828451883,
      "grad_norm": 4.9308977127075195,
      "kl": 0.3037109375,
      "learning_rate": 7.939330543933053e-07,
      "loss": 0.0426,
      "reward": 0.7199100255966187,
      "reward_std": 0.08835803344845772,
      "reward_std_all": 0.08894059807062149,
      "rewards-max/forensics_sinkhorn_reward": 0.9750539362430573,
      "rewards-max/forensics_strict_edge_reward": 0.41336529701948166,
      "rewards-max/forensics_tiou_reward": 0.9581307470798492,
      "rewards/forensics_sinkhorn_reward": 0.6374363005161285,
      "rewards/forensics_strict_edge_reward": 0.02930558263324201,
      "rewards/forensics_tiou_reward": 0.38829539716243744,
      "step": 198
    },
    {
      "completion_length": 20.453125,
      "epoch": 0.8326359832635983,
      "grad_norm": 4.7786126136779785,
      "kl": 0.2275390625,
      "learning_rate": 7.928870292887028e-07,
      "loss": 0.1172,
      "reward": 0.694622278213501,
      "reward_std": 0.07404712215065956,
      "reward_std_all": 0.09320850670337677,
      "rewards-max/forensics_sinkhorn_reward": 0.9820418059825897,
      "rewards-max/forensics_strict_edge_reward": 0.5736939311027527,
      "rewards-max/forensics_tiou_reward": 0.9822457134723663,
      "rewards/forensics_sinkhorn_reward": 0.6399542987346649,
      "rewards/forensics_strict_edge_reward": 0.025799697265028954,
      "rewards/forensics_tiou_reward": 0.3796079456806183,
      "step": 199
    },
    {
      "completion_length": 22.0,
      "epoch": 0.8368200836820083,
      "grad_norm": 5.165499210357666,
      "kl": 0.2578125,
      "learning_rate": 7.918410041841004e-07,
      "loss": 0.0856,
      "reward": 0.7035074532032013,
      "reward_std": 0.10215475037693977,
      "reward_std_all": 0.10496344044804573,
      "rewards-max/forensics_sinkhorn_reward": 0.9775552451610565,
      "rewards-max/forensics_strict_edge_reward": 0.32841892540454865,
      "rewards-max/forensics_tiou_reward": 0.9668327569961548,
      "rewards/forensics_sinkhorn_reward": 0.6305457949638367,
      "rewards/forensics_strict_edge_reward": 0.019224142655730247,
      "rewards/forensics_tiou_reward": 0.36698606610298157,
      "step": 200
    },
    {
      "completion_length": 20.328125,
      "epoch": 0.8410041841004184,
      "grad_norm": 5.607640743255615,
      "kl": 0.3134765625,
      "learning_rate": 7.907949790794979e-07,
      "loss": 0.1282,
      "reward": 0.690526694059372,
      "reward_std": 0.0732532050460577,
      "reward_std_all": 0.09262852929532528,
      "rewards-max/forensics_sinkhorn_reward": 0.9563167989253998,
      "rewards-max/forensics_strict_edge_reward": 0.16784934140741825,
      "rewards-max/forensics_tiou_reward": 0.9003986716270447,
      "rewards/forensics_sinkhorn_reward": 0.6213846206665039,
      "rewards/forensics_strict_edge_reward": 0.019201030023396015,
      "rewards/forensics_tiou_reward": 0.3741897791624069,
      "step": 201
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.8451882845188284,
      "grad_norm": 5.446465015411377,
      "kl": 0.306640625,
      "learning_rate": 7.897489539748954e-07,
      "loss": 0.0954,
      "reward": 0.7147013545036316,
      "reward_std": 0.10738338157534599,
      "reward_std_all": 0.10416274517774582,
      "rewards-max/forensics_sinkhorn_reward": 0.9247649908065796,
      "rewards-max/forensics_strict_edge_reward": 0.012306820135563612,
      "rewards-max/forensics_tiou_reward": 0.834851086139679,
      "rewards/forensics_sinkhorn_reward": 0.6390773355960846,
      "rewards/forensics_strict_edge_reward": 0.0012146014341851696,
      "rewards/forensics_tiou_reward": 0.35359059274196625,
      "step": 202
    },
    {
      "completion_length": 22.296875,
      "epoch": 0.8493723849372385,
      "grad_norm": 4.79252815246582,
      "kl": 0.33056640625,
      "learning_rate": 7.887029288702929e-07,
      "loss": 0.2062,
      "reward": 0.699801504611969,
      "reward_std": 0.06696468777954578,
      "reward_std_all": 0.08648882433772087,
      "rewards-max/forensics_sinkhorn_reward": 0.9564332365989685,
      "rewards-max/forensics_strict_edge_reward": 0.02389150857925415,
      "rewards-max/forensics_tiou_reward": 0.9130278527736664,
      "rewards/forensics_sinkhorn_reward": 0.5940331816673279,
      "rewards/forensics_strict_edge_reward": 0.0012221020879223943,
      "rewards/forensics_tiou_reward": 0.30499011278152466,
      "step": 203
    },
    {
      "completion_length": 19.484375,
      "epoch": 0.8535564853556485,
      "grad_norm": 4.167577743530273,
      "kl": 0.4013671875,
      "learning_rate": 7.876569037656904e-07,
      "loss": 0.1622,
      "reward": 0.7474254667758942,
      "reward_std": 0.0706116259098053,
      "reward_std_all": 0.067515192553401,
      "rewards-max/forensics_sinkhorn_reward": 0.9427282512187958,
      "rewards-max/forensics_strict_edge_reward": 0.11885337345302105,
      "rewards-max/forensics_tiou_reward": 0.8850829601287842,
      "rewards/forensics_sinkhorn_reward": 0.6881535351276398,
      "rewards/forensics_strict_edge_reward": 0.009006137493997812,
      "rewards/forensics_tiou_reward": 0.4044206738471985,
      "step": 204
    },
    {
      "completion_length": 20.140625,
      "epoch": 0.8577405857740585,
      "grad_norm": 4.050464630126953,
      "kl": 0.24658203125,
      "learning_rate": 7.866108786610879e-07,
      "loss": 0.0854,
      "reward": 0.7309167087078094,
      "reward_std": 0.07853349670767784,
      "reward_std_all": 0.07707677409052849,
      "rewards-max/forensics_sinkhorn_reward": 0.9536681175231934,
      "rewards-max/forensics_strict_edge_reward": 0.2637604847550392,
      "rewards-max/forensics_tiou_reward": 0.9096547067165375,
      "rewards/forensics_sinkhorn_reward": 0.675160825252533,
      "rewards/forensics_strict_edge_reward": 0.015057516982778907,
      "rewards/forensics_tiou_reward": 0.4191613048315048,
      "step": 205
    },
    {
      "completion_length": 20.375,
      "epoch": 0.8619246861924686,
      "grad_norm": 6.035658836364746,
      "kl": 0.25390625,
      "learning_rate": 7.855648535564853e-07,
      "loss": 0.1061,
      "reward": 0.6500872373580933,
      "reward_std": 0.16759751737117767,
      "reward_std_all": 0.16113682091236115,
      "rewards-max/forensics_sinkhorn_reward": 0.9463159143924713,
      "rewards-max/forensics_strict_edge_reward": 0.16070556640625,
      "rewards-max/forensics_tiou_reward": 0.888408362865448,
      "rewards/forensics_sinkhorn_reward": 0.5813648402690887,
      "rewards/forensics_strict_edge_reward": 0.014692022814415395,
      "rewards/forensics_tiou_reward": 0.31857647001743317,
      "step": 206
    },
    {
      "completion_length": 20.28125,
      "epoch": 0.8661087866108786,
      "grad_norm": 4.620895862579346,
      "kl": 0.294921875,
      "learning_rate": 7.845188284518827e-07,
      "loss": 0.1026,
      "reward": 0.6755037307739258,
      "reward_std": 0.10675615072250366,
      "reward_std_all": 0.11926915496587753,
      "rewards-max/forensics_sinkhorn_reward": 0.8805698454380035,
      "rewards-max/forensics_strict_edge_reward": 0.008956128731369972,
      "rewards-max/forensics_tiou_reward": 0.7185945212841034,
      "rewards/forensics_sinkhorn_reward": 0.5794994831085205,
      "rewards/forensics_strict_edge_reward": 0.0005583731544902548,
      "rewards/forensics_tiou_reward": 0.26064684987068176,
      "step": 207
    },
    {
      "completion_length": 20.5,
      "epoch": 0.8702928870292888,
      "grad_norm": 4.822731971740723,
      "kl": 0.2314453125,
      "learning_rate": 7.834728033472802e-07,
      "loss": 0.1336,
      "reward": 0.7357246279716492,
      "reward_std": 0.04422017000615597,
      "reward_std_all": 0.056861113756895065,
      "rewards-max/forensics_sinkhorn_reward": 0.9279181361198425,
      "rewards-max/forensics_strict_edge_reward": 0.031220927834510803,
      "rewards-max/forensics_tiou_reward": 0.8525193929672241,
      "rewards/forensics_sinkhorn_reward": 0.6221988797187805,
      "rewards/forensics_strict_edge_reward": 0.0014550034538842738,
      "rewards/forensics_tiou_reward": 0.27739737927913666,
      "step": 208
    },
    {
      "completion_length": 19.296875,
      "epoch": 0.8744769874476988,
      "grad_norm": 4.408634662628174,
      "kl": 0.3193359375,
      "learning_rate": 7.824267782426778e-07,
      "loss": 0.044,
      "reward": 0.7202275395393372,
      "reward_std": 0.06674887984991074,
      "reward_std_all": 0.07814647629857063,
      "rewards-max/forensics_sinkhorn_reward": 0.9683468043804169,
      "rewards-max/forensics_strict_edge_reward": 0.3903922364115715,
      "rewards-max/forensics_tiou_reward": 0.9434461891651154,
      "rewards/forensics_sinkhorn_reward": 0.6884645521640778,
      "rewards/forensics_strict_edge_reward": 0.037731640972197056,
      "rewards/forensics_tiou_reward": 0.43337053060531616,
      "step": 209
    },
    {
      "completion_length": 20.203125,
      "epoch": 0.8786610878661087,
      "grad_norm": 4.497759819030762,
      "kl": 0.310546875,
      "learning_rate": 7.813807531380753e-07,
      "loss": -0.0019,
      "reward": 0.77031210064888,
      "reward_std": 0.03271242417395115,
      "reward_std_all": 0.034881070256233215,
      "rewards-max/forensics_sinkhorn_reward": 0.9553817808628082,
      "rewards-max/forensics_strict_edge_reward": 0.22642836719751358,
      "rewards-max/forensics_tiou_reward": 0.9152238667011261,
      "rewards/forensics_sinkhorn_reward": 0.7052748799324036,
      "rewards/forensics_strict_edge_reward": 0.01773984543979168,
      "rewards/forensics_tiou_reward": 0.4198288023471832,
      "step": 210
    },
    {
      "completion_length": 19.21875,
      "epoch": 0.8828451882845189,
      "grad_norm": 4.926708221435547,
      "kl": 0.25048828125,
      "learning_rate": 7.803347280334728e-07,
      "loss": -0.0213,
      "reward": 0.6945699453353882,
      "reward_std": 0.06722225807607174,
      "reward_std_all": 0.08755525201559067,
      "rewards-max/forensics_sinkhorn_reward": 0.9628848433494568,
      "rewards-max/forensics_strict_edge_reward": 0.20633811503648758,
      "rewards-max/forensics_tiou_reward": 0.9300531148910522,
      "rewards/forensics_sinkhorn_reward": 0.639600932598114,
      "rewards/forensics_strict_edge_reward": 0.01894747605547309,
      "rewards/forensics_tiou_reward": 0.3792068809270859,
      "step": 211
    },
    {
      "completion_length": 19.46875,
      "epoch": 0.8870292887029289,
      "grad_norm": 6.631626129150391,
      "kl": 0.287109375,
      "learning_rate": 7.792887029288703e-07,
      "loss": 0.0774,
      "reward": 0.7551279067993164,
      "reward_std": 0.06357284262776375,
      "reward_std_all": 0.06052180379629135,
      "rewards-max/forensics_sinkhorn_reward": 0.9263469576835632,
      "rewards-max/forensics_strict_edge_reward": 0.08862914890050888,
      "rewards-max/forensics_tiou_reward": 0.8467391431331635,
      "rewards/forensics_sinkhorn_reward": 0.694234699010849,
      "rewards/forensics_strict_edge_reward": 0.007341579534113407,
      "rewards/forensics_tiou_reward": 0.4117412567138672,
      "step": 212
    },
    {
      "completion_length": 20.046875,
      "epoch": 0.891213389121339,
      "grad_norm": 5.299755096435547,
      "kl": 0.26904296875,
      "learning_rate": 7.782426778242678e-07,
      "loss": 0.1075,
      "reward": 0.7167121171951294,
      "reward_std": 0.07757367193698883,
      "reward_std_all": 0.08093937113881111,
      "rewards-max/forensics_sinkhorn_reward": 0.9363547563552856,
      "rewards-max/forensics_strict_edge_reward": 0.09514021500945091,
      "rewards-max/forensics_tiou_reward": 0.8796469867229462,
      "rewards/forensics_sinkhorn_reward": 0.6505784094333649,
      "rewards/forensics_strict_edge_reward": 0.0056483387015759945,
      "rewards/forensics_tiou_reward": 0.36984802782535553,
      "step": 213
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.895397489539749,
      "grad_norm": 5.223036289215088,
      "kl": 0.255859375,
      "learning_rate": 7.771966527196653e-07,
      "loss": 0.1171,
      "reward": 0.7317288517951965,
      "reward_std": 0.05352240242063999,
      "reward_std_all": 0.06332438252866268,
      "rewards-max/forensics_sinkhorn_reward": 0.9655516147613525,
      "rewards-max/forensics_strict_edge_reward": 0.3535345382988453,
      "rewards-max/forensics_tiou_reward": 0.9337237775325775,
      "rewards/forensics_sinkhorn_reward": 0.6457472741603851,
      "rewards/forensics_strict_edge_reward": 0.014892256585881114,
      "rewards/forensics_tiou_reward": 0.35181906819343567,
      "step": 214
    },
    {
      "completion_length": 22.515625,
      "epoch": 0.899581589958159,
      "grad_norm": 4.777080535888672,
      "kl": 0.2578125,
      "learning_rate": 7.761506276150628e-07,
      "loss": 0.168,
      "reward": 0.681799441576004,
      "reward_std": 0.09564986824989319,
      "reward_std_all": 0.10520672053098679,
      "rewards-max/forensics_sinkhorn_reward": 0.9432241916656494,
      "rewards-max/forensics_strict_edge_reward": 0.001116176601499319,
      "rewards-max/forensics_tiou_reward": 0.8643622994422913,
      "rewards/forensics_sinkhorn_reward": 0.5765259563922882,
      "rewards/forensics_strict_edge_reward": 5.6564005490145064e-05,
      "rewards/forensics_tiou_reward": 0.27009208500385284,
      "step": 215
    },
    {
      "completion_length": 19.65625,
      "epoch": 0.9037656903765691,
      "grad_norm": 5.499263763427734,
      "kl": 0.32421875,
      "learning_rate": 7.751046025104601e-07,
      "loss": 0.1235,
      "reward": 0.6726978123188019,
      "reward_std": 0.09687311574816704,
      "reward_std_all": 0.11433961614966393,
      "rewards-max/forensics_sinkhorn_reward": 0.9543121755123138,
      "rewards-max/forensics_strict_edge_reward": 0.18365223240107298,
      "rewards-max/forensics_tiou_reward": 0.9093098640441895,
      "rewards/forensics_sinkhorn_reward": 0.6028346121311188,
      "rewards/forensics_strict_edge_reward": 0.017959509044885635,
      "rewards/forensics_tiou_reward": 0.3357139080762863,
      "step": 216
    },
    {
      "completion_length": 22.46875,
      "epoch": 0.9079497907949791,
      "grad_norm": 5.4732136726379395,
      "kl": 0.3232421875,
      "learning_rate": 7.740585774058576e-07,
      "loss": 0.1249,
      "reward": 0.7045511901378632,
      "reward_std": 0.1200871616601944,
      "reward_std_all": 0.11646539531648159,
      "rewards-max/forensics_sinkhorn_reward": 0.9648829996585846,
      "rewards-max/forensics_strict_edge_reward": 0.24673397094011307,
      "rewards-max/forensics_tiou_reward": 0.9392014145851135,
      "rewards/forensics_sinkhorn_reward": 0.6299004256725311,
      "rewards/forensics_strict_edge_reward": 0.018356477841734886,
      "rewards/forensics_tiou_reward": 0.37254974246025085,
      "step": 217
    },
    {
      "completion_length": 19.890625,
      "epoch": 0.9121338912133892,
      "grad_norm": 3.8892974853515625,
      "kl": 0.2255859375,
      "learning_rate": 7.730125523012552e-07,
      "loss": 0.0577,
      "reward": 0.7081009447574615,
      "reward_std": 0.05336896702647209,
      "reward_std_all": 0.07243559136986732,
      "rewards-max/forensics_sinkhorn_reward": 0.967617928981781,
      "rewards-max/forensics_strict_edge_reward": 0.1977487187832594,
      "rewards-max/forensics_tiou_reward": 0.9374742209911346,
      "rewards/forensics_sinkhorn_reward": 0.6589865684509277,
      "rewards/forensics_strict_edge_reward": 0.0250325296074152,
      "rewards/forensics_tiou_reward": 0.38227762281894684,
      "step": 218
    },
    {
      "completion_length": 21.28125,
      "epoch": 0.9163179916317992,
      "grad_norm": 4.998129844665527,
      "kl": 0.271484375,
      "learning_rate": 7.719665271966527e-07,
      "loss": 0.0519,
      "reward": 0.7152718901634216,
      "reward_std": 0.08320680260658264,
      "reward_std_all": 0.08922704309225082,
      "rewards-max/forensics_sinkhorn_reward": 0.9344898760318756,
      "rewards-max/forensics_strict_edge_reward": 0.05088130757212639,
      "rewards-max/forensics_tiou_reward": 0.8526702523231506,
      "rewards/forensics_sinkhorn_reward": 0.646421492099762,
      "rewards/forensics_strict_edge_reward": 0.003013207286130637,
      "rewards/forensics_tiou_reward": 0.36404040455818176,
      "step": 219
    },
    {
      "completion_length": 20.4375,
      "epoch": 0.9205020920502092,
      "grad_norm": 4.879301071166992,
      "kl": 0.2607421875,
      "learning_rate": 7.709205020920502e-07,
      "loss": 0.0681,
      "reward": 0.7328119874000549,
      "reward_std": 0.08105135336518288,
      "reward_std_all": 0.0834808424115181,
      "rewards-max/forensics_sinkhorn_reward": 0.9759572744369507,
      "rewards-max/forensics_strict_edge_reward": 0.5670294612646103,
      "rewards-max/forensics_tiou_reward": 0.9659081697463989,
      "rewards/forensics_sinkhorn_reward": 0.6918609738349915,
      "rewards/forensics_strict_edge_reward": 0.03754476271569729,
      "rewards/forensics_tiou_reward": 0.44550229609012604,
      "step": 220
    },
    {
      "completion_length": 19.71875,
      "epoch": 0.9246861924686193,
      "grad_norm": 6.227497100830078,
      "kl": 0.306640625,
      "learning_rate": 7.698744769874477e-07,
      "loss": 0.141,
      "reward": 0.6678308844566345,
      "reward_std": 0.12938807904720306,
      "reward_std_all": 0.13453643023967743,
      "rewards-max/forensics_sinkhorn_reward": 0.9687178432941437,
      "rewards-max/forensics_strict_edge_reward": 0.18572804180439562,
      "rewards-max/forensics_tiou_reward": 0.9497002959251404,
      "rewards/forensics_sinkhorn_reward": 0.6019987463951111,
      "rewards/forensics_strict_edge_reward": 0.013163720723241568,
      "rewards/forensics_tiou_reward": 0.3287741094827652,
      "step": 221
    },
    {
      "completion_length": 22.171875,
      "epoch": 0.9288702928870293,
      "grad_norm": 4.645644187927246,
      "kl": 0.26171875,
      "learning_rate": 7.688284518828452e-07,
      "loss": 0.0849,
      "reward": 0.6616452932357788,
      "reward_std": 0.10164998844265938,
      "reward_std_all": 0.12125591933727264,
      "rewards-max/forensics_sinkhorn_reward": 0.9110790193080902,
      "rewards-max/forensics_strict_edge_reward": 0.051892359610064887,
      "rewards-max/forensics_tiou_reward": 0.776103287935257,
      "rewards/forensics_sinkhorn_reward": 0.5486891567707062,
      "rewards/forensics_strict_edge_reward": 0.004287584337362205,
      "rewards/forensics_tiou_reward": 0.26858408004045486,
      "step": 222
    },
    {
      "completion_length": 22.25,
      "epoch": 0.9330543933054394,
      "grad_norm": 4.29263973236084,
      "kl": 0.25341796875,
      "learning_rate": 7.677824267782427e-07,
      "loss": 0.0638,
      "reward": 0.6946006119251251,
      "reward_std": 0.09068462625145912,
      "reward_std_all": 0.10206852853298187,
      "rewards-max/forensics_sinkhorn_reward": 0.9648747146129608,
      "rewards-max/forensics_strict_edge_reward": 0.22055835276842117,
      "rewards-max/forensics_tiou_reward": 0.9301437437534332,
      "rewards/forensics_sinkhorn_reward": 0.5969293713569641,
      "rewards/forensics_strict_edge_reward": 0.010295022279024124,
      "rewards/forensics_tiou_reward": 0.33268865942955017,
      "step": 223
    },
    {
      "completion_length": 20.546875,
      "epoch": 0.9372384937238494,
      "grad_norm": 9.88431453704834,
      "kl": 0.591796875,
      "learning_rate": 7.667364016736401e-07,
      "loss": 0.1952,
      "reward": 0.7374420762062073,
      "reward_std": 0.08559796307235956,
      "reward_std_all": 0.08021692465990782,
      "rewards-max/forensics_sinkhorn_reward": 0.939942330121994,
      "rewards-max/forensics_strict_edge_reward": 0.31683944538235664,
      "rewards-max/forensics_tiou_reward": 0.8841030597686768,
      "rewards/forensics_sinkhorn_reward": 0.687539130449295,
      "rewards/forensics_strict_edge_reward": 0.027055883780121803,
      "rewards/forensics_tiou_reward": 0.4272647649049759,
      "step": 224
    },
    {
      "completion_length": 20.625,
      "epoch": 0.9414225941422594,
      "grad_norm": 5.033627510070801,
      "kl": 0.2607421875,
      "learning_rate": 7.656903765690376e-07,
      "loss": 0.1613,
      "reward": 0.7389708161354065,
      "reward_std": 0.034983862191438675,
      "reward_std_all": 0.05040700361132622,
      "rewards-max/forensics_sinkhorn_reward": 0.9737356603145599,
      "rewards-max/forensics_strict_edge_reward": 0.31173690408468246,
      "rewards-max/forensics_tiou_reward": 0.9591719806194305,
      "rewards/forensics_sinkhorn_reward": 0.6852349638938904,
      "rewards/forensics_strict_edge_reward": 0.02048386074602604,
      "rewards/forensics_tiou_reward": 0.4368036836385727,
      "step": 225
    },
    {
      "completion_length": 20.859375,
      "epoch": 0.9456066945606695,
      "grad_norm": 5.38408899307251,
      "kl": 0.3232421875,
      "learning_rate": 7.64644351464435e-07,
      "loss": 0.0594,
      "reward": 0.6985317766666412,
      "reward_std": 0.09021784737706184,
      "reward_std_all": 0.09620615467429161,
      "rewards-max/forensics_sinkhorn_reward": 0.9218111932277679,
      "rewards-max/forensics_strict_edge_reward": 0.12434585019946098,
      "rewards-max/forensics_tiou_reward": 0.835780143737793,
      "rewards/forensics_sinkhorn_reward": 0.6057334542274475,
      "rewards/forensics_strict_edge_reward": 0.006028227740898728,
      "rewards/forensics_tiou_reward": 0.29355475306510925,
      "step": 226
    },
    {
      "completion_length": 21.328125,
      "epoch": 0.9497907949790795,
      "grad_norm": 5.639710903167725,
      "kl": 0.3125,
      "learning_rate": 7.635983263598326e-07,
      "loss": 0.0549,
      "reward": 0.6939015686511993,
      "reward_std": 0.08223217353224754,
      "reward_std_all": 0.09818992018699646,
      "rewards-max/forensics_sinkhorn_reward": 0.9618691205978394,
      "rewards-max/forensics_strict_edge_reward": 0.2517708130180836,
      "rewards-max/forensics_tiou_reward": 0.9161262810230255,
      "rewards/forensics_sinkhorn_reward": 0.6308459937572479,
      "rewards/forensics_strict_edge_reward": 0.020058519672602415,
      "rewards/forensics_tiou_reward": 0.3845456540584564,
      "step": 227
    },
    {
      "completion_length": 22.109375,
      "epoch": 0.9539748953974896,
      "grad_norm": 4.202918529510498,
      "kl": 0.359375,
      "learning_rate": 7.625523012552301e-07,
      "loss": 0.0755,
      "reward": 0.7124184966087341,
      "reward_std": 0.06368163600564003,
      "reward_std_all": 0.07679053023457527,
      "rewards-max/forensics_sinkhorn_reward": 0.915969580411911,
      "rewards-max/forensics_strict_edge_reward": 0.026626420207321644,
      "rewards-max/forensics_tiou_reward": 0.8124396204948425,
      "rewards/forensics_sinkhorn_reward": 0.6099174916744232,
      "rewards/forensics_strict_edge_reward": 0.001460633531678468,
      "rewards/forensics_tiou_reward": 0.34409740567207336,
      "step": 228
    },
    {
      "completion_length": 21.65625,
      "epoch": 0.9581589958158996,
      "grad_norm": 5.472174167633057,
      "kl": 0.26953125,
      "learning_rate": 7.615062761506276e-07,
      "loss": 0.0154,
      "reward": 0.6924487948417664,
      "reward_std": 0.10149269178509712,
      "reward_std_all": 0.10421555861830711,
      "rewards-max/forensics_sinkhorn_reward": 0.9609431028366089,
      "rewards-max/forensics_strict_edge_reward": 0.41780946403741837,
      "rewards-max/forensics_tiou_reward": 0.9319242835044861,
      "rewards/forensics_sinkhorn_reward": 0.6029556691646576,
      "rewards/forensics_strict_edge_reward": 0.020875100046396255,
      "rewards/forensics_tiou_reward": 0.31719809770584106,
      "step": 229
    },
    {
      "completion_length": 23.734375,
      "epoch": 0.9623430962343096,
      "grad_norm": 6.452548503875732,
      "kl": 0.33740234375,
      "learning_rate": 7.604602510460251e-07,
      "loss": 0.0634,
      "reward": 0.6739404797554016,
      "reward_std": 0.06933488510549068,
      "reward_std_all": 0.09946681559085846,
      "rewards-max/forensics_sinkhorn_reward": 0.8981770575046539,
      "rewards-max/forensics_strict_edge_reward": 0.15367461740970612,
      "rewards-max/forensics_tiou_reward": 0.7974380850791931,
      "rewards/forensics_sinkhorn_reward": 0.5568421185016632,
      "rewards/forensics_strict_edge_reward": 0.012305048119742423,
      "rewards/forensics_tiou_reward": 0.3289639875292778,
      "step": 230
    },
    {
      "completion_length": 23.15625,
      "epoch": 0.9665271966527197,
      "grad_norm": 5.875587463378906,
      "kl": 0.220703125,
      "learning_rate": 7.594142259414226e-07,
      "loss": 0.1638,
      "reward": 0.6920333206653595,
      "reward_std": 0.11943447217345238,
      "reward_std_all": 0.11732860654592514,
      "rewards-max/forensics_sinkhorn_reward": 0.9707545638084412,
      "rewards-max/forensics_strict_edge_reward": 0.31821663305163383,
      "rewards-max/forensics_tiou_reward": 0.9532275497913361,
      "rewards/forensics_sinkhorn_reward": 0.6042298078536987,
      "rewards/forensics_strict_edge_reward": 0.03017473267391324,
      "rewards/forensics_tiou_reward": 0.39425912499427795,
      "step": 231
    },
    {
      "completion_length": 22.125,
      "epoch": 0.9707112970711297,
      "grad_norm": 4.7664594650268555,
      "kl": 0.24755859375,
      "learning_rate": 7.5836820083682e-07,
      "loss": 0.1334,
      "reward": 0.7194902300834656,
      "reward_std": 0.10719345882534981,
      "reward_std_all": 0.10004895552992821,
      "rewards-max/forensics_sinkhorn_reward": 0.9164687991142273,
      "rewards-max/forensics_strict_edge_reward": 0.11588521883822978,
      "rewards-max/forensics_tiou_reward": 0.8205472528934479,
      "rewards/forensics_sinkhorn_reward": 0.6460604667663574,
      "rewards/forensics_strict_edge_reward": 0.004965630545484601,
      "rewards/forensics_tiou_reward": 0.38802000880241394,
      "step": 232
    },
    {
      "completion_length": 24.5,
      "epoch": 0.9748953974895398,
      "grad_norm": 5.845853805541992,
      "kl": 0.259765625,
      "learning_rate": 7.573221757322175e-07,
      "loss": 0.06,
      "reward": 0.6405424475669861,
      "reward_std": 0.10704319179058075,
      "reward_std_all": 0.12797784805297852,
      "rewards-max/forensics_sinkhorn_reward": 0.9285145998001099,
      "rewards-max/forensics_strict_edge_reward": 0.13274266495136544,
      "rewards-max/forensics_tiou_reward": 0.8195031583309174,
      "rewards/forensics_sinkhorn_reward": 0.5158945471048355,
      "rewards/forensics_strict_edge_reward": 0.005797609148430638,
      "rewards/forensics_tiou_reward": 0.2814014181494713,
      "step": 233
    },
    {
      "completion_length": 21.984375,
      "epoch": 0.9790794979079498,
      "grad_norm": 5.526343822479248,
      "kl": 0.31201171875,
      "learning_rate": 7.56276150627615e-07,
      "loss": 0.0822,
      "reward": 0.6956334710121155,
      "reward_std": 0.11709210649132729,
      "reward_std_all": 0.1160833090543747,
      "rewards-max/forensics_sinkhorn_reward": 0.9705070853233337,
      "rewards-max/forensics_strict_edge_reward": 0.34273045510053635,
      "rewards-max/forensics_tiou_reward": 0.953639417886734,
      "rewards/forensics_sinkhorn_reward": 0.6233696043491364,
      "rewards/forensics_strict_edge_reward": 0.019238328095525503,
      "rewards/forensics_tiou_reward": 0.3864114135503769,
      "step": 234
    },
    {
      "completion_length": 21.09375,
      "epoch": 0.9832635983263598,
      "grad_norm": 7.615586757659912,
      "kl": 0.7529296875,
      "learning_rate": 7.552301255230126e-07,
      "loss": 0.1222,
      "reward": 0.7324070334434509,
      "reward_std": 0.0773455910384655,
      "reward_std_all": 0.07724421843886375,
      "rewards-max/forensics_sinkhorn_reward": 0.9710746109485626,
      "rewards-max/forensics_strict_edge_reward": 0.4745236784219742,
      "rewards-max/forensics_tiou_reward": 0.9506377279758453,
      "rewards/forensics_sinkhorn_reward": 0.6733347475528717,
      "rewards/forensics_strict_edge_reward": 0.038334669545292854,
      "rewards/forensics_tiou_reward": 0.4230734407901764,
      "step": 235
    },
    {
      "completion_length": 20.453125,
      "epoch": 0.9874476987447699,
      "grad_norm": 5.034271717071533,
      "kl": 0.25732421875,
      "learning_rate": 7.5418410041841e-07,
      "loss": 0.1307,
      "reward": 0.7382775545120239,
      "reward_std": 0.06989796832203865,
      "reward_std_all": 0.06860821321606636,
      "rewards-max/forensics_sinkhorn_reward": 0.9740655720233917,
      "rewards-max/forensics_strict_edge_reward": 0.2694893293082714,
      "rewards-max/forensics_tiou_reward": 0.9629536867141724,
      "rewards/forensics_sinkhorn_reward": 0.6725009083747864,
      "rewards/forensics_strict_edge_reward": 0.025589708471670747,
      "rewards/forensics_tiou_reward": 0.45027658343315125,
      "step": 236
    },
    {
      "completion_length": 20.734375,
      "epoch": 0.9916317991631799,
      "grad_norm": 4.381447792053223,
      "kl": 0.2900390625,
      "learning_rate": 7.531380753138075e-07,
      "loss": 0.0319,
      "reward": 0.758794903755188,
      "reward_std": 0.05634373612701893,
      "reward_std_all": 0.054766204208135605,
      "rewards-max/forensics_sinkhorn_reward": 0.9527279138565063,
      "rewards-max/forensics_strict_edge_reward": 0.01189633784815669,
      "rewards-max/forensics_tiou_reward": 0.8816069662570953,
      "rewards/forensics_sinkhorn_reward": 0.7050448954105377,
      "rewards/forensics_strict_edge_reward": 0.000960753473918885,
      "rewards/forensics_tiou_reward": 0.4526982307434082,
      "step": 237
    },
    {
      "completion_length": 20.703125,
      "epoch": 0.99581589958159,
      "grad_norm": 5.85205602645874,
      "kl": 0.2431640625,
      "learning_rate": 7.52092050209205e-07,
      "loss": 0.0608,
      "reward": 0.717137336730957,
      "reward_std": 0.06904628500342369,
      "reward_std_all": 0.07754873018711805,
      "rewards-max/forensics_sinkhorn_reward": 0.9258435070514679,
      "rewards-max/forensics_strict_edge_reward": 0.138397173024714,
      "rewards-max/forensics_tiou_reward": 0.8490520715713501,
      "rewards/forensics_sinkhorn_reward": 0.6462241411209106,
      "rewards/forensics_strict_edge_reward": 0.01117898931261152,
      "rewards/forensics_tiou_reward": 0.36416836082935333,
      "step": 238
    },
    {
      "completion_length": 21.285715103149414,
      "epoch": 1.0,
      "grad_norm": 6.7843403816223145,
      "kl": 0.29296875,
      "learning_rate": 7.510460251046025e-07,
      "loss": 0.0101,
      "reward": 0.6354602575302124,
      "reward_std": 0.08988117892295122,
      "reward_std_all": 0.12486503832042217,
      "rewards-max/forensics_sinkhorn_reward": 0.958726316690445,
      "rewards-max/forensics_strict_edge_reward": 0.34077148139476776,
      "rewards-max/forensics_tiou_reward": 0.9263701140880585,
      "rewards/forensics_sinkhorn_reward": 0.6140728741884232,
      "rewards/forensics_strict_edge_reward": 0.049246806651353836,
      "rewards/forensics_tiou_reward": 0.4069260507822037,
      "step": 239
    },
    {
      "completion_length": 20.53125,
      "epoch": 1.00418410041841,
      "grad_norm": 10.263815879821777,
      "kl": 0.6875,
      "learning_rate": 7.5e-07,
      "loss": 0.0152,
      "reward": 0.7726685404777527,
      "reward_std": 0.04655778221786022,
      "reward_std_all": 0.04397745616734028,
      "rewards-max/forensics_sinkhorn_reward": 0.9733178019523621,
      "rewards-max/forensics_strict_edge_reward": 0.6030158251523972,
      "rewards-max/forensics_tiou_reward": 0.9597459435462952,
      "rewards/forensics_sinkhorn_reward": 0.7245742678642273,
      "rewards/forensics_strict_edge_reward": 0.04070581495761871,
      "rewards/forensics_tiou_reward": 0.47796982526779175,
      "step": 240
    },
    {
      "completion_length": 20.015625,
      "epoch": 1.00836820083682,
      "grad_norm": 5.016006946563721,
      "kl": 0.4130859375,
      "learning_rate": 7.489539748953974e-07,
      "loss": 0.0134,
      "reward": 0.7733662128448486,
      "reward_std": 0.03855019807815552,
      "reward_std_all": 0.03811032883822918,
      "rewards-max/forensics_sinkhorn_reward": 0.9772185683250427,
      "rewards-max/forensics_strict_edge_reward": 0.1859860047698021,
      "rewards-max/forensics_tiou_reward": 0.963525265455246,
      "rewards/forensics_sinkhorn_reward": 0.7293819785118103,
      "rewards/forensics_strict_edge_reward": 0.012829666025936604,
      "rewards/forensics_tiou_reward": 0.46614184975624084,
      "step": 241
    },
    {
      "completion_length": 21.046875,
      "epoch": 1.0125523012552302,
      "grad_norm": 5.680051326751709,
      "kl": 0.3583984375,
      "learning_rate": 7.479079497907949e-07,
      "loss": 0.1096,
      "reward": 0.7129886746406555,
      "reward_std": 0.08402472734451294,
      "reward_std_all": 0.0865556076169014,
      "rewards-max/forensics_sinkhorn_reward": 0.9180521368980408,
      "rewards-max/forensics_strict_edge_reward": 0.1787000596523285,
      "rewards-max/forensics_tiou_reward": 0.8400890827178955,
      "rewards/forensics_sinkhorn_reward": 0.6503418385982513,
      "rewards/forensics_strict_edge_reward": 0.008226454257965088,
      "rewards/forensics_tiou_reward": 0.3943044990301132,
      "step": 242
    },
    {
      "completion_length": 21.328125,
      "epoch": 1.0167364016736402,
      "grad_norm": 5.800715446472168,
      "kl": 0.3154296875,
      "learning_rate": 7.468619246861925e-07,
      "loss": -0.0183,
      "reward": 0.7220052778720856,
      "reward_std": 0.06816997192800045,
      "reward_std_all": 0.07394501939415932,
      "rewards-max/forensics_sinkhorn_reward": 0.9650080800056458,
      "rewards-max/forensics_strict_edge_reward": 0.35250507295131683,
      "rewards-max/forensics_tiou_reward": 0.9400874674320221,
      "rewards/forensics_sinkhorn_reward": 0.6866760849952698,
      "rewards/forensics_strict_edge_reward": 0.026974298991262913,
      "rewards/forensics_tiou_reward": 0.4193222224712372,
      "step": 243
    },
    {
      "completion_length": 20.359375,
      "epoch": 1.0209205020920502,
      "grad_norm": 5.1950483322143555,
      "kl": 0.2578125,
      "learning_rate": 7.4581589958159e-07,
      "loss": 0.0617,
      "reward": 0.7412309050559998,
      "reward_std": 0.06346333026885986,
      "reward_std_all": 0.06350195966660976,
      "rewards-max/forensics_sinkhorn_reward": 0.9540898501873016,
      "rewards-max/forensics_strict_edge_reward": 0.5321668684482574,
      "rewards-max/forensics_tiou_reward": 0.9076995849609375,
      "rewards/forensics_sinkhorn_reward": 0.6933325827121735,
      "rewards/forensics_strict_edge_reward": 0.0338446032255888,
      "rewards/forensics_tiou_reward": 0.41645781695842743,
      "step": 244
    },
    {
      "completion_length": 20.171875,
      "epoch": 1.0251046025104602,
      "grad_norm": 4.2725629806518555,
      "kl": 0.2978515625,
      "learning_rate": 7.447698744769874e-07,
      "loss": 0.0173,
      "reward": 0.7302744686603546,
      "reward_std": 0.05419662408530712,
      "reward_std_all": 0.06439521163702011,
      "rewards-max/forensics_sinkhorn_reward": 0.9703651666641235,
      "rewards-max/forensics_strict_edge_reward": 0.13078086450695992,
      "rewards-max/forensics_tiou_reward": 0.9436216652393341,
      "rewards/forensics_sinkhorn_reward": 0.6797276437282562,
      "rewards/forensics_strict_edge_reward": 0.012515742797404528,
      "rewards/forensics_tiou_reward": 0.4172895848751068,
      "step": 245
    },
    {
      "completion_length": 19.59375,
      "epoch": 1.0292887029288702,
      "grad_norm": 5.171640872955322,
      "kl": 0.27001953125,
      "learning_rate": 7.437238493723849e-07,
      "loss": -0.0298,
      "reward": 0.7518593370914459,
      "reward_std": 0.050258444622159004,
      "reward_std_all": 0.05138812214136124,
      "rewards-max/forensics_sinkhorn_reward": 0.9511432945728302,
      "rewards-max/forensics_strict_edge_reward": 0.31600624322891235,
      "rewards-max/forensics_tiou_reward": 0.9080944061279297,
      "rewards/forensics_sinkhorn_reward": 0.6827609837055206,
      "rewards/forensics_strict_edge_reward": 0.026151111349463463,
      "rewards/forensics_tiou_reward": 0.40426434576511383,
      "step": 246
    },
    {
      "completion_length": 22.8125,
      "epoch": 1.0334728033472804,
      "grad_norm": 4.372252464294434,
      "kl": 0.2900390625,
      "learning_rate": 7.426778242677824e-07,
      "loss": 0.1086,
      "reward": 0.746184915304184,
      "reward_std": 0.06085577234625816,
      "reward_std_all": 0.060599204152822495,
      "rewards-max/forensics_sinkhorn_reward": 0.9402990341186523,
      "rewards-max/forensics_strict_edge_reward": 0.2104724422097206,
      "rewards-max/forensics_tiou_reward": 0.8809722661972046,
      "rewards/forensics_sinkhorn_reward": 0.657789558172226,
      "rewards/forensics_strict_edge_reward": 0.009209772106260061,
      "rewards/forensics_tiou_reward": 0.3894611895084381,
      "step": 247
    },
    {
      "completion_length": 20.859375,
      "epoch": 1.0376569037656904,
      "grad_norm": 4.921392440795898,
      "kl": 0.25048828125,
      "learning_rate": 7.416317991631799e-07,
      "loss": 0.1459,
      "reward": 0.6722292900085449,
      "reward_std": 0.10247575864195824,
      "reward_std_all": 0.1146182268857956,
      "rewards-max/forensics_sinkhorn_reward": 0.9745583832263947,
      "rewards-max/forensics_strict_edge_reward": 0.27715054154396057,
      "rewards-max/forensics_tiou_reward": 0.9610633850097656,
      "rewards/forensics_sinkhorn_reward": 0.5918108224868774,
      "rewards/forensics_strict_edge_reward": 0.02072135964408517,
      "rewards/forensics_tiou_reward": 0.3043350279331207,
      "step": 248
    },
    {
      "completion_length": 21.109375,
      "epoch": 1.0418410041841004,
      "grad_norm": 5.481258392333984,
      "kl": 0.2607421875,
      "learning_rate": 7.405857740585773e-07,
      "loss": 0.0284,
      "reward": 0.7483410537242889,
      "reward_std": 0.06830970197916031,
      "reward_std_all": 0.06544230505824089,
      "rewards-max/forensics_sinkhorn_reward": 0.9379848539829254,
      "rewards-max/forensics_strict_edge_reward": 0.19903841707855463,
      "rewards-max/forensics_tiou_reward": 0.8745395839214325,
      "rewards/forensics_sinkhorn_reward": 0.6681229174137115,
      "rewards/forensics_strict_edge_reward": 0.020100158639252186,
      "rewards/forensics_tiou_reward": 0.36414971947669983,
      "step": 249
    },
    {
      "completion_length": 20.453125,
      "epoch": 1.0460251046025104,
      "grad_norm": 5.537356853485107,
      "kl": 0.30859375,
      "learning_rate": 7.395397489539748e-07,
      "loss": 0.0495,
      "reward": 0.7678001523017883,
      "reward_std": 0.01851685903966427,
      "reward_std_all": 0.030330680310726166,
      "rewards-max/forensics_sinkhorn_reward": 0.9769670069217682,
      "rewards-max/forensics_strict_edge_reward": 0.5167631208896637,
      "rewards-max/forensics_tiou_reward": 0.9583216905593872,
      "rewards/forensics_sinkhorn_reward": 0.7327381074428558,
      "rewards/forensics_strict_edge_reward": 0.032549225725233555,
      "rewards/forensics_tiou_reward": 0.46521179378032684,
      "step": 250
    },
    {
      "completion_length": 20.921875,
      "epoch": 1.0502092050209204,
      "grad_norm": 5.4088592529296875,
      "kl": 0.275390625,
      "learning_rate": 7.384937238493723e-07,
      "loss": 0.0741,
      "reward": 0.7418130338191986,
      "reward_std": 0.08045502007007599,
      "reward_std_all": 0.0759919136762619,
      "rewards-max/forensics_sinkhorn_reward": 0.9589093625545502,
      "rewards-max/forensics_strict_edge_reward": 0.05882888182532042,
      "rewards-max/forensics_tiou_reward": 0.8987727165222168,
      "rewards/forensics_sinkhorn_reward": 0.6594116687774658,
      "rewards/forensics_strict_edge_reward": 0.003129347547655925,
      "rewards/forensics_tiou_reward": 0.398367315530777,
      "step": 251
    },
    {
      "completion_length": 21.15625,
      "epoch": 1.0543933054393306,
      "grad_norm": 5.635743618011475,
      "kl": 0.365234375,
      "learning_rate": 7.374476987447699e-07,
      "loss": 0.2162,
      "reward": 0.7168496549129486,
      "reward_std": 0.08575595170259476,
      "reward_std_all": 0.09185581654310226,
      "rewards-max/forensics_sinkhorn_reward": 0.9562524557113647,
      "rewards-max/forensics_strict_edge_reward": 0.40880464762449265,
      "rewards-max/forensics_tiou_reward": 0.9173508584499359,
      "rewards/forensics_sinkhorn_reward": 0.6461709141731262,
      "rewards/forensics_strict_edge_reward": 0.03711455315351486,
      "rewards/forensics_tiou_reward": 0.373213529586792,
      "step": 252
    },
    {
      "completion_length": 21.0625,
      "epoch": 1.0585774058577406,
      "grad_norm": 4.935459136962891,
      "kl": 0.23388671875,
      "learning_rate": 7.364016736401674e-07,
      "loss": 0.2464,
      "reward": 0.7427273690700531,
      "reward_std": 0.059148166328668594,
      "reward_std_all": 0.060412436723709106,
      "rewards-max/forensics_sinkhorn_reward": 0.9488475620746613,
      "rewards-max/forensics_strict_edge_reward": 0.34070829302072525,
      "rewards-max/forensics_tiou_reward": 0.9071642160415649,
      "rewards/forensics_sinkhorn_reward": 0.657074362039566,
      "rewards/forensics_strict_edge_reward": 0.019881437998265028,
      "rewards/forensics_tiou_reward": 0.3714035898447037,
      "step": 253
    },
    {
      "completion_length": 19.453125,
      "epoch": 1.0627615062761506,
      "grad_norm": 5.908838272094727,
      "kl": 0.267578125,
      "learning_rate": 7.353556485355649e-07,
      "loss": 0.0874,
      "reward": 0.6901156902313232,
      "reward_std": 0.07457122951745987,
      "reward_std_all": 0.0943029336631298,
      "rewards-max/forensics_sinkhorn_reward": 0.9444599449634552,
      "rewards-max/forensics_strict_edge_reward": 0.1300692930817604,
      "rewards-max/forensics_tiou_reward": 0.8916686177253723,
      "rewards/forensics_sinkhorn_reward": 0.6452396512031555,
      "rewards/forensics_strict_edge_reward": 0.01043338654562831,
      "rewards/forensics_tiou_reward": 0.37825708091259,
      "step": 254
    },
    {
      "completion_length": 21.09375,
      "epoch": 1.0669456066945606,
      "grad_norm": 6.0237884521484375,
      "kl": 0.3447265625,
      "learning_rate": 7.343096234309623e-07,
      "loss": 0.1113,
      "reward": 0.7018671929836273,
      "reward_std": 0.0868038572371006,
      "reward_std_all": 0.09923921525478363,
      "rewards-max/forensics_sinkhorn_reward": 0.980742484331131,
      "rewards-max/forensics_strict_edge_reward": 0.5314356125891209,
      "rewards-max/forensics_tiou_reward": 0.9786509275436401,
      "rewards/forensics_sinkhorn_reward": 0.6254835724830627,
      "rewards/forensics_strict_edge_reward": 0.02372239006217569,
      "rewards/forensics_tiou_reward": 0.37202315032482147,
      "step": 255
    },
    {
      "completion_length": 21.375,
      "epoch": 1.0711297071129706,
      "grad_norm": 6.203367710113525,
      "kl": 0.3798828125,
      "learning_rate": 7.332635983263598e-07,
      "loss": 0.0307,
      "reward": 0.7210807502269745,
      "reward_std": 0.05680258758366108,
      "reward_std_all": 0.07050857320427895,
      "rewards-max/forensics_sinkhorn_reward": 0.9689191579818726,
      "rewards-max/forensics_strict_edge_reward": 0.5123617630451918,
      "rewards-max/forensics_tiou_reward": 0.9422232508659363,
      "rewards/forensics_sinkhorn_reward": 0.6479473114013672,
      "rewards/forensics_strict_edge_reward": 0.038637656602077186,
      "rewards/forensics_tiou_reward": 0.3953600078821182,
      "step": 256
    },
    {
      "completion_length": 22.328125,
      "epoch": 1.0753138075313808,
      "grad_norm": 4.445494174957275,
      "kl": 0.24169921875,
      "learning_rate": 7.322175732217572e-07,
      "loss": 0.1725,
      "reward": 0.7114525139331818,
      "reward_std": 0.06861026398837566,
      "reward_std_all": 0.08104980364441872,
      "rewards-max/forensics_sinkhorn_reward": 0.9660225808620453,
      "rewards-max/forensics_strict_edge_reward": 0.29929934069514275,
      "rewards-max/forensics_tiou_reward": 0.9341052174568176,
      "rewards/forensics_sinkhorn_reward": 0.6342103183269501,
      "rewards/forensics_strict_edge_reward": 0.015288401860743761,
      "rewards/forensics_tiou_reward": 0.4144172966480255,
      "step": 257
    },
    {
      "completion_length": 22.421875,
      "epoch": 1.0794979079497908,
      "grad_norm": 5.0317864418029785,
      "kl": 0.26318359375,
      "learning_rate": 7.311715481171547e-07,
      "loss": 0.0483,
      "reward": 0.6883804798126221,
      "reward_std": 0.05161470081657171,
      "reward_std_all": 0.08446935936808586,
      "rewards-max/forensics_sinkhorn_reward": 0.9223603308200836,
      "rewards-max/forensics_strict_edge_reward": 0.0815911814570427,
      "rewards-max/forensics_tiou_reward": 0.8406060934066772,
      "rewards/forensics_sinkhorn_reward": 0.5805424153804779,
      "rewards/forensics_strict_edge_reward": 0.0040447848150506616,
      "rewards/forensics_tiou_reward": 0.3214479684829712,
      "step": 258
    },
    {
      "completion_length": 22.1875,
      "epoch": 1.0836820083682008,
      "grad_norm": 5.1123857498168945,
      "kl": 0.3203125,
      "learning_rate": 7.301255230125522e-07,
      "loss": 0.0582,
      "reward": 0.6928124725818634,
      "reward_std": 0.09505556151270866,
      "reward_std_all": 0.10010580345988274,
      "rewards-max/forensics_sinkhorn_reward": 0.9366211295127869,
      "rewards-max/forensics_strict_edge_reward": 0.1849136003293097,
      "rewards-max/forensics_tiou_reward": 0.8706889152526855,
      "rewards/forensics_sinkhorn_reward": 0.588175505399704,
      "rewards/forensics_strict_edge_reward": 0.013222242094343528,
      "rewards/forensics_tiou_reward": 0.2948000133037567,
      "step": 259
    },
    {
      "completion_length": 24.0625,
      "epoch": 1.0878661087866108,
      "grad_norm": 4.542353630065918,
      "kl": 0.224609375,
      "learning_rate": 7.290794979079497e-07,
      "loss": 0.0383,
      "reward": 0.7445971369743347,
      "reward_std": 0.06806101463735104,
      "reward_std_all": 0.06929253600537777,
      "rewards-max/forensics_sinkhorn_reward": 0.9193792641162872,
      "rewards-max/forensics_strict_edge_reward": 0.373468185774982,
      "rewards-max/forensics_tiou_reward": 0.8451119661331177,
      "rewards/forensics_sinkhorn_reward": 0.662115067243576,
      "rewards/forensics_strict_edge_reward": 0.03770581027492881,
      "rewards/forensics_tiou_reward": 0.4718703478574753,
      "step": 260
    },
    {
      "completion_length": 22.5,
      "epoch": 1.0920502092050208,
      "grad_norm": 4.703904628753662,
      "kl": 0.27490234375,
      "learning_rate": 7.280334728033473e-07,
      "loss": 0.0966,
      "reward": 0.6829578280448914,
      "reward_std": 0.07774671167135239,
      "reward_std_all": 0.09839214384555817,
      "rewards-max/forensics_sinkhorn_reward": 0.9211088120937347,
      "rewards-max/forensics_strict_edge_reward": 0.38159448746591806,
      "rewards-max/forensics_tiou_reward": 0.8381596505641937,
      "rewards/forensics_sinkhorn_reward": 0.5994299054145813,
      "rewards/forensics_strict_edge_reward": 0.01901136664673686,
      "rewards/forensics_tiou_reward": 0.36021339893341064,
      "step": 261
    },
    {
      "completion_length": 20.53125,
      "epoch": 1.096234309623431,
      "grad_norm": 6.728747844696045,
      "kl": 0.2822265625,
      "learning_rate": 7.269874476987448e-07,
      "loss": -0.0787,
      "reward": 0.747596949338913,
      "reward_std": 0.04924156703054905,
      "reward_std_all": 0.052310049533843994,
      "rewards-max/forensics_sinkhorn_reward": 0.9636066854000092,
      "rewards-max/forensics_strict_edge_reward": 0.3996507991105318,
      "rewards-max/forensics_tiou_reward": 0.9266205430030823,
      "rewards/forensics_sinkhorn_reward": 0.6590697169303894,
      "rewards/forensics_strict_edge_reward": 0.026849967311136425,
      "rewards/forensics_tiou_reward": 0.3949316442012787,
      "step": 262
    },
    {
      "completion_length": 21.265625,
      "epoch": 1.100418410041841,
      "grad_norm": 6.5718793869018555,
      "kl": 0.3681640625,
      "learning_rate": 7.259414225941423e-07,
      "loss": 0.1134,
      "reward": 0.7100832164287567,
      "reward_std": 0.06144014373421669,
      "reward_std_all": 0.07753051072359085,
      "rewards-max/forensics_sinkhorn_reward": 0.9360170364379883,
      "rewards-max/forensics_strict_edge_reward": 0.012156679295003414,
      "rewards-max/forensics_tiou_reward": 0.8491278886795044,
      "rewards/forensics_sinkhorn_reward": 0.6034077703952789,
      "rewards/forensics_strict_edge_reward": 0.0006931869429536164,
      "rewards/forensics_tiou_reward": 0.3467867821455002,
      "step": 263
    },
    {
      "completion_length": 22.96875,
      "epoch": 1.104602510460251,
      "grad_norm": 5.676671504974365,
      "kl": 0.3212890625,
      "learning_rate": 7.248953974895398e-07,
      "loss": -0.0079,
      "reward": 0.7321009635925293,
      "reward_std": 0.06339921243488789,
      "reward_std_all": 0.06800887733697891,
      "rewards-max/forensics_sinkhorn_reward": 0.9613721966743469,
      "rewards-max/forensics_strict_edge_reward": 0.26087334007024765,
      "rewards-max/forensics_tiou_reward": 0.9285984933376312,
      "rewards/forensics_sinkhorn_reward": 0.6561055183410645,
      "rewards/forensics_strict_edge_reward": 0.01113420445472002,
      "rewards/forensics_tiou_reward": 0.3801916688680649,
      "step": 264
    },
    {
      "completion_length": 24.875,
      "epoch": 1.108786610878661,
      "grad_norm": 6.279118061065674,
      "kl": 0.310546875,
      "learning_rate": 7.238493723849371e-07,
      "loss": 0.087,
      "reward": 0.7332375347614288,
      "reward_std": 0.05559493415057659,
      "reward_std_all": 0.06487737596035004,
      "rewards-max/forensics_sinkhorn_reward": 0.9671784937381744,
      "rewards-max/forensics_strict_edge_reward": 0.36689896136522293,
      "rewards-max/forensics_tiou_reward": 0.9502392113208771,
      "rewards/forensics_sinkhorn_reward": 0.6835304200649261,
      "rewards/forensics_strict_edge_reward": 0.01701222639530897,
      "rewards/forensics_tiou_reward": 0.4581402689218521,
      "step": 265
    },
    {
      "completion_length": 20.9375,
      "epoch": 1.112970711297071,
      "grad_norm": 4.686708450317383,
      "kl": 0.24169921875,
      "learning_rate": 7.228033472803346e-07,
      "loss": 0.0851,
      "reward": 0.7397739589214325,
      "reward_std": 0.07602088339626789,
      "reward_std_all": 0.07261131703853607,
      "rewards-max/forensics_sinkhorn_reward": 0.9819470643997192,
      "rewards-max/forensics_strict_edge_reward": 0.5106222033500671,
      "rewards-max/forensics_tiou_reward": 0.9803114831447601,
      "rewards/forensics_sinkhorn_reward": 0.7122699320316315,
      "rewards/forensics_strict_edge_reward": 0.05918388068675995,
      "rewards/forensics_tiou_reward": 0.5279191732406616,
      "step": 266
    },
    {
      "completion_length": 21.5,
      "epoch": 1.1171548117154813,
      "grad_norm": 3.9910244941711426,
      "kl": 0.2958984375,
      "learning_rate": 7.217573221757321e-07,
      "loss": 0.0444,
      "reward": 0.7502884864807129,
      "reward_std": 0.03689399641007185,
      "reward_std_all": 0.04581005126237869,
      "rewards-max/forensics_sinkhorn_reward": 0.9626511037349701,
      "rewards-max/forensics_strict_edge_reward": 0.2471451535820961,
      "rewards-max/forensics_tiou_reward": 0.928362101316452,
      "rewards/forensics_sinkhorn_reward": 0.683777928352356,
      "rewards/forensics_strict_edge_reward": 0.01870537083595991,
      "rewards/forensics_tiou_reward": 0.47183506190776825,
      "step": 267
    },
    {
      "completion_length": 22.375,
      "epoch": 1.1213389121338913,
      "grad_norm": 5.525880813598633,
      "kl": 0.4091796875,
      "learning_rate": 7.207112970711296e-07,
      "loss": 0.1583,
      "reward": 0.7100293636322021,
      "reward_std": 0.07914900407195091,
      "reward_std_all": 0.08735084161162376,
      "rewards-max/forensics_sinkhorn_reward": 0.9612961113452911,
      "rewards-max/forensics_strict_edge_reward": 0.3800518251955509,
      "rewards-max/forensics_tiou_reward": 0.9217847585678101,
      "rewards/forensics_sinkhorn_reward": 0.6201860010623932,
      "rewards/forensics_strict_edge_reward": 0.03165509086102247,
      "rewards/forensics_tiou_reward": 0.35966165363788605,
      "step": 268
    },
    {
      "completion_length": 19.359375,
      "epoch": 1.1255230125523012,
      "grad_norm": 5.744743347167969,
      "kl": 0.2734375,
      "learning_rate": 7.196652719665272e-07,
      "loss": 0.1047,
      "reward": 0.7264256179332733,
      "reward_std": 0.08828457444906235,
      "reward_std_all": 0.08520390838384628,
      "rewards-max/forensics_sinkhorn_reward": 0.9475939273834229,
      "rewards-max/forensics_strict_edge_reward": 0.17495790868997574,
      "rewards-max/forensics_tiou_reward": 0.8991853892803192,
      "rewards/forensics_sinkhorn_reward": 0.686940997838974,
      "rewards/forensics_strict_edge_reward": 0.012848575599491596,
      "rewards/forensics_tiou_reward": 0.40501391887664795,
      "step": 269
    },
    {
      "completion_length": 20.265625,
      "epoch": 1.1297071129707112,
      "grad_norm": 5.41372013092041,
      "kl": 0.291015625,
      "learning_rate": 7.186192468619247e-07,
      "loss": -0.0255,
      "reward": 0.7451973855495453,
      "reward_std": 0.06838917918503284,
      "reward_std_all": 0.06605757400393486,
      "rewards-max/forensics_sinkhorn_reward": 0.94291752576828,
      "rewards-max/forensics_strict_edge_reward": 0.3893759995698929,
      "rewards-max/forensics_tiou_reward": 0.8966788351535797,
      "rewards/forensics_sinkhorn_reward": 0.6676622629165649,
      "rewards/forensics_strict_edge_reward": 0.02043059840798378,
      "rewards/forensics_tiou_reward": 0.3562321960926056,
      "step": 270
    },
    {
      "completion_length": 20.453125,
      "epoch": 1.1338912133891212,
      "grad_norm": 5.811721324920654,
      "kl": 0.275390625,
      "learning_rate": 7.175732217573222e-07,
      "loss": -0.0201,
      "reward": 0.7288406789302826,
      "reward_std": 0.04638919234275818,
      "reward_std_all": 0.061204537749290466,
      "rewards-max/forensics_sinkhorn_reward": 0.9687959551811218,
      "rewards-max/forensics_strict_edge_reward": 0.5366946756839752,
      "rewards-max/forensics_tiou_reward": 0.950840413570404,
      "rewards/forensics_sinkhorn_reward": 0.6769796311855316,
      "rewards/forensics_strict_edge_reward": 0.05168277584016323,
      "rewards/forensics_tiou_reward": 0.40459395945072174,
      "step": 271
    },
    {
      "completion_length": 20.15625,
      "epoch": 1.1380753138075315,
      "grad_norm": 4.937432765960693,
      "kl": 0.19580078125,
      "learning_rate": 7.165271966527197e-07,
      "loss": 0.0127,
      "reward": 0.7467776536941528,
      "reward_std": 0.05923897959291935,
      "reward_std_all": 0.06175328977406025,
      "rewards-max/forensics_sinkhorn_reward": 0.9692373871803284,
      "rewards-max/forensics_strict_edge_reward": 0.5758009254932404,
      "rewards-max/forensics_tiou_reward": 0.9425206184387207,
      "rewards/forensics_sinkhorn_reward": 0.7045761346817017,
      "rewards/forensics_strict_edge_reward": 0.03498368104919791,
      "rewards/forensics_tiou_reward": 0.4400661587715149,
      "step": 272
    },
    {
      "completion_length": 21.046875,
      "epoch": 1.1422594142259415,
      "grad_norm": 4.817264556884766,
      "kl": 0.25830078125,
      "learning_rate": 7.154811715481172e-07,
      "loss": 0.0552,
      "reward": 0.7358517348766327,
      "reward_std": 0.037537371506914496,
      "reward_std_all": 0.054136719554662704,
      "rewards-max/forensics_sinkhorn_reward": 0.9643257558345795,
      "rewards-max/forensics_strict_edge_reward": 0.2821265831589699,
      "rewards-max/forensics_tiou_reward": 0.9246900975704193,
      "rewards/forensics_sinkhorn_reward": 0.6973925232887268,
      "rewards/forensics_strict_edge_reward": 0.025827436475083232,
      "rewards/forensics_tiou_reward": 0.4589350074529648,
      "step": 273
    },
    {
      "completion_length": 20.078125,
      "epoch": 1.1464435146443515,
      "grad_norm": 4.823574066162109,
      "kl": 0.3271484375,
      "learning_rate": 7.144351464435146e-07,
      "loss": -0.0382,
      "reward": 0.7738975882530212,
      "reward_std": 0.042413657531142235,
      "reward_std_all": 0.04073901288211346,
      "rewards-max/forensics_sinkhorn_reward": 0.978325605392456,
      "rewards-max/forensics_strict_edge_reward": 0.29148174822330475,
      "rewards-max/forensics_tiou_reward": 0.9676380157470703,
      "rewards/forensics_sinkhorn_reward": 0.745051920413971,
      "rewards/forensics_strict_edge_reward": 0.01606342615559697,
      "rewards/forensics_tiou_reward": 0.5011709630489349,
      "step": 274
    },
    {
      "completion_length": 20.296875,
      "epoch": 1.1506276150627615,
      "grad_norm": 5.411806583404541,
      "kl": 0.328125,
      "learning_rate": 7.13389121338912e-07,
      "loss": -0.0141,
      "reward": 0.7114143669605255,
      "reward_std": 0.07801027595996857,
      "reward_std_all": 0.08675945177674294,
      "rewards-max/forensics_sinkhorn_reward": 0.9246987402439117,
      "rewards-max/forensics_strict_edge_reward": 0.3218911364674568,
      "rewards-max/forensics_tiou_reward": 0.8432418406009674,
      "rewards/forensics_sinkhorn_reward": 0.666871041059494,
      "rewards/forensics_strict_edge_reward": 0.025795846711844206,
      "rewards/forensics_tiou_reward": 0.4172651171684265,
      "step": 275
    },
    {
      "completion_length": 20.75,
      "epoch": 1.1548117154811715,
      "grad_norm": 5.120550632476807,
      "kl": 0.2705078125,
      "learning_rate": 7.123430962343095e-07,
      "loss": 0.0322,
      "reward": 0.6976177096366882,
      "reward_std": 0.06993885524570942,
      "reward_std_all": 0.09019111096858978,
      "rewards-max/forensics_sinkhorn_reward": 0.9726499021053314,
      "rewards-max/forensics_strict_edge_reward": 0.6837711930274963,
      "rewards-max/forensics_tiou_reward": 0.9626931548118591,
      "rewards/forensics_sinkhorn_reward": 0.64806067943573,
      "rewards/forensics_strict_edge_reward": 0.040922945365309715,
      "rewards/forensics_tiou_reward": 0.410085067152977,
      "step": 276
    },
    {
      "completion_length": 22.171875,
      "epoch": 1.1589958158995817,
      "grad_norm": 5.308947563171387,
      "kl": 0.330078125,
      "learning_rate": 7.11297071129707e-07,
      "loss": 0.0966,
      "reward": 0.7161239385604858,
      "reward_std": 0.09318790584802628,
      "reward_std_all": 0.09326326102018356,
      "rewards-max/forensics_sinkhorn_reward": 0.9681923687458038,
      "rewards-max/forensics_strict_edge_reward": 0.282395675778389,
      "rewards-max/forensics_tiou_reward": 0.9432263374328613,
      "rewards/forensics_sinkhorn_reward": 0.6659786999225616,
      "rewards/forensics_strict_edge_reward": 0.03601190075278282,
      "rewards/forensics_tiou_reward": 0.44876401126384735,
      "step": 277
    },
    {
      "completion_length": 22.109375,
      "epoch": 1.1631799163179917,
      "grad_norm": 4.7549543380737305,
      "kl": 0.3310546875,
      "learning_rate": 7.102510460251046e-07,
      "loss": 0.1584,
      "reward": 0.6954726874828339,
      "reward_std": 0.10867515578866005,
      "reward_std_all": 0.11002407595515251,
      "rewards-max/forensics_sinkhorn_reward": 0.8710275292396545,
      "rewards-max/forensics_strict_edge_reward": 0.07157937821466476,
      "rewards-max/forensics_tiou_reward": 0.7232389450073242,
      "rewards/forensics_sinkhorn_reward": 0.6089811027050018,
      "rewards/forensics_strict_edge_reward": 0.00773655237571802,
      "rewards/forensics_tiou_reward": 0.3446768522262573,
      "step": 278
    },
    {
      "completion_length": 19.46875,
      "epoch": 1.1673640167364017,
      "grad_norm": 5.893594264984131,
      "kl": 0.30859375,
      "learning_rate": 7.092050209205021e-07,
      "loss": 0.1113,
      "reward": 0.7156315445899963,
      "reward_std": 0.045330699533224106,
      "reward_std_all": 0.06721262633800507,
      "rewards-max/forensics_sinkhorn_reward": 0.9295646250247955,
      "rewards-max/forensics_strict_edge_reward": 0.04409551480785012,
      "rewards-max/forensics_tiou_reward": 0.8445656597614288,
      "rewards/forensics_sinkhorn_reward": 0.6760548949241638,
      "rewards/forensics_strict_edge_reward": 0.002424579666694626,
      "rewards/forensics_tiou_reward": 0.40035396814346313,
      "step": 279
    },
    {
      "completion_length": 21.484375,
      "epoch": 1.1715481171548117,
      "grad_norm": 5.200986862182617,
      "kl": 0.3046875,
      "learning_rate": 7.081589958158996e-07,
      "loss": 0.1539,
      "reward": 0.7029594480991364,
      "reward_std": 0.09214338101446629,
      "reward_std_all": 0.10023186355829239,
      "rewards-max/forensics_sinkhorn_reward": 0.9405328333377838,
      "rewards-max/forensics_strict_edge_reward": 0.18362712860107422,
      "rewards-max/forensics_tiou_reward": 0.8811480700969696,
      "rewards/forensics_sinkhorn_reward": 0.6196843385696411,
      "rewards/forensics_strict_edge_reward": 0.02295078430324793,
      "rewards/forensics_tiou_reward": 0.3593141436576843,
      "step": 280
    },
    {
      "completion_length": 19.921875,
      "epoch": 1.1757322175732217,
      "grad_norm": 5.616764545440674,
      "kl": 0.27490234375,
      "learning_rate": 7.071129707112971e-07,
      "loss": 0.0484,
      "reward": 0.7084561586380005,
      "reward_std": 0.07175228744745255,
      "reward_std_all": 0.08426500670611858,
      "rewards-max/forensics_sinkhorn_reward": 0.9288599789142609,
      "rewards-max/forensics_strict_edge_reward": 0.23752502351999283,
      "rewards-max/forensics_tiou_reward": 0.8629685342311859,
      "rewards/forensics_sinkhorn_reward": 0.6380473971366882,
      "rewards/forensics_strict_edge_reward": 0.013521367684006691,
      "rewards/forensics_tiou_reward": 0.38656581938266754,
      "step": 281
    },
    {
      "completion_length": 21.9375,
      "epoch": 1.1799163179916319,
      "grad_norm": 4.969091415405273,
      "kl": 0.3642578125,
      "learning_rate": 7.060669456066946e-07,
      "loss": 0.1406,
      "reward": 0.737311065196991,
      "reward_std": 0.05721690412610769,
      "reward_std_all": 0.06293437629938126,
      "rewards-max/forensics_sinkhorn_reward": 0.9236908257007599,
      "rewards-max/forensics_strict_edge_reward": 0.5001444219669793,
      "rewards-max/forensics_tiou_reward": 0.831083744764328,
      "rewards/forensics_sinkhorn_reward": 0.6664744019508362,
      "rewards/forensics_strict_edge_reward": 0.07435987088501861,
      "rewards/forensics_tiou_reward": 0.3915198743343353,
      "step": 282
    },
    {
      "completion_length": 21.5,
      "epoch": 1.1841004184100419,
      "grad_norm": 6.475979804992676,
      "kl": 0.2939453125,
      "learning_rate": 7.05020920502092e-07,
      "loss": 0.146,
      "reward": 0.7580174803733826,
      "reward_std": 0.04635380022227764,
      "reward_std_all": 0.04862461891025305,
      "rewards-max/forensics_sinkhorn_reward": 0.9796597063541412,
      "rewards-max/forensics_strict_edge_reward": 0.31592947617173195,
      "rewards-max/forensics_tiou_reward": 0.9735823571681976,
      "rewards/forensics_sinkhorn_reward": 0.6888420283794403,
      "rewards/forensics_strict_edge_reward": 0.02199454791843891,
      "rewards/forensics_tiou_reward": 0.4365353584289551,
      "step": 283
    },
    {
      "completion_length": 19.375,
      "epoch": 1.1882845188284519,
      "grad_norm": 5.802641868591309,
      "kl": 0.30859375,
      "learning_rate": 7.039748953974895e-07,
      "loss": 0.145,
      "reward": 0.7531943321228027,
      "reward_std": 0.03761492110788822,
      "reward_std_all": 0.04737559612840414,
      "rewards-max/forensics_sinkhorn_reward": 0.9541741609573364,
      "rewards-max/forensics_strict_edge_reward": 0.5410424992442131,
      "rewards-max/forensics_tiou_reward": 0.921400398015976,
      "rewards/forensics_sinkhorn_reward": 0.6862848103046417,
      "rewards/forensics_strict_edge_reward": 0.040547013748437166,
      "rewards/forensics_tiou_reward": 0.38895589113235474,
      "step": 284
    },
    {
      "completion_length": 22.5,
      "epoch": 1.1924686192468619,
      "grad_norm": 6.934317588806152,
      "kl": 0.4248046875,
      "learning_rate": 7.029288702928869e-07,
      "loss": 0.1367,
      "reward": 0.689582109451294,
      "reward_std": 0.10366281121969223,
      "reward_std_all": 0.11256732791662216,
      "rewards-max/forensics_sinkhorn_reward": 0.9674770832061768,
      "rewards-max/forensics_strict_edge_reward": 0.6393096446990967,
      "rewards-max/forensics_tiou_reward": 0.9437437355518341,
      "rewards/forensics_sinkhorn_reward": 0.631547212600708,
      "rewards/forensics_strict_edge_reward": 0.047392044216394424,
      "rewards/forensics_tiou_reward": 0.4129645824432373,
      "step": 285
    },
    {
      "completion_length": 20.671875,
      "epoch": 1.196652719665272,
      "grad_norm": 4.668708801269531,
      "kl": 0.3076171875,
      "learning_rate": 7.018828451882844e-07,
      "loss": 0.0588,
      "reward": 0.6919618844985962,
      "reward_std": 0.07704031467437744,
      "reward_std_all": 0.0928926169872284,
      "rewards-max/forensics_sinkhorn_reward": 0.9783588647842407,
      "rewards-max/forensics_strict_edge_reward": 0.21820376813411713,
      "rewards-max/forensics_tiou_reward": 0.9639179408550262,
      "rewards/forensics_sinkhorn_reward": 0.6353910565376282,
      "rewards/forensics_strict_edge_reward": 0.017695868154987693,
      "rewards/forensics_tiou_reward": 0.3877592235803604,
      "step": 286
    },
    {
      "completion_length": 20.390625,
      "epoch": 1.200836820083682,
      "grad_norm": 5.409516334533691,
      "kl": 0.23779296875,
      "learning_rate": 7.00836820083682e-07,
      "loss": 0.0234,
      "reward": 0.6722516417503357,
      "reward_std": 0.09952875226736069,
      "reward_std_all": 0.11647244915366173,
      "rewards-max/forensics_sinkhorn_reward": 0.9278503656387329,
      "rewards-max/forensics_strict_edge_reward": 0.07334584556519985,
      "rewards-max/forensics_tiou_reward": 0.8448033034801483,
      "rewards/forensics_sinkhorn_reward": 0.6014326810836792,
      "rewards/forensics_strict_edge_reward": 0.004619751940481365,
      "rewards/forensics_tiou_reward": 0.31709985435009,
      "step": 287
    },
    {
      "completion_length": 20.234375,
      "epoch": 1.205020920502092,
      "grad_norm": 6.943659782409668,
      "kl": 0.3408203125,
      "learning_rate": 6.997907949790795e-07,
      "loss": -0.0443,
      "reward": 0.7598115801811218,
      "reward_std": 0.04154087044298649,
      "reward_std_all": 0.044507257640361786,
      "rewards-max/forensics_sinkhorn_reward": 0.960777997970581,
      "rewards-max/forensics_strict_edge_reward": 0.4827088415622711,
      "rewards-max/forensics_tiou_reward": 0.9341850280761719,
      "rewards/forensics_sinkhorn_reward": 0.7212287783622742,
      "rewards/forensics_strict_edge_reward": 0.03280920349061489,
      "rewards/forensics_tiou_reward": 0.4612586498260498,
      "step": 288
    },
    {
      "completion_length": 20.5625,
      "epoch": 1.209205020920502,
      "grad_norm": 7.673379421234131,
      "kl": 0.29736328125,
      "learning_rate": 6.98744769874477e-07,
      "loss": 0.0672,
      "reward": 0.7187427282333374,
      "reward_std": 0.06327681615948677,
      "reward_std_all": 0.07381679862737656,
      "rewards-max/forensics_sinkhorn_reward": 0.9730832278728485,
      "rewards-max/forensics_strict_edge_reward": 0.38056008517742157,
      "rewards-max/forensics_tiou_reward": 0.9546814560890198,
      "rewards/forensics_sinkhorn_reward": 0.6894802749156952,
      "rewards/forensics_strict_edge_reward": 0.018159614875912666,
      "rewards/forensics_tiou_reward": 0.4455071836709976,
      "step": 289
    },
    {
      "completion_length": 21.84375,
      "epoch": 1.213389121338912,
      "grad_norm": 4.404791831970215,
      "kl": 0.296875,
      "learning_rate": 6.976987447698745e-07,
      "loss": 0.1007,
      "reward": 0.72098708152771,
      "reward_std": 0.049562305212020874,
      "reward_std_all": 0.06503269635140896,
      "rewards-max/forensics_sinkhorn_reward": 0.9818319082260132,
      "rewards-max/forensics_strict_edge_reward": 0.3254753649234772,
      "rewards-max/forensics_tiou_reward": 0.9748597145080566,
      "rewards/forensics_sinkhorn_reward": 0.6574923098087311,
      "rewards/forensics_strict_edge_reward": 0.021513513289391994,
      "rewards/forensics_tiou_reward": 0.4178762882947922,
      "step": 290
    },
    {
      "completion_length": 20.375,
      "epoch": 1.217573221757322,
      "grad_norm": 6.0823140144348145,
      "kl": 0.2578125,
      "learning_rate": 6.966527196652719e-07,
      "loss": 0.0428,
      "reward": 0.7571783363819122,
      "reward_std": 0.06689701974391937,
      "reward_std_all": 0.06190577894449234,
      "rewards-max/forensics_sinkhorn_reward": 0.9618112742900848,
      "rewards-max/forensics_strict_edge_reward": 0.3391399891115725,
      "rewards-max/forensics_tiou_reward": 0.9336270093917847,
      "rewards/forensics_sinkhorn_reward": 0.7305716574192047,
      "rewards/forensics_strict_edge_reward": 0.040705637831706554,
      "rewards/forensics_tiou_reward": 0.48133715987205505,
      "step": 291
    },
    {
      "completion_length": 20.953125,
      "epoch": 1.2217573221757323,
      "grad_norm": 5.331935405731201,
      "kl": 0.265625,
      "learning_rate": 6.956066945606694e-07,
      "loss": 0.0828,
      "reward": 0.7507474422454834,
      "reward_std": 0.05874476209282875,
      "reward_std_all": 0.05853298306465149,
      "rewards-max/forensics_sinkhorn_reward": 0.9570826590061188,
      "rewards-max/forensics_strict_edge_reward": 0.22151389345526695,
      "rewards-max/forensics_tiou_reward": 0.9198703169822693,
      "rewards/forensics_sinkhorn_reward": 0.6920550465583801,
      "rewards/forensics_strict_edge_reward": 0.012812266824766994,
      "rewards/forensics_tiou_reward": 0.4421239346265793,
      "step": 292
    },
    {
      "completion_length": 20.90625,
      "epoch": 1.2259414225941423,
      "grad_norm": 5.839554786682129,
      "kl": 0.2958984375,
      "learning_rate": 6.945606694560669e-07,
      "loss": 0.0608,
      "reward": 0.7189038395881653,
      "reward_std": 0.10003599524497986,
      "reward_std_all": 0.09508595243096352,
      "rewards-max/forensics_sinkhorn_reward": 0.9633159041404724,
      "rewards-max/forensics_strict_edge_reward": 0.42878545820713043,
      "rewards-max/forensics_tiou_reward": 0.9371602237224579,
      "rewards/forensics_sinkhorn_reward": 0.6471402943134308,
      "rewards/forensics_strict_edge_reward": 0.03017742745578289,
      "rewards/forensics_tiou_reward": 0.4058590680360794,
      "step": 293
    },
    {
      "completion_length": 20.34375,
      "epoch": 1.2301255230125523,
      "grad_norm": 5.967104434967041,
      "kl": 0.302734375,
      "learning_rate": 6.935146443514644e-07,
      "loss": 0.1568,
      "reward": 0.7555884420871735,
      "reward_std": 0.046018531545996666,
      "reward_std_all": 0.0476690586656332,
      "rewards-max/forensics_sinkhorn_reward": 0.9375809133052826,
      "rewards-max/forensics_strict_edge_reward": 0.06630663760006428,
      "rewards-max/forensics_tiou_reward": 0.8664887845516205,
      "rewards/forensics_sinkhorn_reward": 0.6943409740924835,
      "rewards/forensics_strict_edge_reward": 0.0041724692564457655,
      "rewards/forensics_tiou_reward": 0.40775974094867706,
      "step": 294
    },
    {
      "completion_length": 20.6875,
      "epoch": 1.2343096234309623,
      "grad_norm": 4.7712578773498535,
      "kl": 0.294921875,
      "learning_rate": 6.924686192468618e-07,
      "loss": 0.0755,
      "reward": 0.6917470097541809,
      "reward_std": 0.051753558218479156,
      "reward_std_all": 0.0812096856534481,
      "rewards-max/forensics_sinkhorn_reward": 0.9670523107051849,
      "rewards-max/forensics_strict_edge_reward": 0.2392681771889329,
      "rewards-max/forensics_tiou_reward": 0.9438088834285736,
      "rewards/forensics_sinkhorn_reward": 0.6183327734470367,
      "rewards/forensics_strict_edge_reward": 0.011334611510392278,
      "rewards/forensics_tiou_reward": 0.3922661393880844,
      "step": 295
    },
    {
      "completion_length": 19.859375,
      "epoch": 1.2384937238493725,
      "grad_norm": 4.7310333251953125,
      "kl": 0.330078125,
      "learning_rate": 6.914225941422594e-07,
      "loss": 0.0718,
      "reward": 0.7575635313987732,
      "reward_std": 0.05146120674908161,
      "reward_std_all": 0.05079099163413048,
      "rewards-max/forensics_sinkhorn_reward": 0.9621972143650055,
      "rewards-max/forensics_strict_edge_reward": 0.32777850329875946,
      "rewards-max/forensics_tiou_reward": 0.9339486658573151,
      "rewards/forensics_sinkhorn_reward": 0.7140895426273346,
      "rewards/forensics_strict_edge_reward": 0.022600538097321987,
      "rewards/forensics_tiou_reward": 0.43596790730953217,
      "step": 296
    },
    {
      "completion_length": 21.25,
      "epoch": 1.2426778242677825,
      "grad_norm": 6.860005855560303,
      "kl": 0.4521484375,
      "learning_rate": 6.903765690376569e-07,
      "loss": 0.0184,
      "reward": 0.6833628416061401,
      "reward_std": 0.08702775835990906,
      "reward_std_all": 0.10281749814748764,
      "rewards-max/forensics_sinkhorn_reward": 0.9569408595561981,
      "rewards-max/forensics_strict_edge_reward": 0.3072382062673569,
      "rewards-max/forensics_tiou_reward": 0.922799825668335,
      "rewards/forensics_sinkhorn_reward": 0.6042942404747009,
      "rewards/forensics_strict_edge_reward": 0.018836085218936205,
      "rewards/forensics_tiou_reward": 0.3037617653608322,
      "step": 297
    },
    {
      "completion_length": 20.78125,
      "epoch": 1.2468619246861925,
      "grad_norm": 7.030130863189697,
      "kl": 0.2861328125,
      "learning_rate": 6.893305439330544e-07,
      "loss": 0.0682,
      "reward": 0.6868405640125275,
      "reward_std": 0.062466371804475784,
      "reward_std_all": 0.09029397740960121,
      "rewards-max/forensics_sinkhorn_reward": 0.9667437672615051,
      "rewards-max/forensics_strict_edge_reward": 0.3174005262553692,
      "rewards-max/forensics_tiou_reward": 0.9285447299480438,
      "rewards/forensics_sinkhorn_reward": 0.6381552517414093,
      "rewards/forensics_strict_edge_reward": 0.028787018032744527,
      "rewards/forensics_tiou_reward": 0.3903830349445343,
      "step": 298
    },
    {
      "completion_length": 22.265625,
      "epoch": 1.2510460251046025,
      "grad_norm": 5.2615532875061035,
      "kl": 0.27099609375,
      "learning_rate": 6.882845188284518e-07,
      "loss": 0.0656,
      "reward": 0.7472319602966309,
      "reward_std": 0.06345664337277412,
      "reward_std_all": 0.061917271465063095,
      "rewards-max/forensics_sinkhorn_reward": 0.9438942968845367,
      "rewards-max/forensics_strict_edge_reward": 0.4273863285779953,
      "rewards-max/forensics_tiou_reward": 0.8787153959274292,
      "rewards/forensics_sinkhorn_reward": 0.6656781733036041,
      "rewards/forensics_strict_edge_reward": 0.03559849224984646,
      "rewards/forensics_tiou_reward": 0.4403180480003357,
      "step": 299
    },
    {
      "completion_length": 21.25,
      "epoch": 1.2552301255230125,
      "grad_norm": 5.571683406829834,
      "kl": 0.31640625,
      "learning_rate": 6.872384937238493e-07,
      "loss": 0.1289,
      "reward": 0.7022594213485718,
      "reward_std": 0.048019133508205414,
      "reward_std_all": 0.07540664449334145,
      "rewards-max/forensics_sinkhorn_reward": 0.9484546482563019,
      "rewards-max/forensics_strict_edge_reward": 0.29233212023973465,
      "rewards-max/forensics_tiou_reward": 0.9066206216812134,
      "rewards/forensics_sinkhorn_reward": 0.5951856970787048,
      "rewards/forensics_strict_edge_reward": 0.013865047134459019,
      "rewards/forensics_tiou_reward": 0.30919376015663147,
      "step": 300
    },
    {
      "completion_length": 21.4375,
      "epoch": 1.2594142259414225,
      "grad_norm": 5.529702186584473,
      "kl": 0.32421875,
      "learning_rate": 6.861924686192468e-07,
      "loss": 0.038,
      "reward": 0.7326129972934723,
      "reward_std": 0.06115315482020378,
      "reward_std_all": 0.06525096669793129,
      "rewards-max/forensics_sinkhorn_reward": 0.9831356406211853,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9849620759487152,
      "rewards/forensics_sinkhorn_reward": 0.6498611867427826,
      "rewards/forensics_strict_edge_reward": 0.031638662330806255,
      "rewards/forensics_tiou_reward": 0.38330915570259094,
      "step": 301
    },
    {
      "completion_length": 21.28125,
      "epoch": 1.2635983263598327,
      "grad_norm": 5.129769325256348,
      "kl": 0.3173828125,
      "learning_rate": 6.851464435146443e-07,
      "loss": 0.1208,
      "reward": 0.7460393011569977,
      "reward_std": 0.07854705303907394,
      "reward_std_all": 0.0734499879181385,
      "rewards-max/forensics_sinkhorn_reward": 0.9730245172977448,
      "rewards-max/forensics_strict_edge_reward": 0.34696415439248085,
      "rewards-max/forensics_tiou_reward": 0.9585398137569427,
      "rewards/forensics_sinkhorn_reward": 0.6749107837677002,
      "rewards/forensics_strict_edge_reward": 0.022300497628748417,
      "rewards/forensics_tiou_reward": 0.45180293917655945,
      "step": 302
    },
    {
      "completion_length": 21.96875,
      "epoch": 1.2677824267782427,
      "grad_norm": 6.377272129058838,
      "kl": 0.3427734375,
      "learning_rate": 6.841004184100418e-07,
      "loss": 0.2079,
      "reward": 0.6833070516586304,
      "reward_std": 0.11725638434290886,
      "reward_std_all": 0.11747675761580467,
      "rewards-max/forensics_sinkhorn_reward": 0.9631252884864807,
      "rewards-max/forensics_strict_edge_reward": 0.5081118941307068,
      "rewards-max/forensics_tiou_reward": 0.937112033367157,
      "rewards/forensics_sinkhorn_reward": 0.6237468421459198,
      "rewards/forensics_strict_edge_reward": 0.03924329672008753,
      "rewards/forensics_tiou_reward": 0.37655380368232727,
      "step": 303
    },
    {
      "completion_length": 21.09375,
      "epoch": 1.2719665271966527,
      "grad_norm": 6.458245277404785,
      "kl": 0.3330078125,
      "learning_rate": 6.830543933054394e-07,
      "loss": -0.022,
      "reward": 0.7517873644828796,
      "reward_std": 0.07746486738324165,
      "reward_std_all": 0.07105372101068497,
      "rewards-max/forensics_sinkhorn_reward": 0.9708572328090668,
      "rewards-max/forensics_strict_edge_reward": 0.5950735807418823,
      "rewards-max/forensics_tiou_reward": 0.9572498798370361,
      "rewards/forensics_sinkhorn_reward": 0.6973008215427399,
      "rewards/forensics_strict_edge_reward": 0.047552487812936306,
      "rewards/forensics_tiou_reward": 0.48613592982292175,
      "step": 304
    },
    {
      "completion_length": 20.25,
      "epoch": 1.2761506276150627,
      "grad_norm": 5.511967182159424,
      "kl": 0.3681640625,
      "learning_rate": 6.820083682008368e-07,
      "loss": 0.1425,
      "reward": 0.6778889000415802,
      "reward_std": 0.09900233522057533,
      "reward_std_all": 0.11688397824764252,
      "rewards-max/forensics_sinkhorn_reward": 0.9803718328475952,
      "rewards-max/forensics_strict_edge_reward": 0.6557035446166992,
      "rewards-max/forensics_tiou_reward": 0.9761851727962494,
      "rewards/forensics_sinkhorn_reward": 0.6116900146007538,
      "rewards/forensics_strict_edge_reward": 0.03637626115232706,
      "rewards/forensics_tiou_reward": 0.35539279878139496,
      "step": 305
    },
    {
      "completion_length": 20.203125,
      "epoch": 1.280334728033473,
      "grad_norm": 5.910262107849121,
      "kl": 0.298828125,
      "learning_rate": 6.809623430962343e-07,
      "loss": 0.1586,
      "reward": 0.7084982097148895,
      "reward_std": 0.058267127722501755,
      "reward_std_all": 0.07916367799043655,
      "rewards-max/forensics_sinkhorn_reward": 0.9686716496944427,
      "rewards-max/forensics_strict_edge_reward": 0.3420734480023384,
      "rewards-max/forensics_tiou_reward": 0.945890873670578,
      "rewards/forensics_sinkhorn_reward": 0.6659131050109863,
      "rewards/forensics_strict_edge_reward": 0.019409795757383108,
      "rewards/forensics_tiou_reward": 0.40642550587654114,
      "step": 306
    },
    {
      "completion_length": 22.546875,
      "epoch": 1.284518828451883,
      "grad_norm": 6.111764430999756,
      "kl": 0.25341796875,
      "learning_rate": 6.799163179916317e-07,
      "loss": 0.0403,
      "reward": 0.73848956823349,
      "reward_std": 0.08970540016889572,
      "reward_std_all": 0.08318790420889854,
      "rewards-max/forensics_sinkhorn_reward": 0.9653835892677307,
      "rewards-max/forensics_strict_edge_reward": 0.3977031335234642,
      "rewards-max/forensics_tiou_reward": 0.9420799016952515,
      "rewards/forensics_sinkhorn_reward": 0.6589479446411133,
      "rewards/forensics_strict_edge_reward": 0.028378714807331562,
      "rewards/forensics_tiou_reward": 0.428754523396492,
      "step": 307
    },
    {
      "completion_length": 20.59375,
      "epoch": 1.288702928870293,
      "grad_norm": 4.3902411460876465,
      "kl": 0.337890625,
      "learning_rate": 6.788702928870292e-07,
      "loss": 0.0891,
      "reward": 0.7305803894996643,
      "reward_std": 0.04291406832635403,
      "reward_std_all": 0.05986184999346733,
      "rewards-max/forensics_sinkhorn_reward": 0.9835834503173828,
      "rewards-max/forensics_strict_edge_reward": 0.16330911964178085,
      "rewards-max/forensics_tiou_reward": 0.984925240278244,
      "rewards/forensics_sinkhorn_reward": 0.6643323302268982,
      "rewards/forensics_strict_edge_reward": 0.009930451633408666,
      "rewards/forensics_tiou_reward": 0.3908986747264862,
      "step": 308
    },
    {
      "completion_length": 20.578125,
      "epoch": 1.292887029288703,
      "grad_norm": 6.543727874755859,
      "kl": 0.4853515625,
      "learning_rate": 6.778242677824267e-07,
      "loss": 0.0832,
      "reward": 0.7283049821853638,
      "reward_std": 0.07112754881381989,
      "reward_std_all": 0.07235599681735039,
      "rewards-max/forensics_sinkhorn_reward": 0.960137277841568,
      "rewards-max/forensics_strict_edge_reward": 0.42061087489128113,
      "rewards-max/forensics_tiou_reward": 0.9282712638378143,
      "rewards/forensics_sinkhorn_reward": 0.6624591946601868,
      "rewards/forensics_strict_edge_reward": 0.021041087806224823,
      "rewards/forensics_tiou_reward": 0.4426434636116028,
      "step": 309
    },
    {
      "completion_length": 20.25,
      "epoch": 1.297071129707113,
      "grad_norm": 4.964821815490723,
      "kl": 0.3193359375,
      "learning_rate": 6.767782426778242e-07,
      "loss": -0.0055,
      "reward": 0.729735791683197,
      "reward_std": 0.07737122103571892,
      "reward_std_all": 0.07788483798503876,
      "rewards-max/forensics_sinkhorn_reward": 0.9729414284229279,
      "rewards-max/forensics_strict_edge_reward": 0.2672666907310486,
      "rewards-max/forensics_tiou_reward": 0.9507744312286377,
      "rewards/forensics_sinkhorn_reward": 0.7067297697067261,
      "rewards/forensics_strict_edge_reward": 0.022594827227294445,
      "rewards/forensics_tiou_reward": 0.4818980246782303,
      "step": 310
    },
    {
      "completion_length": 19.484375,
      "epoch": 1.301255230125523,
      "grad_norm": 5.7502665519714355,
      "kl": 0.400390625,
      "learning_rate": 6.757322175732217e-07,
      "loss": 0.0768,
      "reward": 0.7538015246391296,
      "reward_std": 0.06740200705826283,
      "reward_std_all": 0.0633252989500761,
      "rewards-max/forensics_sinkhorn_reward": 0.9471302330493927,
      "rewards-max/forensics_strict_edge_reward": 0.3330773785710335,
      "rewards-max/forensics_tiou_reward": 0.9042232036590576,
      "rewards/forensics_sinkhorn_reward": 0.7187202870845795,
      "rewards/forensics_strict_edge_reward": 0.019347605295479298,
      "rewards/forensics_tiou_reward": 0.46320129930973053,
      "step": 311
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.3054393305439331,
      "grad_norm": 6.908195495605469,
      "kl": 0.3212890625,
      "learning_rate": 6.746861924686192e-07,
      "loss": 0.0927,
      "reward": 0.6680946946144104,
      "reward_std": 0.09473831951618195,
      "reward_std_all": 0.11062188073992729,
      "rewards-max/forensics_sinkhorn_reward": 0.8996167480945587,
      "rewards-max/forensics_strict_edge_reward": 0.07242675591260195,
      "rewards-max/forensics_tiou_reward": 0.7866577208042145,
      "rewards/forensics_sinkhorn_reward": 0.553074836730957,
      "rewards/forensics_strict_edge_reward": 0.0049136035377159715,
      "rewards/forensics_tiou_reward": 0.290701299905777,
      "step": 312
    },
    {
      "completion_length": 21.015625,
      "epoch": 1.3096234309623431,
      "grad_norm": 5.475022792816162,
      "kl": 0.3662109375,
      "learning_rate": 6.736401673640168e-07,
      "loss": 0.2811,
      "reward": 0.7459441125392914,
      "reward_std": 0.07762224785983562,
      "reward_std_all": 0.07212866470217705,
      "rewards-max/forensics_sinkhorn_reward": 0.9816945195198059,
      "rewards-max/forensics_strict_edge_reward": 0.44343748688697815,
      "rewards-max/forensics_tiou_reward": 0.9806602299213409,
      "rewards/forensics_sinkhorn_reward": 0.658801257610321,
      "rewards/forensics_strict_edge_reward": 0.05264921858906746,
      "rewards/forensics_tiou_reward": 0.4826244115829468,
      "step": 313
    },
    {
      "completion_length": 19.6875,
      "epoch": 1.3138075313807531,
      "grad_norm": 6.151768207550049,
      "kl": 0.322265625,
      "learning_rate": 6.725941422594143e-07,
      "loss": 0.1037,
      "reward": 0.7682581841945648,
      "reward_std": 0.044314319267868996,
      "reward_std_all": 0.043012285605072975,
      "rewards-max/forensics_sinkhorn_reward": 0.9462916553020477,
      "rewards-max/forensics_strict_edge_reward": 0.26853641122579575,
      "rewards-max/forensics_tiou_reward": 0.8992149829864502,
      "rewards/forensics_sinkhorn_reward": 0.7576715648174286,
      "rewards/forensics_strict_edge_reward": 0.02383694052696228,
      "rewards/forensics_tiou_reward": 0.5424575805664062,
      "step": 314
    },
    {
      "completion_length": 20.125,
      "epoch": 1.3179916317991631,
      "grad_norm": 6.935332298278809,
      "kl": 0.2578125,
      "learning_rate": 6.715481171548117e-07,
      "loss": 0.0414,
      "reward": 0.7575020492076874,
      "reward_std": 0.03853140398859978,
      "reward_std_all": 0.04303992912173271,
      "rewards-max/forensics_sinkhorn_reward": 0.9525054395198822,
      "rewards-max/forensics_strict_edge_reward": 0.3443178366869688,
      "rewards-max/forensics_tiou_reward": 0.8969803154468536,
      "rewards/forensics_sinkhorn_reward": 0.7127469182014465,
      "rewards/forensics_strict_edge_reward": 0.02792320610024035,
      "rewards/forensics_tiou_reward": 0.4667501300573349,
      "step": 315
    },
    {
      "completion_length": 20.875,
      "epoch": 1.3221757322175733,
      "grad_norm": 6.597142696380615,
      "kl": 0.302734375,
      "learning_rate": 6.705020920502091e-07,
      "loss": 0.0784,
      "reward": 0.7237750887870789,
      "reward_std": 0.0663918536156416,
      "reward_std_all": 0.07297496870160103,
      "rewards-max/forensics_sinkhorn_reward": 0.9281657934188843,
      "rewards-max/forensics_strict_edge_reward": 0.34279249608516693,
      "rewards-max/forensics_tiou_reward": 0.854468435049057,
      "rewards/forensics_sinkhorn_reward": 0.670475959777832,
      "rewards/forensics_strict_edge_reward": 0.032728907419368625,
      "rewards/forensics_tiou_reward": 0.46926388144493103,
      "step": 316
    },
    {
      "completion_length": 21.5,
      "epoch": 1.3263598326359833,
      "grad_norm": 5.202786922454834,
      "kl": 0.404296875,
      "learning_rate": 6.694560669456066e-07,
      "loss": 0.1206,
      "reward": 0.739029049873352,
      "reward_std": 0.07363832741975784,
      "reward_std_all": 0.07193014211952686,
      "rewards-max/forensics_sinkhorn_reward": 0.9724274277687073,
      "rewards-max/forensics_strict_edge_reward": 0.6011503338813782,
      "rewards-max/forensics_tiou_reward": 0.9589043259620667,
      "rewards/forensics_sinkhorn_reward": 0.6700388193130493,
      "rewards/forensics_strict_edge_reward": 0.04086586460471153,
      "rewards/forensics_tiou_reward": 0.4364643692970276,
      "step": 317
    },
    {
      "completion_length": 19.6875,
      "epoch": 1.3305439330543933,
      "grad_norm": 4.5457258224487305,
      "kl": 0.30859375,
      "learning_rate": 6.684100418410041e-07,
      "loss": 0.0823,
      "reward": 0.7352396547794342,
      "reward_std": 0.037429291754961014,
      "reward_std_all": 0.05756979249417782,
      "rewards-max/forensics_sinkhorn_reward": 0.970054417848587,
      "rewards-max/forensics_strict_edge_reward": 0.4697008728981018,
      "rewards-max/forensics_tiou_reward": 0.9512118101119995,
      "rewards/forensics_sinkhorn_reward": 0.6980648636817932,
      "rewards/forensics_strict_edge_reward": 0.032781449146568775,
      "rewards/forensics_tiou_reward": 0.42899972200393677,
      "step": 318
    },
    {
      "completion_length": 20.234375,
      "epoch": 1.3347280334728033,
      "grad_norm": 6.000026226043701,
      "kl": 0.4111328125,
      "learning_rate": 6.673640167364016e-07,
      "loss": 0.0813,
      "reward": 0.7340326309204102,
      "reward_std": 0.048289031721651554,
      "reward_std_all": 0.061289943754673004,
      "rewards-max/forensics_sinkhorn_reward": 0.9288490414619446,
      "rewards-max/forensics_strict_edge_reward": 0.23604796826839447,
      "rewards-max/forensics_tiou_reward": 0.8543278574943542,
      "rewards/forensics_sinkhorn_reward": 0.6637733280658722,
      "rewards/forensics_strict_edge_reward": 0.010000124108046293,
      "rewards/forensics_tiou_reward": 0.38839341700077057,
      "step": 319
    },
    {
      "completion_length": 20.546875,
      "epoch": 1.3389121338912133,
      "grad_norm": 6.16050386428833,
      "kl": 0.24951171875,
      "learning_rate": 6.663179916317991e-07,
      "loss": 0.1443,
      "reward": 0.7257636189460754,
      "reward_std": 0.056109463796019554,
      "reward_std_all": 0.07116309925913811,
      "rewards-max/forensics_sinkhorn_reward": 0.9832309782505035,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9836986660957336,
      "rewards/forensics_sinkhorn_reward": 0.6775293052196503,
      "rewards/forensics_strict_edge_reward": 0.045782670378685,
      "rewards/forensics_tiou_reward": 0.4414111524820328,
      "step": 320
    },
    {
      "completion_length": 20.484375,
      "epoch": 1.3430962343096233,
      "grad_norm": 4.7137579917907715,
      "kl": 0.29296875,
      "learning_rate": 6.652719665271967e-07,
      "loss": -0.055,
      "reward": 0.7733777463436127,
      "reward_std": 0.02845308557152748,
      "reward_std_all": 0.03176207561045885,
      "rewards-max/forensics_sinkhorn_reward": 0.9774341881275177,
      "rewards-max/forensics_strict_edge_reward": 0.45442917943000793,
      "rewards-max/forensics_tiou_reward": 0.9674332141876221,
      "rewards/forensics_sinkhorn_reward": 0.7390527427196503,
      "rewards/forensics_strict_edge_reward": 0.027589570730924606,
      "rewards/forensics_tiou_reward": 0.4522760808467865,
      "step": 321
    },
    {
      "completion_length": 19.96875,
      "epoch": 1.3472803347280335,
      "grad_norm": 7.949156761169434,
      "kl": 0.3662109375,
      "learning_rate": 6.642259414225942e-07,
      "loss": 0.0779,
      "reward": 0.7105337381362915,
      "reward_std": 0.08389031887054443,
      "reward_std_all": 0.08908619359135628,
      "rewards-max/forensics_sinkhorn_reward": 0.9821226894855499,
      "rewards-max/forensics_strict_edge_reward": 0.38767390325665474,
      "rewards-max/forensics_tiou_reward": 0.9759470224380493,
      "rewards/forensics_sinkhorn_reward": 0.6783856749534607,
      "rewards/forensics_strict_edge_reward": 0.01981935719959438,
      "rewards/forensics_tiou_reward": 0.4188491404056549,
      "step": 322
    },
    {
      "completion_length": 19.6875,
      "epoch": 1.3514644351464435,
      "grad_norm": 5.794936180114746,
      "kl": 0.310546875,
      "learning_rate": 6.631799163179917e-07,
      "loss": -0.0063,
      "reward": 0.6986989676952362,
      "reward_std": 0.02718257624655962,
      "reward_std_all": 0.06691677263006568,
      "rewards-max/forensics_sinkhorn_reward": 0.9764258563518524,
      "rewards-max/forensics_strict_edge_reward": 0.2945442795753479,
      "rewards-max/forensics_tiou_reward": 0.9648978114128113,
      "rewards/forensics_sinkhorn_reward": 0.6593092083930969,
      "rewards/forensics_strict_edge_reward": 0.01357288658618927,
      "rewards/forensics_tiou_reward": 0.41728852689266205,
      "step": 323
    },
    {
      "completion_length": 19.359375,
      "epoch": 1.3556485355648535,
      "grad_norm": 5.485504627227783,
      "kl": 0.4208984375,
      "learning_rate": 6.621338912133892e-07,
      "loss": -0.024,
      "reward": 0.7957438826560974,
      "reward_std": 0.013565458706580102,
      "reward_std_all": 0.01419946737587452,
      "rewards-max/forensics_sinkhorn_reward": 0.9753878712654114,
      "rewards-max/forensics_strict_edge_reward": 0.2692512273788452,
      "rewards-max/forensics_tiou_reward": 0.9588492810726166,
      "rewards/forensics_sinkhorn_reward": 0.8010739684104919,
      "rewards/forensics_strict_edge_reward": 0.026666104327887297,
      "rewards/forensics_tiou_reward": 0.5619127154350281,
      "step": 324
    },
    {
      "completion_length": 20.125,
      "epoch": 1.3598326359832635,
      "grad_norm": 6.727545738220215,
      "kl": 0.384765625,
      "learning_rate": 6.610878661087865e-07,
      "loss": 0.0256,
      "reward": 0.7440242171287537,
      "reward_std": 0.04323656111955643,
      "reward_std_all": 0.05361815728247166,
      "rewards-max/forensics_sinkhorn_reward": 0.9753364026546478,
      "rewards-max/forensics_strict_edge_reward": 0.4935835525393486,
      "rewards-max/forensics_tiou_reward": 0.9597883522510529,
      "rewards/forensics_sinkhorn_reward": 0.6856612265110016,
      "rewards/forensics_strict_edge_reward": 0.029298210982233286,
      "rewards/forensics_tiou_reward": 0.42166899144649506,
      "step": 325
    },
    {
      "completion_length": 20.515625,
      "epoch": 1.3640167364016738,
      "grad_norm": 8.83599853515625,
      "kl": 0.34375,
      "learning_rate": 6.60041841004184e-07,
      "loss": -0.005,
      "reward": 0.7500342130661011,
      "reward_std": 0.012550319777801633,
      "reward_std_all": 0.035230450332164764,
      "rewards-max/forensics_sinkhorn_reward": 0.9673956632614136,
      "rewards-max/forensics_strict_edge_reward": 0.34384459257125854,
      "rewards-max/forensics_tiou_reward": 0.9428177773952484,
      "rewards/forensics_sinkhorn_reward": 0.6898816525936127,
      "rewards/forensics_strict_edge_reward": 0.021375369280576706,
      "rewards/forensics_tiou_reward": 0.41276177763938904,
      "step": 326
    },
    {
      "completion_length": 21.046875,
      "epoch": 1.3682008368200838,
      "grad_norm": 5.394491195678711,
      "kl": 0.4482421875,
      "learning_rate": 6.589958158995815e-07,
      "loss": 0.1251,
      "reward": 0.6343050599098206,
      "reward_std": 0.10531711205840111,
      "reward_std_all": 0.1327713206410408,
      "rewards-max/forensics_sinkhorn_reward": 0.9156837463378906,
      "rewards-max/forensics_strict_edge_reward": 0.004623777931556106,
      "rewards-max/forensics_tiou_reward": 0.8100517988204956,
      "rewards/forensics_sinkhorn_reward": 0.5471821129322052,
      "rewards/forensics_strict_edge_reward": 0.0005519441037904471,
      "rewards/forensics_tiou_reward": 0.2941691428422928,
      "step": 327
    },
    {
      "completion_length": 23.078125,
      "epoch": 1.3723849372384938,
      "grad_norm": 5.109241485595703,
      "kl": 0.4755859375,
      "learning_rate": 6.57949790794979e-07,
      "loss": 0.108,
      "reward": 0.7466883361339569,
      "reward_std": 0.058062536641955376,
      "reward_std_all": 0.058207325637340546,
      "rewards-max/forensics_sinkhorn_reward": 0.9358937740325928,
      "rewards-max/forensics_strict_edge_reward": 0.14661087095737457,
      "rewards-max/forensics_tiou_reward": 0.8493654727935791,
      "rewards/forensics_sinkhorn_reward": 0.6865655779838562,
      "rewards/forensics_strict_edge_reward": 0.01298220083117485,
      "rewards/forensics_tiou_reward": 0.44069693982601166,
      "step": 328
    },
    {
      "completion_length": 21.015625,
      "epoch": 1.3765690376569037,
      "grad_norm": 5.67556095123291,
      "kl": 0.3544921875,
      "learning_rate": 6.569037656903765e-07,
      "loss": 0.0843,
      "reward": 0.7142748534679413,
      "reward_std": 0.07961460761725903,
      "reward_std_all": 0.08414532616734505,
      "rewards-max/forensics_sinkhorn_reward": 0.9499431252479553,
      "rewards-max/forensics_strict_edge_reward": 0.12697420455515385,
      "rewards-max/forensics_tiou_reward": 0.8939896523952484,
      "rewards/forensics_sinkhorn_reward": 0.6463456153869629,
      "rewards/forensics_strict_edge_reward": 0.009177779080346227,
      "rewards/forensics_tiou_reward": 0.38923509418964386,
      "step": 329
    },
    {
      "completion_length": 21.25,
      "epoch": 1.3807531380753137,
      "grad_norm": 8.164499282836914,
      "kl": 0.3076171875,
      "learning_rate": 6.558577405857741e-07,
      "loss": 0.0667,
      "reward": 0.7576728165149689,
      "reward_std": 0.0512995682656765,
      "reward_std_all": 0.050735730677843094,
      "rewards-max/forensics_sinkhorn_reward": 0.9739361107349396,
      "rewards-max/forensics_strict_edge_reward": 0.2533908262848854,
      "rewards-max/forensics_tiou_reward": 0.9438464939594269,
      "rewards/forensics_sinkhorn_reward": 0.7066149413585663,
      "rewards/forensics_strict_edge_reward": 0.04131447523832321,
      "rewards/forensics_tiou_reward": 0.5058133453130722,
      "step": 330
    },
    {
      "completion_length": 22.171875,
      "epoch": 1.3849372384937237,
      "grad_norm": 23.917482376098633,
      "kl": 0.326171875,
      "learning_rate": 6.548117154811716e-07,
      "loss": 0.1813,
      "reward": 0.6975154280662537,
      "reward_std": 0.0634789876639843,
      "reward_std_all": 0.0849098190665245,
      "rewards-max/forensics_sinkhorn_reward": 0.9257499873638153,
      "rewards-max/forensics_strict_edge_reward": 0.10880552046000957,
      "rewards-max/forensics_tiou_reward": 0.8533532619476318,
      "rewards/forensics_sinkhorn_reward": 0.5871744751930237,
      "rewards/forensics_strict_edge_reward": 0.0064210264245048165,
      "rewards/forensics_tiou_reward": 0.33856940269470215,
      "step": 331
    },
    {
      "completion_length": 20.375,
      "epoch": 1.389121338912134,
      "grad_norm": 4.935762405395508,
      "kl": 0.3544921875,
      "learning_rate": 6.537656903765691e-07,
      "loss": 0.0706,
      "reward": 0.7255370020866394,
      "reward_std": 0.05906838923692703,
      "reward_std_all": 0.07139801979064941,
      "rewards-max/forensics_sinkhorn_reward": 0.9809918403625488,
      "rewards-max/forensics_strict_edge_reward": 0.692571371793747,
      "rewards-max/forensics_tiou_reward": 0.9781451225280762,
      "rewards/forensics_sinkhorn_reward": 0.6542418897151947,
      "rewards/forensics_strict_edge_reward": 0.0352795859798789,
      "rewards/forensics_tiou_reward": 0.3850897550582886,
      "step": 332
    },
    {
      "completion_length": 20.59375,
      "epoch": 1.393305439330544,
      "grad_norm": 5.050448417663574,
      "kl": 0.302734375,
      "learning_rate": 6.527196652719665e-07,
      "loss": 0.0694,
      "reward": 0.7477313280105591,
      "reward_std": 0.060213714838027954,
      "reward_std_all": 0.06026175245642662,
      "rewards-max/forensics_sinkhorn_reward": 0.9527264833450317,
      "rewards-max/forensics_strict_edge_reward": 0.5206717550754547,
      "rewards-max/forensics_tiou_reward": 0.9085529446601868,
      "rewards/forensics_sinkhorn_reward": 0.6675804257392883,
      "rewards/forensics_strict_edge_reward": 0.029296133667230606,
      "rewards/forensics_tiou_reward": 0.40639808773994446,
      "step": 333
    },
    {
      "completion_length": 22.625,
      "epoch": 1.397489539748954,
      "grad_norm": 5.14325475692749,
      "kl": 0.2607421875,
      "learning_rate": 6.51673640167364e-07,
      "loss": 0.0718,
      "reward": 0.7338097393512726,
      "reward_std": 0.0976376011967659,
      "reward_std_all": 0.08983501046895981,
      "rewards-max/forensics_sinkhorn_reward": 0.9563211500644684,
      "rewards-max/forensics_strict_edge_reward": 0.29233212023973465,
      "rewards-max/forensics_tiou_reward": 0.9115324020385742,
      "rewards/forensics_sinkhorn_reward": 0.6761124730110168,
      "rewards/forensics_strict_edge_reward": 0.018567598424851894,
      "rewards/forensics_tiou_reward": 0.4703802317380905,
      "step": 334
    },
    {
      "completion_length": 22.546875,
      "epoch": 1.401673640167364,
      "grad_norm": 6.576317310333252,
      "kl": 0.3046875,
      "learning_rate": 6.506276150627614e-07,
      "loss": -0.0115,
      "reward": 0.7379060089588165,
      "reward_std": 0.07075753062963486,
      "reward_std_all": 0.06978487223386765,
      "rewards-max/forensics_sinkhorn_reward": 0.9749369323253632,
      "rewards-max/forensics_strict_edge_reward": 0.1802845485508442,
      "rewards-max/forensics_tiou_reward": 0.9548335671424866,
      "rewards/forensics_sinkhorn_reward": 0.6434165239334106,
      "rewards/forensics_strict_edge_reward": 0.01491499156691134,
      "rewards/forensics_tiou_reward": 0.410367488861084,
      "step": 335
    },
    {
      "completion_length": 20.484375,
      "epoch": 1.4058577405857742,
      "grad_norm": 5.7756829261779785,
      "kl": 0.3173828125,
      "learning_rate": 6.495815899581589e-07,
      "loss": 0.2125,
      "reward": 0.7116163074970245,
      "reward_std": 0.06507093086838722,
      "reward_std_all": 0.07775519415736198,
      "rewards-max/forensics_sinkhorn_reward": 0.9373482763767242,
      "rewards-max/forensics_strict_edge_reward": 0.1376016093418002,
      "rewards-max/forensics_tiou_reward": 0.8542213141918182,
      "rewards/forensics_sinkhorn_reward": 0.6090943217277527,
      "rewards/forensics_strict_edge_reward": 0.011051669425796717,
      "rewards/forensics_tiou_reward": 0.3465346544981003,
      "step": 336
    },
    {
      "completion_length": 21.171875,
      "epoch": 1.4100418410041842,
      "grad_norm": 6.921660423278809,
      "kl": 0.51953125,
      "learning_rate": 6.485355648535564e-07,
      "loss": 0.0741,
      "reward": 0.7673178613185883,
      "reward_std": 0.039484485052526,
      "reward_std_all": 0.040349071845412254,
      "rewards-max/forensics_sinkhorn_reward": 0.9112433791160583,
      "rewards-max/forensics_strict_edge_reward": 0.0021028527116868645,
      "rewards-max/forensics_tiou_reward": 0.7947789430618286,
      "rewards/forensics_sinkhorn_reward": 0.670598030090332,
      "rewards/forensics_strict_edge_reward": 0.00019962285114161205,
      "rewards/forensics_tiou_reward": 0.40220925211906433,
      "step": 337
    },
    {
      "completion_length": 21.09375,
      "epoch": 1.4142259414225942,
      "grad_norm": 5.379278659820557,
      "kl": 0.3095703125,
      "learning_rate": 6.47489539748954e-07,
      "loss": 0.088,
      "reward": 0.6868326365947723,
      "reward_std": 0.07777215540409088,
      "reward_std_all": 0.09495237842202187,
      "rewards-max/forensics_sinkhorn_reward": 0.965366780757904,
      "rewards-max/forensics_strict_edge_reward": 0.33076252043247223,
      "rewards-max/forensics_tiou_reward": 0.9395604431629181,
      "rewards/forensics_sinkhorn_reward": 0.592924028635025,
      "rewards/forensics_strict_edge_reward": 0.027012805454432964,
      "rewards/forensics_tiou_reward": 0.3376488983631134,
      "step": 338
    },
    {
      "completion_length": 22.015625,
      "epoch": 1.4184100418410042,
      "grad_norm": 5.762954235076904,
      "kl": 0.4326171875,
      "learning_rate": 6.464435146443515e-07,
      "loss": 0.1617,
      "reward": 0.7731323540210724,
      "reward_std": 0.03447718173265457,
      "reward_std_all": 0.03676849417388439,
      "rewards-max/forensics_sinkhorn_reward": 0.9777272641658783,
      "rewards-max/forensics_strict_edge_reward": 0.7169294059276581,
      "rewards-max/forensics_tiou_reward": 0.9729442298412323,
      "rewards/forensics_sinkhorn_reward": 0.7062550187110901,
      "rewards/forensics_strict_edge_reward": 0.044922349974513054,
      "rewards/forensics_tiou_reward": 0.4487673491239548,
      "step": 339
    },
    {
      "completion_length": 21.140625,
      "epoch": 1.4225941422594142,
      "grad_norm": 5.8581976890563965,
      "kl": 0.388671875,
      "learning_rate": 6.45397489539749e-07,
      "loss": 0.0795,
      "reward": 0.7406424880027771,
      "reward_std": 0.06226141378283501,
      "reward_std_all": 0.06382528133690357,
      "rewards-max/forensics_sinkhorn_reward": 0.9700223207473755,
      "rewards-max/forensics_strict_edge_reward": 0.4231441020965576,
      "rewards-max/forensics_tiou_reward": 0.9517200291156769,
      "rewards/forensics_sinkhorn_reward": 0.6713517904281616,
      "rewards/forensics_strict_edge_reward": 0.0292791947722435,
      "rewards/forensics_tiou_reward": 0.4594668447971344,
      "step": 340
    },
    {
      "completion_length": 20.40625,
      "epoch": 1.4267782426778242,
      "grad_norm": 5.051840782165527,
      "kl": 0.3798828125,
      "learning_rate": 6.443514644351464e-07,
      "loss": 0.1105,
      "reward": 0.7354446053504944,
      "reward_std": 0.06485719233751297,
      "reward_std_all": 0.07034531235694885,
      "rewards-max/forensics_sinkhorn_reward": 0.9638958871364594,
      "rewards-max/forensics_strict_edge_reward": 0.34018705040216446,
      "rewards-max/forensics_tiou_reward": 0.938501238822937,
      "rewards/forensics_sinkhorn_reward": 0.6744904816150665,
      "rewards/forensics_strict_edge_reward": 0.01897844299674034,
      "rewards/forensics_tiou_reward": 0.42653296887874603,
      "step": 341
    },
    {
      "completion_length": 21.171875,
      "epoch": 1.4309623430962344,
      "grad_norm": 5.699254035949707,
      "kl": 0.283203125,
      "learning_rate": 6.433054393305439e-07,
      "loss": 0.0302,
      "reward": 0.7435804009437561,
      "reward_std": 0.0789840891957283,
      "reward_std_all": 0.07440879568457603,
      "rewards-max/forensics_sinkhorn_reward": 0.968302309513092,
      "rewards-max/forensics_strict_edge_reward": 0.503647968173027,
      "rewards-max/forensics_tiou_reward": 0.9482647776603699,
      "rewards/forensics_sinkhorn_reward": 0.7070243656635284,
      "rewards/forensics_strict_edge_reward": 0.05282638035714626,
      "rewards/forensics_tiou_reward": 0.4699995666742325,
      "step": 342
    },
    {
      "completion_length": 21.96875,
      "epoch": 1.4351464435146444,
      "grad_norm": 5.192465782165527,
      "kl": 0.2666015625,
      "learning_rate": 6.422594142259414e-07,
      "loss": 0.1686,
      "reward": 0.7185202538967133,
      "reward_std": 0.08283041790127754,
      "reward_std_all": 0.084597859531641,
      "rewards-max/forensics_sinkhorn_reward": 0.9360659122467041,
      "rewards-max/forensics_strict_edge_reward": 0.3175234720110893,
      "rewards-max/forensics_tiou_reward": 0.8790394067764282,
      "rewards/forensics_sinkhorn_reward": 0.6113806664943695,
      "rewards/forensics_strict_edge_reward": 0.03132488392293453,
      "rewards/forensics_tiou_reward": 0.4010847359895706,
      "step": 343
    },
    {
      "completion_length": 22.34375,
      "epoch": 1.4393305439330544,
      "grad_norm": 5.805377006530762,
      "kl": 0.3232421875,
      "learning_rate": 6.412133891213389e-07,
      "loss": -0.0276,
      "reward": 0.7179188132286072,
      "reward_std": 0.06857092306017876,
      "reward_std_all": 0.07617262750864029,
      "rewards-max/forensics_sinkhorn_reward": 0.9476390182971954,
      "rewards-max/forensics_strict_edge_reward": 0.17452361434698105,
      "rewards-max/forensics_tiou_reward": 0.8915059864521027,
      "rewards/forensics_sinkhorn_reward": 0.6019231379032135,
      "rewards/forensics_strict_edge_reward": 0.012462712358683348,
      "rewards/forensics_tiou_reward": 0.31800153851509094,
      "step": 344
    },
    {
      "completion_length": 21.90625,
      "epoch": 1.4435146443514644,
      "grad_norm": 4.481995582580566,
      "kl": 0.3056640625,
      "learning_rate": 6.401673640167363e-07,
      "loss": 0.1668,
      "reward": 0.7042614817619324,
      "reward_std": 0.07132109813392162,
      "reward_std_all": 0.08994656056165695,
      "rewards-max/forensics_sinkhorn_reward": 0.9577053487300873,
      "rewards-max/forensics_strict_edge_reward": 0.11057738214731216,
      "rewards-max/forensics_tiou_reward": 0.8935226500034332,
      "rewards/forensics_sinkhorn_reward": 0.6330990493297577,
      "rewards/forensics_strict_edge_reward": 0.008694184245541692,
      "rewards/forensics_tiou_reward": 0.4024984538555145,
      "step": 345
    },
    {
      "completion_length": 20.515625,
      "epoch": 1.4476987447698746,
      "grad_norm": 6.2201948165893555,
      "kl": 0.3544921875,
      "learning_rate": 6.391213389121338e-07,
      "loss": 0.1366,
      "reward": 0.69999560713768,
      "reward_std": 0.09684910252690315,
      "reward_std_all": 0.09794536605477333,
      "rewards-max/forensics_sinkhorn_reward": 0.9742391705513,
      "rewards-max/forensics_strict_edge_reward": 0.36565569043159485,
      "rewards-max/forensics_tiou_reward": 0.959576427936554,
      "rewards/forensics_sinkhorn_reward": 0.6373265981674194,
      "rewards/forensics_strict_edge_reward": 0.021546821808442473,
      "rewards/forensics_tiou_reward": 0.39060178399086,
      "step": 346
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.4518828451882846,
      "grad_norm": 4.939878940582275,
      "kl": 0.2197265625,
      "learning_rate": 6.380753138075314e-07,
      "loss": 0.1218,
      "reward": 0.7147096991539001,
      "reward_std": 0.057357028126716614,
      "reward_std_all": 0.07232062704861164,
      "rewards-max/forensics_sinkhorn_reward": 0.9221262037754059,
      "rewards-max/forensics_strict_edge_reward": 0.09324931725859642,
      "rewards-max/forensics_tiou_reward": 0.8363551795482635,
      "rewards/forensics_sinkhorn_reward": 0.5964336395263672,
      "rewards/forensics_strict_edge_reward": 0.0065599558292888105,
      "rewards/forensics_tiou_reward": 0.3213738352060318,
      "step": 347
    },
    {
      "completion_length": 20.90625,
      "epoch": 1.4560669456066946,
      "grad_norm": 6.8511128425598145,
      "kl": 0.3076171875,
      "learning_rate": 6.370292887029289e-07,
      "loss": 0.1367,
      "reward": 0.7788570821285248,
      "reward_std": 0.03234955947846174,
      "reward_std_all": 0.03291620314121246,
      "rewards-max/forensics_sinkhorn_reward": 0.9597437679767609,
      "rewards-max/forensics_strict_edge_reward": 0.2117384374141693,
      "rewards-max/forensics_tiou_reward": 0.9246671497821808,
      "rewards/forensics_sinkhorn_reward": 0.7100566327571869,
      "rewards/forensics_strict_edge_reward": 0.009561762679368258,
      "rewards/forensics_tiou_reward": 0.4098316431045532,
      "step": 348
    },
    {
      "completion_length": 21.03125,
      "epoch": 1.4602510460251046,
      "grad_norm": 4.804234981536865,
      "kl": 0.3466796875,
      "learning_rate": 6.359832635983263e-07,
      "loss": 0.0391,
      "reward": 0.7218728959560394,
      "reward_std": 0.06599757447838783,
      "reward_std_all": 0.07416720688343048,
      "rewards-max/forensics_sinkhorn_reward": 0.9850215315818787,
      "rewards-max/forensics_strict_edge_reward": 0.8053101599216461,
      "rewards-max/forensics_tiou_reward": 0.9889598786830902,
      "rewards/forensics_sinkhorn_reward": 0.6664988994598389,
      "rewards/forensics_strict_edge_reward": 0.043601082637906075,
      "rewards/forensics_tiou_reward": 0.43366724252700806,
      "step": 349
    },
    {
      "completion_length": 19.9375,
      "epoch": 1.4644351464435146,
      "grad_norm": 5.005899429321289,
      "kl": 0.3525390625,
      "learning_rate": 6.349372384937238e-07,
      "loss": -0.0048,
      "reward": 0.7854938805103302,
      "reward_std": 0.0194944036193192,
      "reward_std_all": 0.02128674346022308,
      "rewards-max/forensics_sinkhorn_reward": 0.9625091850757599,
      "rewards-max/forensics_strict_edge_reward": 0.48699167370796204,
      "rewards-max/forensics_tiou_reward": 0.93763068318367,
      "rewards/forensics_sinkhorn_reward": 0.7799558937549591,
      "rewards/forensics_strict_edge_reward": 0.028208360075950623,
      "rewards/forensics_tiou_reward": 0.5368409603834152,
      "step": 350
    },
    {
      "completion_length": 20.0625,
      "epoch": 1.4686192468619246,
      "grad_norm": 5.15999698638916,
      "kl": 0.3046875,
      "learning_rate": 6.338912133891213e-07,
      "loss": 0.2173,
      "reward": 0.749069482088089,
      "reward_std": 0.06233518198132515,
      "reward_std_all": 0.0629095509648323,
      "rewards-max/forensics_sinkhorn_reward": 0.9593102633953094,
      "rewards-max/forensics_strict_edge_reward": 0.11123155988752842,
      "rewards-max/forensics_tiou_reward": 0.9111490249633789,
      "rewards/forensics_sinkhorn_reward": 0.7028302550315857,
      "rewards/forensics_strict_edge_reward": 0.008109120186418295,
      "rewards/forensics_tiou_reward": 0.4346032738685608,
      "step": 351
    },
    {
      "completion_length": 19.765625,
      "epoch": 1.4728033472803348,
      "grad_norm": 4.787565231323242,
      "kl": 0.298828125,
      "learning_rate": 6.328451882845188e-07,
      "loss": -0.0643,
      "reward": 0.7240971624851227,
      "reward_std": 0.06352906674146652,
      "reward_std_all": 0.07126634567975998,
      "rewards-max/forensics_sinkhorn_reward": 0.9709278047084808,
      "rewards-max/forensics_strict_edge_reward": 0.36305589973926544,
      "rewards-max/forensics_tiou_reward": 0.9525978863239288,
      "rewards/forensics_sinkhorn_reward": 0.6644163727760315,
      "rewards/forensics_strict_edge_reward": 0.0240301126614213,
      "rewards/forensics_tiou_reward": 0.380098819732666,
      "step": 352
    },
    {
      "completion_length": 21.59375,
      "epoch": 1.4769874476987448,
      "grad_norm": 4.834192276000977,
      "kl": 0.3427734375,
      "learning_rate": 6.317991631799163e-07,
      "loss": 0.086,
      "reward": 0.7162717282772064,
      "reward_std": 0.09105497598648071,
      "reward_std_all": 0.09108598530292511,
      "rewards-max/forensics_sinkhorn_reward": 0.9610118567943573,
      "rewards-max/forensics_strict_edge_reward": 0.1992451057303697,
      "rewards-max/forensics_tiou_reward": 0.9100753366947174,
      "rewards/forensics_sinkhorn_reward": 0.6302315294742584,
      "rewards/forensics_strict_edge_reward": 0.019427644438110292,
      "rewards/forensics_tiou_reward": 0.3701069504022598,
      "step": 353
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.4811715481171548,
      "grad_norm": 4.513514995574951,
      "kl": 0.27734375,
      "learning_rate": 6.307531380753138e-07,
      "loss": 0.0987,
      "reward": 0.6920731663703918,
      "reward_std": 0.06946751475334167,
      "reward_std_all": 0.08864204213023186,
      "rewards-max/forensics_sinkhorn_reward": 0.9421712458133698,
      "rewards-max/forensics_strict_edge_reward": 0.14972888678312302,
      "rewards-max/forensics_tiou_reward": 0.87260502576828,
      "rewards/forensics_sinkhorn_reward": 0.6264613270759583,
      "rewards/forensics_strict_edge_reward": 0.012659960892051458,
      "rewards/forensics_tiou_reward": 0.3642908185720444,
      "step": 354
    },
    {
      "completion_length": 21.546875,
      "epoch": 1.4853556485355648,
      "grad_norm": 4.994217395782471,
      "kl": 0.298828125,
      "learning_rate": 6.297071129707112e-07,
      "loss": 0.0722,
      "reward": 0.7475643157958984,
      "reward_std": 0.05176609195768833,
      "reward_std_all": 0.05664210766553879,
      "rewards-max/forensics_sinkhorn_reward": 0.9625896215438843,
      "rewards-max/forensics_strict_edge_reward": 0.3965419679880142,
      "rewards-max/forensics_tiou_reward": 0.9261155128479004,
      "rewards/forensics_sinkhorn_reward": 0.6873111128807068,
      "rewards/forensics_strict_edge_reward": 0.03472193982452154,
      "rewards/forensics_tiou_reward": 0.45141512155532837,
      "step": 355
    },
    {
      "completion_length": 20.078125,
      "epoch": 1.489539748953975,
      "grad_norm": 4.973355770111084,
      "kl": 0.1962890625,
      "learning_rate": 6.286610878661088e-07,
      "loss": 0.0699,
      "reward": 0.714320957660675,
      "reward_std": 0.08300702646374702,
      "reward_std_all": 0.08665255829691887,
      "rewards-max/forensics_sinkhorn_reward": 0.9581153094768524,
      "rewards-max/forensics_strict_edge_reward": 0.3332277685403824,
      "rewards-max/forensics_tiou_reward": 0.9222277104854584,
      "rewards/forensics_sinkhorn_reward": 0.6693119704723358,
      "rewards/forensics_strict_edge_reward": 0.026604071725159883,
      "rewards/forensics_tiou_reward": 0.4171058237552643,
      "step": 356
    },
    {
      "completion_length": 22.015625,
      "epoch": 1.493723849372385,
      "grad_norm": 5.588532447814941,
      "kl": 0.2734375,
      "learning_rate": 6.276150627615063e-07,
      "loss": -0.0462,
      "reward": 0.7266486585140228,
      "reward_std": 0.06079642288386822,
      "reward_std_all": 0.06944001838564873,
      "rewards-max/forensics_sinkhorn_reward": 0.9842230081558228,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9864233732223511,
      "rewards/forensics_sinkhorn_reward": 0.673890233039856,
      "rewards/forensics_strict_edge_reward": 0.04267779923975468,
      "rewards/forensics_tiou_reward": 0.41672712564468384,
      "step": 357
    },
    {
      "completion_length": 21.0625,
      "epoch": 1.497907949790795,
      "grad_norm": 4.963912487030029,
      "kl": 0.2998046875,
      "learning_rate": 6.265690376569037e-07,
      "loss": 0.1152,
      "reward": 0.7420946061611176,
      "reward_std": 0.028410212136805058,
      "reward_std_all": 0.04697468411177397,
      "rewards-max/forensics_sinkhorn_reward": 0.9650227725505829,
      "rewards-max/forensics_strict_edge_reward": 0.3773459494113922,
      "rewards-max/forensics_tiou_reward": 0.9354409277439117,
      "rewards/forensics_sinkhorn_reward": 0.7091969847679138,
      "rewards/forensics_strict_edge_reward": 0.03216506680473685,
      "rewards/forensics_tiou_reward": 0.49570050835609436,
      "step": 358
    },
    {
      "completion_length": 22.0625,
      "epoch": 1.502092050209205,
      "grad_norm": 4.660717010498047,
      "kl": 0.361328125,
      "learning_rate": 6.255230125523012e-07,
      "loss": 0.0826,
      "reward": 0.7602362930774689,
      "reward_std": 0.042631082236766815,
      "reward_std_all": 0.04482070729136467,
      "rewards-max/forensics_sinkhorn_reward": 0.9484026730060577,
      "rewards-max/forensics_strict_edge_reward": 0.38803622126579285,
      "rewards-max/forensics_tiou_reward": 0.891352117061615,
      "rewards/forensics_sinkhorn_reward": 0.6969269812107086,
      "rewards/forensics_strict_edge_reward": 0.021562746725976467,
      "rewards/forensics_tiou_reward": 0.47443172335624695,
      "step": 359
    },
    {
      "completion_length": 19.359375,
      "epoch": 1.506276150627615,
      "grad_norm": 4.627596855163574,
      "kl": 0.2392578125,
      "learning_rate": 6.244769874476987e-07,
      "loss": 0.0621,
      "reward": 0.7574318647384644,
      "reward_std": 0.03045611921697855,
      "reward_std_all": 0.041241273283958435,
      "rewards-max/forensics_sinkhorn_reward": 0.9727621674537659,
      "rewards-max/forensics_strict_edge_reward": 0.4293667823076248,
      "rewards-max/forensics_tiou_reward": 0.9543812572956085,
      "rewards/forensics_sinkhorn_reward": 0.7166793346405029,
      "rewards/forensics_strict_edge_reward": 0.040921084582805634,
      "rewards/forensics_tiou_reward": 0.4576532542705536,
      "step": 360
    },
    {
      "completion_length": 20.359375,
      "epoch": 1.510460251046025,
      "grad_norm": 5.3944783210754395,
      "kl": 0.384765625,
      "learning_rate": 6.234309623430962e-07,
      "loss": 0.1353,
      "reward": 0.7366406619548798,
      "reward_std": 0.06083653122186661,
      "reward_std_all": 0.06643299013376236,
      "rewards-max/forensics_sinkhorn_reward": 0.979587972164154,
      "rewards-max/forensics_strict_edge_reward": 0.47266893833875656,
      "rewards-max/forensics_tiou_reward": 0.9716598093509674,
      "rewards/forensics_sinkhorn_reward": 0.7092792391777039,
      "rewards/forensics_strict_edge_reward": 0.02891513518989086,
      "rewards/forensics_tiou_reward": 0.4612499326467514,
      "step": 361
    },
    {
      "completion_length": 19.640625,
      "epoch": 1.514644351464435,
      "grad_norm": 4.687530517578125,
      "kl": 0.3427734375,
      "learning_rate": 6.223849372384937e-07,
      "loss": 0.1634,
      "reward": 0.7435522079467773,
      "reward_std": 0.07450877036899328,
      "reward_std_all": 0.07084850780665874,
      "rewards-max/forensics_sinkhorn_reward": 0.9662267863750458,
      "rewards-max/forensics_strict_edge_reward": 0.45430223643779755,
      "rewards-max/forensics_tiou_reward": 0.9437541365623474,
      "rewards/forensics_sinkhorn_reward": 0.6984884142875671,
      "rewards/forensics_strict_edge_reward": 0.03636947553604841,
      "rewards/forensics_tiou_reward": 0.4737794101238251,
      "step": 362
    },
    {
      "completion_length": 22.46875,
      "epoch": 1.5188284518828452,
      "grad_norm": 4.36262845993042,
      "kl": 0.26171875,
      "learning_rate": 6.213389121338912e-07,
      "loss": 0.1454,
      "reward": 0.7269719541072845,
      "reward_std": 0.0702333003282547,
      "reward_std_all": 0.07430924847722054,
      "rewards-max/forensics_sinkhorn_reward": 0.9726028740406036,
      "rewards-max/forensics_strict_edge_reward": 0.13572769053280354,
      "rewards-max/forensics_tiou_reward": 0.9527595341205597,
      "rewards/forensics_sinkhorn_reward": 0.6704823672771454,
      "rewards/forensics_strict_edge_reward": 0.011130121652968228,
      "rewards/forensics_tiou_reward": 0.44769734144210815,
      "step": 363
    },
    {
      "completion_length": 22.421875,
      "epoch": 1.5230125523012552,
      "grad_norm": 4.705061912536621,
      "kl": 0.267578125,
      "learning_rate": 6.202928870292888e-07,
      "loss": 0.2684,
      "reward": 0.7134746313095093,
      "reward_std": 0.057784028351306915,
      "reward_std_all": 0.07434027642011642,
      "rewards-max/forensics_sinkhorn_reward": 0.9827489852905273,
      "rewards-max/forensics_strict_edge_reward": 0.21100818365812302,
      "rewards-max/forensics_tiou_reward": 0.9818975031375885,
      "rewards/forensics_sinkhorn_reward": 0.6701890230178833,
      "rewards/forensics_strict_edge_reward": 0.01399281993508339,
      "rewards/forensics_tiou_reward": 0.4633985608816147,
      "step": 364
    },
    {
      "completion_length": 20.8125,
      "epoch": 1.5271966527196654,
      "grad_norm": 4.512877464294434,
      "kl": 0.287109375,
      "learning_rate": 6.192468619246862e-07,
      "loss": 0.0522,
      "reward": 0.7502400875091553,
      "reward_std": 0.06089301407337189,
      "reward_std_all": 0.0595649890601635,
      "rewards-max/forensics_sinkhorn_reward": 0.9678507745265961,
      "rewards-max/forensics_strict_edge_reward": 0.5317803770303726,
      "rewards-max/forensics_tiou_reward": 0.9489706456661224,
      "rewards/forensics_sinkhorn_reward": 0.7045783400535583,
      "rewards/forensics_strict_edge_reward": 0.03761482052505016,
      "rewards/forensics_tiou_reward": 0.4779476821422577,
      "step": 365
    },
    {
      "completion_length": 20.890625,
      "epoch": 1.5313807531380754,
      "grad_norm": 5.243403911590576,
      "kl": 0.43359375,
      "learning_rate": 6.182008368200836e-07,
      "loss": 0.1153,
      "reward": 0.647572785615921,
      "reward_std": 0.11710432171821594,
      "reward_std_all": 0.13664427027106285,
      "rewards-max/forensics_sinkhorn_reward": 0.9190747439861298,
      "rewards-max/forensics_strict_edge_reward": 0.035673994570970535,
      "rewards-max/forensics_tiou_reward": 0.8180605173110962,
      "rewards/forensics_sinkhorn_reward": 0.5554516762495041,
      "rewards/forensics_strict_edge_reward": 0.0021521590533666313,
      "rewards/forensics_tiou_reward": 0.2986220568418503,
      "step": 366
    },
    {
      "completion_length": 22.515625,
      "epoch": 1.5355648535564854,
      "grad_norm": 4.277680397033691,
      "kl": 0.3330078125,
      "learning_rate": 6.171548117154811e-07,
      "loss": 0.0813,
      "reward": 0.7404772639274597,
      "reward_std": 0.05932823196053505,
      "reward_std_all": 0.06600966863334179,
      "rewards-max/forensics_sinkhorn_reward": 0.9746724665164948,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9574402868747711,
      "rewards/forensics_sinkhorn_reward": 0.6845270991325378,
      "rewards/forensics_strict_edge_reward": 0.02177668083459139,
      "rewards/forensics_tiou_reward": 0.44193513691425323,
      "step": 367
    },
    {
      "completion_length": 21.359375,
      "epoch": 1.5397489539748954,
      "grad_norm": 5.048942565917969,
      "kl": 0.3466796875,
      "learning_rate": 6.161087866108786e-07,
      "loss": -0.0027,
      "reward": 0.7652219533920288,
      "reward_std": 0.05109707824885845,
      "reward_std_all": 0.048699657432734966,
      "rewards-max/forensics_sinkhorn_reward": 0.9732483327388763,
      "rewards-max/forensics_strict_edge_reward": 0.47400766611099243,
      "rewards-max/forensics_tiou_reward": 0.9598242938518524,
      "rewards/forensics_sinkhorn_reward": 0.7458584606647491,
      "rewards/forensics_strict_edge_reward": 0.04314751364290714,
      "rewards/forensics_tiou_reward": 0.5748967826366425,
      "step": 368
    },
    {
      "completion_length": 20.65625,
      "epoch": 1.5439330543933054,
      "grad_norm": 5.0535078048706055,
      "kl": 0.26904296875,
      "learning_rate": 6.150627615062761e-07,
      "loss": 0.0547,
      "reward": 0.6789366602897644,
      "reward_std": 0.10283024236559868,
      "reward_std_all": 0.11179081723093987,
      "rewards-max/forensics_sinkhorn_reward": 0.9807242751121521,
      "rewards-max/forensics_strict_edge_reward": 0.7514603435993195,
      "rewards-max/forensics_tiou_reward": 0.9795326888561249,
      "rewards/forensics_sinkhorn_reward": 0.6089808642864227,
      "rewards/forensics_strict_edge_reward": 0.03577323257923126,
      "rewards/forensics_tiou_reward": 0.3835226595401764,
      "step": 369
    },
    {
      "completion_length": 22.296875,
      "epoch": 1.5481171548117154,
      "grad_norm": 4.227104663848877,
      "kl": 0.24755859375,
      "learning_rate": 6.140167364016736e-07,
      "loss": 0.0951,
      "reward": 0.7688070833683014,
      "reward_std": 0.04987512528896332,
      "reward_std_all": 0.04674042388796806,
      "rewards-max/forensics_sinkhorn_reward": 0.9518202841281891,
      "rewards-max/forensics_strict_edge_reward": 0.18439566614688374,
      "rewards-max/forensics_tiou_reward": 0.8855659365653992,
      "rewards/forensics_sinkhorn_reward": 0.7109605073928833,
      "rewards/forensics_strict_edge_reward": 0.014834367986622965,
      "rewards/forensics_tiou_reward": 0.5088105350732803,
      "step": 370
    },
    {
      "completion_length": 23.78125,
      "epoch": 1.5523012552301254,
      "grad_norm": 5.4179301261901855,
      "kl": 0.326171875,
      "learning_rate": 6.129707112970711e-07,
      "loss": -0.006,
      "reward": 0.6931373178958893,
      "reward_std": 0.10714816674590111,
      "reward_std_all": 0.10925194248557091,
      "rewards-max/forensics_sinkhorn_reward": 0.9282764792442322,
      "rewards-max/forensics_strict_edge_reward": 0.14202839753124863,
      "rewards-max/forensics_tiou_reward": 0.8450861871242523,
      "rewards/forensics_sinkhorn_reward": 0.5740023255348206,
      "rewards/forensics_strict_edge_reward": 0.007319456490222365,
      "rewards/forensics_tiou_reward": 0.3106169253587723,
      "step": 371
    },
    {
      "completion_length": 22.203125,
      "epoch": 1.5564853556485354,
      "grad_norm": 6.149571418762207,
      "kl": 0.3515625,
      "learning_rate": 6.119246861924686e-07,
      "loss": 0.0427,
      "reward": 0.705662190914154,
      "reward_std": 0.09213547222316265,
      "reward_std_all": 0.09810916148126125,
      "rewards-max/forensics_sinkhorn_reward": 0.9734344184398651,
      "rewards-max/forensics_strict_edge_reward": 0.2051875637844205,
      "rewards-max/forensics_tiou_reward": 0.9573954641819,
      "rewards/forensics_sinkhorn_reward": 0.6449226438999176,
      "rewards/forensics_strict_edge_reward": 0.009428625111468136,
      "rewards/forensics_tiou_reward": 0.42379236221313477,
      "step": 372
    },
    {
      "completion_length": 22.09375,
      "epoch": 1.5606694560669456,
      "grad_norm": 5.884457588195801,
      "kl": 0.361328125,
      "learning_rate": 6.108786610878662e-07,
      "loss": 0.1157,
      "reward": 0.7378393113613129,
      "reward_std": 0.045897262170910835,
      "reward_std_all": 0.05623314902186394,
      "rewards-max/forensics_sinkhorn_reward": 0.9602579772472382,
      "rewards-max/forensics_strict_edge_reward": 0.2097843885421753,
      "rewards-max/forensics_tiou_reward": 0.9231168627738953,
      "rewards/forensics_sinkhorn_reward": 0.6859400868415833,
      "rewards/forensics_strict_edge_reward": 0.015348992310464382,
      "rewards/forensics_tiou_reward": 0.44791021943092346,
      "step": 373
    },
    {
      "completion_length": 22.15625,
      "epoch": 1.5648535564853556,
      "grad_norm": 7.652299404144287,
      "kl": 0.30859375,
      "learning_rate": 6.098326359832637e-07,
      "loss": 0.1061,
      "reward": 0.7663557529449463,
      "reward_std": 0.05848261062055826,
      "reward_std_all": 0.05399237759411335,
      "rewards-max/forensics_sinkhorn_reward": 0.9712152779102325,
      "rewards-max/forensics_strict_edge_reward": 0.3850628435611725,
      "rewards-max/forensics_tiou_reward": 0.9541001915931702,
      "rewards/forensics_sinkhorn_reward": 0.6872902810573578,
      "rewards/forensics_strict_edge_reward": 0.023317434825003147,
      "rewards/forensics_tiou_reward": 0.48456971347332,
      "step": 374
    },
    {
      "completion_length": 20.90625,
      "epoch": 1.5690376569037658,
      "grad_norm": 5.505954742431641,
      "kl": 0.33984375,
      "learning_rate": 6.08786610878661e-07,
      "loss": 0.0991,
      "reward": 0.7361464500427246,
      "reward_std": 0.06568093225359917,
      "reward_std_all": 0.07020440325140953,
      "rewards-max/forensics_sinkhorn_reward": 0.9680846035480499,
      "rewards-max/forensics_strict_edge_reward": 0.13281402736902237,
      "rewards-max/forensics_tiou_reward": 0.9471602737903595,
      "rewards/forensics_sinkhorn_reward": 0.7056028246879578,
      "rewards/forensics_strict_edge_reward": 0.01433911919593811,
      "rewards/forensics_tiou_reward": 0.5146302878856659,
      "step": 375
    },
    {
      "completion_length": 19.65625,
      "epoch": 1.5732217573221758,
      "grad_norm": 7.687255382537842,
      "kl": 0.333984375,
      "learning_rate": 6.077405857740585e-07,
      "loss": -0.0236,
      "reward": 0.7709713280200958,
      "reward_std": 0.03500334173440933,
      "reward_std_all": 0.03651972021907568,
      "rewards-max/forensics_sinkhorn_reward": 0.9590862989425659,
      "rewards-max/forensics_strict_edge_reward": 0.3439207598567009,
      "rewards-max/forensics_tiou_reward": 0.9255056381225586,
      "rewards/forensics_sinkhorn_reward": 0.708152711391449,
      "rewards/forensics_strict_edge_reward": 0.03411304019391537,
      "rewards/forensics_tiou_reward": 0.4360814839601517,
      "step": 376
    },
    {
      "completion_length": 22.9375,
      "epoch": 1.5774058577405858,
      "grad_norm": 5.585402965545654,
      "kl": 0.3134765625,
      "learning_rate": 6.06694560669456e-07,
      "loss": 0.0948,
      "reward": 0.7470651566982269,
      "reward_std": 0.0591875072568655,
      "reward_std_all": 0.05894423648715019,
      "rewards-max/forensics_sinkhorn_reward": 0.9479630589485168,
      "rewards-max/forensics_strict_edge_reward": 0.427202383056283,
      "rewards-max/forensics_tiou_reward": 0.8870764374732971,
      "rewards/forensics_sinkhorn_reward": 0.6560353636741638,
      "rewards/forensics_strict_edge_reward": 0.030061023076996207,
      "rewards/forensics_tiou_reward": 0.42056845128536224,
      "step": 377
    },
    {
      "completion_length": 21.875,
      "epoch": 1.5815899581589958,
      "grad_norm": 6.499553203582764,
      "kl": 0.3095703125,
      "learning_rate": 6.056485355648535e-07,
      "loss": 0.0503,
      "reward": 0.7293033301830292,
      "reward_std": 0.05310477130115032,
      "reward_std_all": 0.06963061168789864,
      "rewards-max/forensics_sinkhorn_reward": 0.9785219132900238,
      "rewards-max/forensics_strict_edge_reward": 0.5872748494148254,
      "rewards-max/forensics_tiou_reward": 0.9715951085090637,
      "rewards/forensics_sinkhorn_reward": 0.6789517402648926,
      "rewards/forensics_strict_edge_reward": 0.036256343126297,
      "rewards/forensics_tiou_reward": 0.4615572988986969,
      "step": 378
    },
    {
      "completion_length": 22.609375,
      "epoch": 1.5857740585774058,
      "grad_norm": 4.982816696166992,
      "kl": 0.4287109375,
      "learning_rate": 6.04602510460251e-07,
      "loss": 0.1093,
      "reward": 0.7448067665100098,
      "reward_std": 0.05000703036785126,
      "reward_std_all": 0.05483405850827694,
      "rewards-max/forensics_sinkhorn_reward": 0.9322164058685303,
      "rewards-max/forensics_strict_edge_reward": 0.2663590111187659,
      "rewards-max/forensics_tiou_reward": 0.8336657285690308,
      "rewards/forensics_sinkhorn_reward": 0.6715274155139923,
      "rewards/forensics_strict_edge_reward": 0.0183987898635678,
      "rewards/forensics_tiou_reward": 0.416573241353035,
      "step": 379
    },
    {
      "completion_length": 19.484375,
      "epoch": 1.5899581589958158,
      "grad_norm": 7.084120273590088,
      "kl": 0.4375,
      "learning_rate": 6.035564853556485e-07,
      "loss": 0.0791,
      "reward": 0.7612677216529846,
      "reward_std": 0.06139550730586052,
      "reward_std_all": 0.05731269344687462,
      "rewards-max/forensics_sinkhorn_reward": 0.9681828320026398,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9510689079761505,
      "rewards/forensics_sinkhorn_reward": 0.7263400554656982,
      "rewards/forensics_strict_edge_reward": 0.0595848523080349,
      "rewards/forensics_tiou_reward": 0.45963720977306366,
      "step": 380
    },
    {
      "completion_length": 21.546875,
      "epoch": 1.5941422594142258,
      "grad_norm": 5.763733863830566,
      "kl": 0.3876953125,
      "learning_rate": 6.02510460251046e-07,
      "loss": 0.0588,
      "reward": 0.7483764588832855,
      "reward_std": 0.03449405916035175,
      "reward_std_all": 0.04643349163234234,
      "rewards-max/forensics_sinkhorn_reward": 0.9656250476837158,
      "rewards-max/forensics_strict_edge_reward": 0.45270346850156784,
      "rewards-max/forensics_tiou_reward": 0.9391629099845886,
      "rewards/forensics_sinkhorn_reward": 0.6733009517192841,
      "rewards/forensics_strict_edge_reward": 0.024095927365124226,
      "rewards/forensics_tiou_reward": 0.4141360968351364,
      "step": 381
    },
    {
      "completion_length": 21.484375,
      "epoch": 1.5983263598326358,
      "grad_norm": 5.57182502746582,
      "kl": 0.283203125,
      "learning_rate": 6.014644351464436e-07,
      "loss": -0.0623,
      "reward": 0.7739968001842499,
      "reward_std": 0.012214530259370804,
      "reward_std_all": 0.02400848176330328,
      "rewards-max/forensics_sinkhorn_reward": 0.9696043431758881,
      "rewards-max/forensics_strict_edge_reward": 0.37157370895147324,
      "rewards-max/forensics_tiou_reward": 0.9475788772106171,
      "rewards/forensics_sinkhorn_reward": 0.7430481314659119,
      "rewards/forensics_strict_edge_reward": 0.03839271143078804,
      "rewards/forensics_tiou_reward": 0.5065649598836899,
      "step": 382
    },
    {
      "completion_length": 21.953125,
      "epoch": 1.602510460251046,
      "grad_norm": 5.379236698150635,
      "kl": 0.36328125,
      "learning_rate": 6.00418410041841e-07,
      "loss": 0.0676,
      "reward": 0.742140382528305,
      "reward_std": 0.056368330493569374,
      "reward_std_all": 0.05969628319144249,
      "rewards-max/forensics_sinkhorn_reward": 0.9694546461105347,
      "rewards-max/forensics_strict_edge_reward": 0.326234370470047,
      "rewards-max/forensics_tiou_reward": 0.9450261294841766,
      "rewards/forensics_sinkhorn_reward": 0.6734442114830017,
      "rewards/forensics_strict_edge_reward": 0.01890890020877123,
      "rewards/forensics_tiou_reward": 0.44420966506004333,
      "step": 383
    },
    {
      "completion_length": 22.109375,
      "epoch": 1.606694560669456,
      "grad_norm": 5.215673446655273,
      "kl": 0.26806640625,
      "learning_rate": 5.993723849372385e-07,
      "loss": 0.0434,
      "reward": 0.7478226125240326,
      "reward_std": 0.026291060727089643,
      "reward_std_all": 0.04349929094314575,
      "rewards-max/forensics_sinkhorn_reward": 0.9852192997932434,
      "rewards-max/forensics_strict_edge_reward": 0.19309754483401775,
      "rewards-max/forensics_tiou_reward": 0.9876252710819244,
      "rewards/forensics_sinkhorn_reward": 0.6798511147499084,
      "rewards/forensics_strict_edge_reward": 0.023022048932034522,
      "rewards/forensics_tiou_reward": 0.510462760925293,
      "step": 384
    },
    {
      "completion_length": 22.125,
      "epoch": 1.6108786610878663,
      "grad_norm": 4.861598014831543,
      "kl": 0.439453125,
      "learning_rate": 5.983263598326359e-07,
      "loss": -0.0003,
      "reward": 0.6701061725616455,
      "reward_std": 0.09796851500868797,
      "reward_std_all": 0.11640539020299911,
      "rewards-max/forensics_sinkhorn_reward": 0.9728967845439911,
      "rewards-max/forensics_strict_edge_reward": 0.3967413753271103,
      "rewards-max/forensics_tiou_reward": 0.9571502506732941,
      "rewards/forensics_sinkhorn_reward": 0.6107592880725861,
      "rewards/forensics_strict_edge_reward": 0.025931249372661114,
      "rewards/forensics_tiou_reward": 0.3797173500061035,
      "step": 385
    },
    {
      "completion_length": 22.375,
      "epoch": 1.6150627615062763,
      "grad_norm": 6.145466327667236,
      "kl": 0.515625,
      "learning_rate": 5.972803347280334e-07,
      "loss": -0.069,
      "reward": 0.7458474040031433,
      "reward_std": 0.04512437433004379,
      "reward_std_all": 0.05113282427191734,
      "rewards-max/forensics_sinkhorn_reward": 0.9181273579597473,
      "rewards-max/forensics_strict_edge_reward": 0.019640328478999436,
      "rewards-max/forensics_tiou_reward": 0.8189067840576172,
      "rewards/forensics_sinkhorn_reward": 0.6556274592876434,
      "rewards/forensics_strict_edge_reward": 0.001544587103126105,
      "rewards/forensics_tiou_reward": 0.39323198795318604,
      "step": 386
    },
    {
      "completion_length": 20.296875,
      "epoch": 1.6192468619246863,
      "grad_norm": 10.511856079101562,
      "kl": 0.4140625,
      "learning_rate": 5.962343096234309e-07,
      "loss": 0.1204,
      "reward": 0.7173782289028168,
      "reward_std": 0.06446296349167824,
      "reward_std_all": 0.07237523421645164,
      "rewards-max/forensics_sinkhorn_reward": 0.9831103980541229,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9830532968044281,
      "rewards/forensics_sinkhorn_reward": 0.6174429357051849,
      "rewards/forensics_strict_edge_reward": 0.05475768819451332,
      "rewards/forensics_tiou_reward": 0.36716900765895844,
      "step": 387
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.6234309623430963,
      "grad_norm": 5.5434794425964355,
      "kl": 0.3134765625,
      "learning_rate": 5.951882845188284e-07,
      "loss": 0.1587,
      "reward": 0.7216686010360718,
      "reward_std": 0.06279426999390125,
      "reward_std_all": 0.07486680150032043,
      "rewards-max/forensics_sinkhorn_reward": 0.9615877568721771,
      "rewards-max/forensics_strict_edge_reward": 0.2621621936559677,
      "rewards-max/forensics_tiou_reward": 0.9308885633945465,
      "rewards/forensics_sinkhorn_reward": 0.6262286603450775,
      "rewards/forensics_strict_edge_reward": 0.01648756582289934,
      "rewards/forensics_tiou_reward": 0.3841332346200943,
      "step": 388
    },
    {
      "completion_length": 21.4375,
      "epoch": 1.6276150627615062,
      "grad_norm": 5.702893257141113,
      "kl": 0.4384765625,
      "learning_rate": 5.941422594142259e-07,
      "loss": 0.1795,
      "reward": 0.7069648206233978,
      "reward_std": 0.09214404225349426,
      "reward_std_all": 0.09581317752599716,
      "rewards-max/forensics_sinkhorn_reward": 0.9532130360603333,
      "rewards-max/forensics_strict_edge_reward": 0.18962457589805126,
      "rewards-max/forensics_tiou_reward": 0.9117099940776825,
      "rewards/forensics_sinkhorn_reward": 0.6224155128002167,
      "rewards/forensics_strict_edge_reward": 0.010719758924096823,
      "rewards/forensics_tiou_reward": 0.41493166983127594,
      "step": 389
    },
    {
      "completion_length": 22.234375,
      "epoch": 1.6317991631799162,
      "grad_norm": 5.849832534790039,
      "kl": 0.3955078125,
      "learning_rate": 5.930962343096235e-07,
      "loss": 0.0119,
      "reward": 0.703152984380722,
      "reward_std": 0.07743551768362522,
      "reward_std_all": 0.08456329256296158,
      "rewards-max/forensics_sinkhorn_reward": 0.972509354352951,
      "rewards-max/forensics_strict_edge_reward": 0.402368500828743,
      "rewards-max/forensics_tiou_reward": 0.9533432722091675,
      "rewards/forensics_sinkhorn_reward": 0.6227516531944275,
      "rewards/forensics_strict_edge_reward": 0.026937715709209442,
      "rewards/forensics_tiou_reward": 0.4078354835510254,
      "step": 390
    },
    {
      "completion_length": 23.0625,
      "epoch": 1.6359832635983262,
      "grad_norm": 6.858505725860596,
      "kl": 0.2939453125,
      "learning_rate": 5.92050209205021e-07,
      "loss": -0.013,
      "reward": 0.7061334550380707,
      "reward_std": 0.0725005529820919,
      "reward_std_all": 0.08274871483445168,
      "rewards-max/forensics_sinkhorn_reward": 0.9810014367103577,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9779469966888428,
      "rewards/forensics_sinkhorn_reward": 0.6426362097263336,
      "rewards/forensics_strict_edge_reward": 0.03426122572273016,
      "rewards/forensics_tiou_reward": 0.4414882957935333,
      "step": 391
    },
    {
      "completion_length": 22.125,
      "epoch": 1.6401673640167362,
      "grad_norm": 7.156951904296875,
      "kl": 0.423828125,
      "learning_rate": 5.910041841004184e-07,
      "loss": 0.0027,
      "reward": 0.7444148361682892,
      "reward_std": 0.08086244016885757,
      "reward_std_all": 0.07687725871801376,
      "rewards-max/forensics_sinkhorn_reward": 0.9344858825206757,
      "rewards-max/forensics_strict_edge_reward": 0.12345641478896141,
      "rewards-max/forensics_tiou_reward": 0.8312570154666901,
      "rewards/forensics_sinkhorn_reward": 0.6833135485649109,
      "rewards/forensics_strict_edge_reward": 0.006642868043854833,
      "rewards/forensics_tiou_reward": 0.4456072151660919,
      "step": 392
    },
    {
      "completion_length": 23.890625,
      "epoch": 1.6443514644351465,
      "grad_norm": 6.060782432556152,
      "kl": 0.322265625,
      "learning_rate": 5.899581589958159e-07,
      "loss": 0.011,
      "reward": 0.7303098142147064,
      "reward_std": 0.022992667742073536,
      "reward_std_all": 0.04726656340062618,
      "rewards-max/forensics_sinkhorn_reward": 0.9563013315200806,
      "rewards-max/forensics_strict_edge_reward": 0.21629484742879868,
      "rewards-max/forensics_tiou_reward": 0.9143508076667786,
      "rewards/forensics_sinkhorn_reward": 0.6507452726364136,
      "rewards/forensics_strict_edge_reward": 0.013898471370339394,
      "rewards/forensics_tiou_reward": 0.4112683832645416,
      "step": 393
    },
    {
      "completion_length": 20.046875,
      "epoch": 1.6485355648535565,
      "grad_norm": 6.183030605316162,
      "kl": 0.3798828125,
      "learning_rate": 5.889121338912134e-07,
      "loss": 0.0424,
      "reward": 0.7464107573032379,
      "reward_std": 0.0605726633220911,
      "reward_std_all": 0.0631045363843441,
      "rewards-max/forensics_sinkhorn_reward": 0.955994188785553,
      "rewards-max/forensics_strict_edge_reward": 0.3443078324198723,
      "rewards-max/forensics_tiou_reward": 0.9239334762096405,
      "rewards/forensics_sinkhorn_reward": 0.7156724333763123,
      "rewards/forensics_strict_edge_reward": 0.04108541226014495,
      "rewards/forensics_tiou_reward": 0.46259529888629913,
      "step": 394
    },
    {
      "completion_length": 21.296875,
      "epoch": 1.6527196652719667,
      "grad_norm": 5.039670467376709,
      "kl": 0.2802734375,
      "learning_rate": 5.878661087866108e-07,
      "loss": 0.0552,
      "reward": 0.77930548787117,
      "reward_std": 0.03193635027855635,
      "reward_std_all": 0.031631252728402615,
      "rewards-max/forensics_sinkhorn_reward": 0.9765019416809082,
      "rewards-max/forensics_strict_edge_reward": 0.4608110338449478,
      "rewards-max/forensics_tiou_reward": 0.964250773191452,
      "rewards/forensics_sinkhorn_reward": 0.7514167129993439,
      "rewards/forensics_strict_edge_reward": 0.042105323635041714,
      "rewards/forensics_tiou_reward": 0.5476257503032684,
      "step": 395
    },
    {
      "completion_length": 20.984375,
      "epoch": 1.6569037656903767,
      "grad_norm": 6.2166619300842285,
      "kl": 0.40625,
      "learning_rate": 5.868200836820083e-07,
      "loss": 0.0122,
      "reward": 0.738553524017334,
      "reward_std": 0.06411497481167316,
      "reward_std_all": 0.06494316458702087,
      "rewards-max/forensics_sinkhorn_reward": 0.9720553457736969,
      "rewards-max/forensics_strict_edge_reward": 0.1412667902186513,
      "rewards-max/forensics_tiou_reward": 0.9487152695655823,
      "rewards/forensics_sinkhorn_reward": 0.6812441945075989,
      "rewards/forensics_strict_edge_reward": 0.007157553802244365,
      "rewards/forensics_tiou_reward": 0.44916093349456787,
      "step": 396
    },
    {
      "completion_length": 20.890625,
      "epoch": 1.6610878661087867,
      "grad_norm": 6.165449619293213,
      "kl": 0.3759765625,
      "learning_rate": 5.857740585774058e-07,
      "loss": 0.032,
      "reward": 0.6965705752372742,
      "reward_std": 0.03385252319276333,
      "reward_std_all": 0.069800466299057,
      "rewards-max/forensics_sinkhorn_reward": 0.9616619646549225,
      "rewards-max/forensics_strict_edge_reward": 0.48574335873126984,
      "rewards-max/forensics_tiou_reward": 0.9361702799797058,
      "rewards/forensics_sinkhorn_reward": 0.6452353298664093,
      "rewards/forensics_strict_edge_reward": 0.04594471678137779,
      "rewards/forensics_tiou_reward": 0.4022551625967026,
      "step": 397
    },
    {
      "completion_length": 21.171875,
      "epoch": 1.6652719665271967,
      "grad_norm": 4.5214033126831055,
      "kl": 0.2578125,
      "learning_rate": 5.847280334728033e-07,
      "loss": 0.1089,
      "reward": 0.6963033378124237,
      "reward_std": 0.0440601697191596,
      "reward_std_all": 0.07482004165649414,
      "rewards-max/forensics_sinkhorn_reward": 0.9756232798099518,
      "rewards-max/forensics_strict_edge_reward": 0.5669038891792297,
      "rewards-max/forensics_tiou_reward": 0.9670599400997162,
      "rewards/forensics_sinkhorn_reward": 0.6507860720157623,
      "rewards/forensics_strict_edge_reward": 0.03978351131081581,
      "rewards/forensics_tiou_reward": 0.4284755438566208,
      "step": 398
    },
    {
      "completion_length": 21.484375,
      "epoch": 1.6694560669456067,
      "grad_norm": 6.235031604766846,
      "kl": 0.306640625,
      "learning_rate": 5.836820083682009e-07,
      "loss": 0.107,
      "reward": 0.7091722190380096,
      "reward_std": 0.08507545292377472,
      "reward_std_all": 0.09266472235321999,
      "rewards-max/forensics_sinkhorn_reward": 0.9179374277591705,
      "rewards-max/forensics_strict_edge_reward": 0.18085789936594665,
      "rewards-max/forensics_tiou_reward": 0.8225165605545044,
      "rewards/forensics_sinkhorn_reward": 0.6336694359779358,
      "rewards/forensics_strict_edge_reward": 0.017055482312571257,
      "rewards/forensics_tiou_reward": 0.3835704028606415,
      "step": 399
    },
    {
      "completion_length": 20.578125,
      "epoch": 1.6736401673640167,
      "grad_norm": 5.693579196929932,
      "kl": 0.3037109375,
      "learning_rate": 5.826359832635983e-07,
      "loss": 0.1411,
      "reward": 0.7243046462535858,
      "reward_std": 0.06611137837171555,
      "reward_std_all": 0.07740030996501446,
      "rewards-max/forensics_sinkhorn_reward": 0.9553569853305817,
      "rewards-max/forensics_strict_edge_reward": 0.4496074318885803,
      "rewards-max/forensics_tiou_reward": 0.9208479523658752,
      "rewards/forensics_sinkhorn_reward": 0.6580355167388916,
      "rewards/forensics_strict_edge_reward": 0.02498909831047058,
      "rewards/forensics_tiou_reward": 0.4105204641819,
      "step": 400
    },
    {
      "completion_length": 20.46875,
      "epoch": 1.6778242677824267,
      "grad_norm": 5.515850067138672,
      "kl": 0.427734375,
      "learning_rate": 5.815899581589958e-07,
      "loss": 0.1373,
      "reward": 0.7547091841697693,
      "reward_std": 0.04495588783174753,
      "reward_std_all": 0.05306335166096687,
      "rewards-max/forensics_sinkhorn_reward": 0.9750484824180603,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9676505625247955,
      "rewards/forensics_sinkhorn_reward": 0.7162733972072601,
      "rewards/forensics_strict_edge_reward": 0.05904507078230381,
      "rewards/forensics_tiou_reward": 0.4874037802219391,
      "step": 401
    },
    {
      "completion_length": 22.546875,
      "epoch": 1.6820083682008367,
      "grad_norm": 6.415983200073242,
      "kl": 0.3896484375,
      "learning_rate": 5.805439330543933e-07,
      "loss": 0.062,
      "reward": 0.7040258049964905,
      "reward_std": 0.08788670226931572,
      "reward_std_all": 0.09137216582894325,
      "rewards-max/forensics_sinkhorn_reward": 0.9726536571979523,
      "rewards-max/forensics_strict_edge_reward": 0.28743013530038297,
      "rewards-max/forensics_tiou_reward": 0.9459576606750488,
      "rewards/forensics_sinkhorn_reward": 0.6350689828395844,
      "rewards/forensics_strict_edge_reward": 0.012545740348286927,
      "rewards/forensics_tiou_reward": 0.41707929968833923,
      "step": 402
    },
    {
      "completion_length": 20.359375,
      "epoch": 1.6861924686192469,
      "grad_norm": 5.413698196411133,
      "kl": 0.291015625,
      "learning_rate": 5.794979079497908e-07,
      "loss": 0.1805,
      "reward": 0.7171506881713867,
      "reward_std": 0.08360270783305168,
      "reward_std_all": 0.08874546736478806,
      "rewards-max/forensics_sinkhorn_reward": 0.9556054174900055,
      "rewards-max/forensics_strict_edge_reward": 0.30519454553723335,
      "rewards-max/forensics_tiou_reward": 0.9223961532115936,
      "rewards/forensics_sinkhorn_reward": 0.6857580542564392,
      "rewards/forensics_strict_edge_reward": 0.03623908408917487,
      "rewards/forensics_tiou_reward": 0.4380967319011688,
      "step": 403
    },
    {
      "completion_length": 21.328125,
      "epoch": 1.6903765690376569,
      "grad_norm": 10.012383460998535,
      "kl": 0.3232421875,
      "learning_rate": 5.784518828451883e-07,
      "loss": 0.029,
      "reward": 0.678540050983429,
      "reward_std": 0.09584001451730728,
      "reward_std_all": 0.11099822074174881,
      "rewards-max/forensics_sinkhorn_reward": 0.964150607585907,
      "rewards-max/forensics_strict_edge_reward": 0.37793944776058197,
      "rewards-max/forensics_tiou_reward": 0.9321412742137909,
      "rewards/forensics_sinkhorn_reward": 0.6282684206962585,
      "rewards/forensics_strict_edge_reward": 0.02698384178802371,
      "rewards/forensics_tiou_reward": 0.39234110713005066,
      "step": 404
    },
    {
      "completion_length": 22.984375,
      "epoch": 1.694560669456067,
      "grad_norm": 6.549920082092285,
      "kl": 0.32421875,
      "learning_rate": 5.774058577405857e-07,
      "loss": 0.1333,
      "reward": 0.7103103995323181,
      "reward_std": 0.10251650586724281,
      "reward_std_all": 0.10099661350250244,
      "rewards-max/forensics_sinkhorn_reward": 0.9303835332393646,
      "rewards-max/forensics_strict_edge_reward": 0.11679134517908096,
      "rewards-max/forensics_tiou_reward": 0.860673189163208,
      "rewards/forensics_sinkhorn_reward": 0.6095786690711975,
      "rewards/forensics_strict_edge_reward": 0.011324249673634768,
      "rewards/forensics_tiou_reward": 0.395119845867157,
      "step": 405
    },
    {
      "completion_length": 21.5625,
      "epoch": 1.698744769874477,
      "grad_norm": 5.661251068115234,
      "kl": 0.447265625,
      "learning_rate": 5.763598326359832e-07,
      "loss": 0.0636,
      "reward": 0.7537543773651123,
      "reward_std": 0.06443304382264614,
      "reward_std_all": 0.06121627613902092,
      "rewards-max/forensics_sinkhorn_reward": 0.9456246495246887,
      "rewards-max/forensics_strict_edge_reward": 0.282395675778389,
      "rewards-max/forensics_tiou_reward": 0.8970468938350677,
      "rewards/forensics_sinkhorn_reward": 0.7153401970863342,
      "rewards/forensics_strict_edge_reward": 0.02252234984189272,
      "rewards/forensics_tiou_reward": 0.4700646549463272,
      "step": 406
    },
    {
      "completion_length": 20.15625,
      "epoch": 1.702928870292887,
      "grad_norm": 5.104758262634277,
      "kl": 0.294921875,
      "learning_rate": 5.753138075313807e-07,
      "loss": 0.1895,
      "reward": 0.7598950266838074,
      "reward_std": 0.038365477696061134,
      "reward_std_all": 0.04497882630676031,
      "rewards-max/forensics_sinkhorn_reward": 0.9276784360408783,
      "rewards-max/forensics_strict_edge_reward": 0.1433630809187889,
      "rewards-max/forensics_tiou_reward": 0.8517444431781769,
      "rewards/forensics_sinkhorn_reward": 0.6863488852977753,
      "rewards/forensics_strict_edge_reward": 0.013440668524708599,
      "rewards/forensics_tiou_reward": 0.41274283826351166,
      "step": 407
    },
    {
      "completion_length": 21.46875,
      "epoch": 1.707112970711297,
      "grad_norm": 5.95016622543335,
      "kl": 0.3828125,
      "learning_rate": 5.742677824267782e-07,
      "loss": 0.1174,
      "reward": 0.7639044225215912,
      "reward_std": 0.05059765558689833,
      "reward_std_all": 0.048699441365897655,
      "rewards-max/forensics_sinkhorn_reward": 0.9710731208324432,
      "rewards-max/forensics_strict_edge_reward": 0.17697150073945522,
      "rewards-max/forensics_tiou_reward": 0.9524831771850586,
      "rewards/forensics_sinkhorn_reward": 0.6954561471939087,
      "rewards/forensics_strict_edge_reward": 0.011801953543908894,
      "rewards/forensics_tiou_reward": 0.4659448266029358,
      "step": 408
    },
    {
      "completion_length": 19.625,
      "epoch": 1.711297071129707,
      "grad_norm": 5.300720691680908,
      "kl": 0.3447265625,
      "learning_rate": 5.732217573221757e-07,
      "loss": 0.1069,
      "reward": 0.7629981637001038,
      "reward_std": 0.06155920773744583,
      "reward_std_all": 0.05836094543337822,
      "rewards-max/forensics_sinkhorn_reward": 0.981786698102951,
      "rewards-max/forensics_strict_edge_reward": 0.5484400391578674,
      "rewards-max/forensics_tiou_reward": 0.9801054000854492,
      "rewards/forensics_sinkhorn_reward": 0.7456393539905548,
      "rewards/forensics_strict_edge_reward": 0.056348055601119995,
      "rewards/forensics_tiou_reward": 0.5082775205373764,
      "step": 409
    },
    {
      "completion_length": 22.078125,
      "epoch": 1.715481171548117,
      "grad_norm": 5.920342922210693,
      "kl": 0.3076171875,
      "learning_rate": 5.721757322175732e-07,
      "loss": 0.1226,
      "reward": 0.7308003306388855,
      "reward_std": 0.06661459244787693,
      "reward_std_all": 0.0704702278599143,
      "rewards-max/forensics_sinkhorn_reward": 0.9784463942050934,
      "rewards-max/forensics_strict_edge_reward": 0.40730567276477814,
      "rewards-max/forensics_tiou_reward": 0.9648532867431641,
      "rewards/forensics_sinkhorn_reward": 0.6416116654872894,
      "rewards/forensics_strict_edge_reward": 0.025479089468717575,
      "rewards/forensics_tiou_reward": 0.46001701056957245,
      "step": 410
    },
    {
      "completion_length": 20.796875,
      "epoch": 1.719665271966527,
      "grad_norm": 4.567623138427734,
      "kl": 0.26416015625,
      "learning_rate": 5.711297071129707e-07,
      "loss": 0.0461,
      "reward": 0.7677324116230011,
      "reward_std": 0.03545221034437418,
      "reward_std_all": 0.03834181930869818,
      "rewards-max/forensics_sinkhorn_reward": 0.9811290502548218,
      "rewards-max/forensics_strict_edge_reward": 0.43483966588974,
      "rewards-max/forensics_tiou_reward": 0.9779461324214935,
      "rewards/forensics_sinkhorn_reward": 0.7389746606349945,
      "rewards/forensics_strict_edge_reward": 0.03062918409705162,
      "rewards/forensics_tiou_reward": 0.48465201258659363,
      "step": 411
    },
    {
      "completion_length": 24.296875,
      "epoch": 1.723849372384937,
      "grad_norm": 6.117485046386719,
      "kl": 0.41796875,
      "learning_rate": 5.700836820083682e-07,
      "loss": 0.0424,
      "reward": 0.6593337655067444,
      "reward_std": 0.10018067248165607,
      "reward_std_all": 0.12284646928310394,
      "rewards-max/forensics_sinkhorn_reward": 0.9695274829864502,
      "rewards-max/forensics_strict_edge_reward": 0.1799003705382347,
      "rewards-max/forensics_tiou_reward": 0.9471915364265442,
      "rewards/forensics_sinkhorn_reward": 0.5203496515750885,
      "rewards/forensics_strict_edge_reward": 0.011093234177678823,
      "rewards/forensics_tiou_reward": 0.32704365253448486,
      "step": 412
    },
    {
      "completion_length": 19.921875,
      "epoch": 1.7280334728033473,
      "grad_norm": 5.8058881759643555,
      "kl": 0.6181640625,
      "learning_rate": 5.690376569037657e-07,
      "loss": 0.0514,
      "reward": 0.7390938997268677,
      "reward_std": 0.05683173984289169,
      "reward_std_all": 0.06652606092393398,
      "rewards-max/forensics_sinkhorn_reward": 0.9658577144145966,
      "rewards-max/forensics_strict_edge_reward": 0.513735294342041,
      "rewards-max/forensics_tiou_reward": 0.9334705173969269,
      "rewards/forensics_sinkhorn_reward": 0.6986011862754822,
      "rewards/forensics_strict_edge_reward": 0.032357076182961464,
      "rewards/forensics_tiou_reward": 0.4487753212451935,
      "step": 413
    },
    {
      "completion_length": 20.40625,
      "epoch": 1.7322175732217573,
      "grad_norm": 6.32830810546875,
      "kl": 0.2919921875,
      "learning_rate": 5.679916317991632e-07,
      "loss": 0.0377,
      "reward": 0.7472144365310669,
      "reward_std": 0.09663534164428711,
      "reward_std_all": 0.08645780757069588,
      "rewards-max/forensics_sinkhorn_reward": 0.9614188969135284,
      "rewards-max/forensics_strict_edge_reward": 0.3441537916660309,
      "rewards-max/forensics_tiou_reward": 0.9273979067802429,
      "rewards/forensics_sinkhorn_reward": 0.6783054769039154,
      "rewards/forensics_strict_edge_reward": 0.022602534852921963,
      "rewards/forensics_tiou_reward": 0.4261074960231781,
      "step": 414
    },
    {
      "completion_length": 20.109375,
      "epoch": 1.7364016736401675,
      "grad_norm": 4.682008266448975,
      "kl": 0.3115234375,
      "learning_rate": 5.669456066945606e-07,
      "loss": -0.0909,
      "reward": 0.7612695097923279,
      "reward_std": 0.03221470769494772,
      "reward_std_all": 0.03943778574466705,
      "rewards-max/forensics_sinkhorn_reward": 0.9771714806556702,
      "rewards-max/forensics_strict_edge_reward": 0.43558479845523834,
      "rewards-max/forensics_tiou_reward": 0.9608210027217865,
      "rewards/forensics_sinkhorn_reward": 0.7211789488792419,
      "rewards/forensics_strict_edge_reward": 0.03206549026072025,
      "rewards/forensics_tiou_reward": 0.463705837726593,
      "step": 415
    },
    {
      "completion_length": 20.28125,
      "epoch": 1.7405857740585775,
      "grad_norm": 8.636760711669922,
      "kl": 0.29736328125,
      "learning_rate": 5.65899581589958e-07,
      "loss": 0.0505,
      "reward": 0.7568645179271698,
      "reward_std": 0.04761507082730532,
      "reward_std_all": 0.048619190230965614,
      "rewards-max/forensics_sinkhorn_reward": 0.9283100962638855,
      "rewards-max/forensics_strict_edge_reward": 0.2986167371273041,
      "rewards-max/forensics_tiou_reward": 0.8620910942554474,
      "rewards/forensics_sinkhorn_reward": 0.6663349568843842,
      "rewards/forensics_strict_edge_reward": 0.015732870902866125,
      "rewards/forensics_tiou_reward": 0.3792997896671295,
      "step": 416
    },
    {
      "completion_length": 19.90625,
      "epoch": 1.7447698744769875,
      "grad_norm": 7.704935073852539,
      "kl": 0.3623046875,
      "learning_rate": 5.648535564853556e-07,
      "loss": 0.0757,
      "reward": 0.7140272557735443,
      "reward_std": 0.09223879314959049,
      "reward_std_all": 0.09332720935344696,
      "rewards-max/forensics_sinkhorn_reward": 0.959135890007019,
      "rewards-max/forensics_strict_edge_reward": 0.34998635947704315,
      "rewards-max/forensics_tiou_reward": 0.9274216592311859,
      "rewards/forensics_sinkhorn_reward": 0.6857978403568268,
      "rewards/forensics_strict_edge_reward": 0.027134311385452747,
      "rewards/forensics_tiou_reward": 0.4548293650150299,
      "step": 417
    },
    {
      "completion_length": 23.9375,
      "epoch": 1.7489539748953975,
      "grad_norm": 6.051887512207031,
      "kl": 0.30859375,
      "learning_rate": 5.638075313807531e-07,
      "loss": -0.0085,
      "reward": 0.7101353108882904,
      "reward_std": 0.07606969401240349,
      "reward_std_all": 0.08402223140001297,
      "rewards-max/forensics_sinkhorn_reward": 0.9840258061885834,
      "rewards-max/forensics_strict_edge_reward": 0.3226740062236786,
      "rewards-max/forensics_tiou_reward": 0.984986275434494,
      "rewards/forensics_sinkhorn_reward": 0.6390396952629089,
      "rewards/forensics_strict_edge_reward": 0.02565435878932476,
      "rewards/forensics_tiou_reward": 0.40984347462654114,
      "step": 418
    },
    {
      "completion_length": 19.8125,
      "epoch": 1.7531380753138075,
      "grad_norm": 8.380234718322754,
      "kl": 0.421875,
      "learning_rate": 5.627615062761506e-07,
      "loss": 0.0273,
      "reward": 0.7489690482616425,
      "reward_std": 0.05641471967101097,
      "reward_std_all": 0.05635879188776016,
      "rewards-max/forensics_sinkhorn_reward": 0.9652703106403351,
      "rewards-max/forensics_strict_edge_reward": 0.37661072611808777,
      "rewards-max/forensics_tiou_reward": 0.939361035823822,
      "rewards/forensics_sinkhorn_reward": 0.7273151576519012,
      "rewards/forensics_strict_edge_reward": 0.018468322232365608,
      "rewards/forensics_tiou_reward": 0.47140468657016754,
      "step": 419
    },
    {
      "completion_length": 20.375,
      "epoch": 1.7573221757322175,
      "grad_norm": 4.860785007476807,
      "kl": 0.314453125,
      "learning_rate": 5.617154811715481e-07,
      "loss": 0.0417,
      "reward": 0.75266894698143,
      "reward_std": 0.057075755670666695,
      "reward_std_all": 0.05593211576342583,
      "rewards-max/forensics_sinkhorn_reward": 0.9707319140434265,
      "rewards-max/forensics_strict_edge_reward": 0.36953114718198776,
      "rewards-max/forensics_tiou_reward": 0.9349538683891296,
      "rewards/forensics_sinkhorn_reward": 0.7144066393375397,
      "rewards/forensics_strict_edge_reward": 0.03788252058438957,
      "rewards/forensics_tiou_reward": 0.4587303251028061,
      "step": 420
    },
    {
      "completion_length": 19.65625,
      "epoch": 1.7615062761506275,
      "grad_norm": 6.377721309661865,
      "kl": 0.2841796875,
      "learning_rate": 5.606694560669456e-07,
      "loss": 0.0382,
      "reward": 0.777081698179245,
      "reward_std": 0.03098463104106486,
      "reward_std_all": 0.03151174634695053,
      "rewards-max/forensics_sinkhorn_reward": 0.9541351497173309,
      "rewards-max/forensics_strict_edge_reward": 0.3607722967863083,
      "rewards-max/forensics_tiou_reward": 0.9117966890335083,
      "rewards/forensics_sinkhorn_reward": 0.7128230035305023,
      "rewards/forensics_strict_edge_reward": 0.021095373667776585,
      "rewards/forensics_tiou_reward": 0.4403914660215378,
      "step": 421
    },
    {
      "completion_length": 21.359375,
      "epoch": 1.7656903765690377,
      "grad_norm": 6.345876693725586,
      "kl": 0.25927734375,
      "learning_rate": 5.596234309623431e-07,
      "loss": 0.1063,
      "reward": 0.7309178709983826,
      "reward_std": 0.07995044067502022,
      "reward_std_all": 0.07768882811069489,
      "rewards-max/forensics_sinkhorn_reward": 0.9344325661659241,
      "rewards-max/forensics_strict_edge_reward": 0.00971356243826449,
      "rewards-max/forensics_tiou_reward": 0.8533742725849152,
      "rewards/forensics_sinkhorn_reward": 0.665178120136261,
      "rewards/forensics_strict_edge_reward": 0.0007282289443537593,
      "rewards/forensics_tiou_reward": 0.4072566330432892,
      "step": 422
    },
    {
      "completion_length": 21.78125,
      "epoch": 1.7698744769874477,
      "grad_norm": 7.231476306915283,
      "kl": 0.259765625,
      "learning_rate": 5.585774058577406e-07,
      "loss": 0.2418,
      "reward": 0.7016841471195221,
      "reward_std": 0.08355272840708494,
      "reward_std_all": 0.09289229661226273,
      "rewards-max/forensics_sinkhorn_reward": 0.9865658283233643,
      "rewards-max/forensics_strict_edge_reward": 0.7580138444900513,
      "rewards-max/forensics_tiou_reward": 0.9914429187774658,
      "rewards/forensics_sinkhorn_reward": 0.6035438477993011,
      "rewards/forensics_strict_edge_reward": 0.06960518658161163,
      "rewards/forensics_tiou_reward": 0.4187684804201126,
      "step": 423
    },
    {
      "completion_length": 21.4375,
      "epoch": 1.7740585774058577,
      "grad_norm": 7.466754913330078,
      "kl": 0.2470703125,
      "learning_rate": 5.575313807531381e-07,
      "loss": 0.2017,
      "reward": 0.6575785279273987,
      "reward_std": 0.09087914973497391,
      "reward_std_all": 0.12090572342276573,
      "rewards-max/forensics_sinkhorn_reward": 0.9379916787147522,
      "rewards-max/forensics_strict_edge_reward": 0.09246509149670601,
      "rewards-max/forensics_tiou_reward": 0.8637826144695282,
      "rewards/forensics_sinkhorn_reward": 0.5784932971000671,
      "rewards/forensics_strict_edge_reward": 0.0049571453128010035,
      "rewards/forensics_tiou_reward": 0.3620518743991852,
      "step": 424
    },
    {
      "completion_length": 21.3125,
      "epoch": 1.778242677824268,
      "grad_norm": 7.047855377197266,
      "kl": 0.4306640625,
      "learning_rate": 5.564853556485354e-07,
      "loss": -0.0005,
      "reward": 0.7197735011577606,
      "reward_std": 0.07564081996679306,
      "reward_std_all": 0.08004079014062881,
      "rewards-max/forensics_sinkhorn_reward": 0.9691689610481262,
      "rewards-max/forensics_strict_edge_reward": 0.4231441020965576,
      "rewards-max/forensics_tiou_reward": 0.9500760138034821,
      "rewards/forensics_sinkhorn_reward": 0.6331100165843964,
      "rewards/forensics_strict_edge_reward": 0.027400503866374493,
      "rewards/forensics_tiou_reward": 0.3795176148414612,
      "step": 425
    },
    {
      "completion_length": 19.796875,
      "epoch": 1.782426778242678,
      "grad_norm": 7.681845664978027,
      "kl": 0.3125,
      "learning_rate": 5.55439330543933e-07,
      "loss": -0.0394,
      "reward": 0.7604480981826782,
      "reward_std": 0.015662764199078083,
      "reward_std_all": 0.03207329474389553,
      "rewards-max/forensics_sinkhorn_reward": 0.9769213497638702,
      "rewards-max/forensics_strict_edge_reward": 0.6762872636318207,
      "rewards-max/forensics_tiou_reward": 0.9704182147979736,
      "rewards/forensics_sinkhorn_reward": 0.7329451441764832,
      "rewards/forensics_strict_edge_reward": 0.04248763993382454,
      "rewards/forensics_tiou_reward": 0.48454591631889343,
      "step": 426
    },
    {
      "completion_length": 21.671875,
      "epoch": 1.786610878661088,
      "grad_norm": 6.261417388916016,
      "kl": 0.271484375,
      "learning_rate": 5.543933054393305e-07,
      "loss": 0.1627,
      "reward": 0.7548426389694214,
      "reward_std": 0.06545221246778965,
      "reward_std_all": 0.06229640915989876,
      "rewards-max/forensics_sinkhorn_reward": 0.9691084921360016,
      "rewards-max/forensics_strict_edge_reward": 0.2655768468976021,
      "rewards-max/forensics_tiou_reward": 0.9519563913345337,
      "rewards/forensics_sinkhorn_reward": 0.6782117486000061,
      "rewards/forensics_strict_edge_reward": 0.015554960817098618,
      "rewards/forensics_tiou_reward": 0.44996584951877594,
      "step": 427
    },
    {
      "completion_length": 21.203125,
      "epoch": 1.790794979079498,
      "grad_norm": 5.917301177978516,
      "kl": 0.4208984375,
      "learning_rate": 5.53347280334728e-07,
      "loss": 0.1132,
      "reward": 0.742215096950531,
      "reward_std": 0.05201033316552639,
      "reward_std_all": 0.05600925721228123,
      "rewards-max/forensics_sinkhorn_reward": 0.9470334351062775,
      "rewards-max/forensics_strict_edge_reward": 0.08486727625131607,
      "rewards-max/forensics_tiou_reward": 0.8883429169654846,
      "rewards/forensics_sinkhorn_reward": 0.6698501706123352,
      "rewards/forensics_strict_edge_reward": 0.007614918984472752,
      "rewards/forensics_tiou_reward": 0.4364684671163559,
      "step": 428
    },
    {
      "completion_length": 22.3125,
      "epoch": 1.794979079497908,
      "grad_norm": 5.891467094421387,
      "kl": 0.265625,
      "learning_rate": 5.523012552301255e-07,
      "loss": 0.0224,
      "reward": 0.7086846828460693,
      "reward_std": 0.05951886065304279,
      "reward_std_all": 0.07836328446865082,
      "rewards-max/forensics_sinkhorn_reward": 0.9774833023548126,
      "rewards-max/forensics_strict_edge_reward": 0.45164985954761505,
      "rewards-max/forensics_tiou_reward": 0.9694257974624634,
      "rewards/forensics_sinkhorn_reward": 0.6510060727596283,
      "rewards/forensics_strict_edge_reward": 0.04208669438958168,
      "rewards/forensics_tiou_reward": 0.43844640254974365,
      "step": 429
    },
    {
      "completion_length": 22.640625,
      "epoch": 1.799163179916318,
      "grad_norm": 6.0302581787109375,
      "kl": 0.2841796875,
      "learning_rate": 5.51255230125523e-07,
      "loss": 0.104,
      "reward": 0.7475929856300354,
      "reward_std": 0.06426284834742546,
      "reward_std_all": 0.06338583491742611,
      "rewards-max/forensics_sinkhorn_reward": 0.953134149312973,
      "rewards-max/forensics_strict_edge_reward": 0.3340276926755905,
      "rewards-max/forensics_tiou_reward": 0.9162312150001526,
      "rewards/forensics_sinkhorn_reward": 0.6659985184669495,
      "rewards/forensics_strict_edge_reward": 0.025768672581762075,
      "rewards/forensics_tiou_reward": 0.40293389558792114,
      "step": 430
    },
    {
      "completion_length": 21.6875,
      "epoch": 1.803347280334728,
      "grad_norm": 4.725098133087158,
      "kl": 0.3955078125,
      "learning_rate": 5.502092050209205e-07,
      "loss": 0.1109,
      "reward": 0.7452191710472107,
      "reward_std": 0.05488232895731926,
      "reward_std_all": 0.058290936052799225,
      "rewards-max/forensics_sinkhorn_reward": 0.9651280343532562,
      "rewards-max/forensics_strict_edge_reward": 0.2516818791627884,
      "rewards-max/forensics_tiou_reward": 0.9393100142478943,
      "rewards/forensics_sinkhorn_reward": 0.6866897642612457,
      "rewards/forensics_strict_edge_reward": 0.023434459697455168,
      "rewards/forensics_tiou_reward": 0.5074435919523239,
      "step": 431
    },
    {
      "completion_length": 22.078125,
      "epoch": 1.8075313807531381,
      "grad_norm": 8.863959312438965,
      "kl": 0.4814453125,
      "learning_rate": 5.49163179916318e-07,
      "loss": 0.1339,
      "reward": 0.7087978720664978,
      "reward_std": 0.05688724108040333,
      "reward_std_all": 0.07404317520558834,
      "rewards-max/forensics_sinkhorn_reward": 0.9562842547893524,
      "rewards-max/forensics_strict_edge_reward": 0.2896316945552826,
      "rewards-max/forensics_tiou_reward": 0.9185910224914551,
      "rewards/forensics_sinkhorn_reward": 0.6354124546051025,
      "rewards/forensics_strict_edge_reward": 0.022472009994089603,
      "rewards/forensics_tiou_reward": 0.38596197962760925,
      "step": 432
    },
    {
      "completion_length": 20.71875,
      "epoch": 1.8117154811715481,
      "grad_norm": 4.542468070983887,
      "kl": 0.337890625,
      "learning_rate": 5.481171548117156e-07,
      "loss": 0.0316,
      "reward": 0.7520789504051208,
      "reward_std": 0.06956644542515278,
      "reward_std_all": 0.06522995326668024,
      "rewards-max/forensics_sinkhorn_reward": 0.9800777733325958,
      "rewards-max/forensics_strict_edge_reward": 0.42625628411769867,
      "rewards-max/forensics_tiou_reward": 0.9693683385848999,
      "rewards/forensics_sinkhorn_reward": 0.7221168279647827,
      "rewards/forensics_strict_edge_reward": 0.0530579574406147,
      "rewards/forensics_tiou_reward": 0.5079831779003143,
      "step": 433
    },
    {
      "completion_length": 27.96875,
      "epoch": 1.8158995815899581,
      "grad_norm": 6.601071357727051,
      "kl": 0.2978515625,
      "learning_rate": 5.47071129707113e-07,
      "loss": 0.0151,
      "reward": 0.6398802101612091,
      "reward_std": 0.1041659265756607,
      "reward_std_all": 0.13110711425542831,
      "rewards-max/forensics_sinkhorn_reward": 0.9580026865005493,
      "rewards-max/forensics_strict_edge_reward": 0.2896316945552826,
      "rewards-max/forensics_tiou_reward": 0.8996796011924744,
      "rewards/forensics_sinkhorn_reward": 0.5465076267719269,
      "rewards/forensics_strict_edge_reward": 0.013962168246507645,
      "rewards/forensics_tiou_reward": 0.3084364980459213,
      "step": 434
    },
    {
      "completion_length": 22.46875,
      "epoch": 1.8200836820083683,
      "grad_norm": 5.733487606048584,
      "kl": 0.2939453125,
      "learning_rate": 5.460251046025104e-07,
      "loss": 0.1339,
      "reward": 0.7706891894340515,
      "reward_std": 0.05534241907298565,
      "reward_std_all": 0.050588177517056465,
      "rewards-max/forensics_sinkhorn_reward": 0.9776870310306549,
      "rewards-max/forensics_strict_edge_reward": 0.4856214374303818,
      "rewards-max/forensics_tiou_reward": 0.9682987332344055,
      "rewards/forensics_sinkhorn_reward": 0.6986582577228546,
      "rewards/forensics_strict_edge_reward": 0.05251770094037056,
      "rewards/forensics_tiou_reward": 0.5307221412658691,
      "step": 435
    },
    {
      "completion_length": 24.203125,
      "epoch": 1.8242677824267783,
      "grad_norm": 6.367354869842529,
      "kl": 0.4365234375,
      "learning_rate": 5.449790794979079e-07,
      "loss": 0.1294,
      "reward": 0.7140898108482361,
      "reward_std": 0.06237117201089859,
      "reward_std_all": 0.07925029844045639,
      "rewards-max/forensics_sinkhorn_reward": 0.9629834294319153,
      "rewards-max/forensics_strict_edge_reward": 0.25900475680828094,
      "rewards-max/forensics_tiou_reward": 0.935257077217102,
      "rewards/forensics_sinkhorn_reward": 0.6354672312736511,
      "rewards/forensics_strict_edge_reward": 0.02185205928981304,
      "rewards/forensics_tiou_reward": 0.434331551194191,
      "step": 436
    },
    {
      "completion_length": 20.8125,
      "epoch": 1.8284518828451883,
      "grad_norm": 4.634284973144531,
      "kl": 0.34765625,
      "learning_rate": 5.439330543933054e-07,
      "loss": -0.008,
      "reward": 0.7764453887939453,
      "reward_std": 0.03617231734097004,
      "reward_std_all": 0.035260921344161034,
      "rewards-max/forensics_sinkhorn_reward": 0.9850653409957886,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9893198311328888,
      "rewards/forensics_sinkhorn_reward": 0.7329888343811035,
      "rewards/forensics_strict_edge_reward": 0.06792524829506874,
      "rewards/forensics_tiou_reward": 0.5270535945892334,
      "step": 437
    },
    {
      "completion_length": 19.828125,
      "epoch": 1.8326359832635983,
      "grad_norm": 5.421773433685303,
      "kl": 0.3291015625,
      "learning_rate": 5.428870292887029e-07,
      "loss": 0.0775,
      "reward": 0.7001250386238098,
      "reward_std": 0.04934663139283657,
      "reward_std_all": 0.07956377230584621,
      "rewards-max/forensics_sinkhorn_reward": 0.9662443101406097,
      "rewards-max/forensics_strict_edge_reward": 0.42822691053152084,
      "rewards-max/forensics_tiou_reward": 0.9402118027210236,
      "rewards/forensics_sinkhorn_reward": 0.6541924178600311,
      "rewards/forensics_strict_edge_reward": 0.046563167590647936,
      "rewards/forensics_tiou_reward": 0.41654519736766815,
      "step": 438
    },
    {
      "completion_length": 20.90625,
      "epoch": 1.8368200836820083,
      "grad_norm": 5.663573741912842,
      "kl": 0.3525390625,
      "learning_rate": 5.418410041841004e-07,
      "loss": 0.1002,
      "reward": 0.7493637800216675,
      "reward_std": 0.06266222521662712,
      "reward_std_all": 0.06119091250002384,
      "rewards-max/forensics_sinkhorn_reward": 0.973837822675705,
      "rewards-max/forensics_strict_edge_reward": 0.5500316619873047,
      "rewards-max/forensics_tiou_reward": 0.9655985832214355,
      "rewards/forensics_sinkhorn_reward": 0.709044337272644,
      "rewards/forensics_strict_edge_reward": 0.03328697755932808,
      "rewards/forensics_tiou_reward": 0.4823470115661621,
      "step": 439
    },
    {
      "completion_length": 21.4375,
      "epoch": 1.8410041841004183,
      "grad_norm": 5.214901447296143,
      "kl": 0.2939453125,
      "learning_rate": 5.407949790794979e-07,
      "loss": 0.1949,
      "reward": 0.7314854860305786,
      "reward_std": 0.10934564471244812,
      "reward_std_all": 0.09889092110097408,
      "rewards-max/forensics_sinkhorn_reward": 0.9521254897117615,
      "rewards-max/forensics_strict_edge_reward": 0.14064522460103035,
      "rewards-max/forensics_tiou_reward": 0.8855640590190887,
      "rewards/forensics_sinkhorn_reward": 0.71331387758255,
      "rewards/forensics_strict_edge_reward": 0.0159152802079916,
      "rewards/forensics_tiou_reward": 0.489361509680748,
      "step": 440
    },
    {
      "completion_length": 20.015625,
      "epoch": 1.8451882845188283,
      "grad_norm": 5.896451950073242,
      "kl": 0.46875,
      "learning_rate": 5.397489539748954e-07,
      "loss": 0.0141,
      "reward": 0.6930214464664459,
      "reward_std": 0.07091425359249115,
      "reward_std_all": 0.08950716257095337,
      "rewards-max/forensics_sinkhorn_reward": 0.9772482514381409,
      "rewards-max/forensics_strict_edge_reward": 0.3005804941058159,
      "rewards-max/forensics_tiou_reward": 0.9652259051799774,
      "rewards/forensics_sinkhorn_reward": 0.634372740983963,
      "rewards/forensics_strict_edge_reward": 0.01806088211014867,
      "rewards/forensics_tiou_reward": 0.3869343101978302,
      "step": 441
    },
    {
      "completion_length": 23.796875,
      "epoch": 1.8493723849372385,
      "grad_norm": 6.12571907043457,
      "kl": 0.3642578125,
      "learning_rate": 5.387029288702928e-07,
      "loss": 0.1019,
      "reward": 0.7153217196464539,
      "reward_std": 0.08105314522981644,
      "reward_std_all": 0.08278217166662216,
      "rewards-max/forensics_sinkhorn_reward": 0.9430942535400391,
      "rewards-max/forensics_strict_edge_reward": 0.44867993146181107,
      "rewards-max/forensics_tiou_reward": 0.8799863159656525,
      "rewards/forensics_sinkhorn_reward": 0.6356790661811829,
      "rewards/forensics_strict_edge_reward": 0.03107314556837082,
      "rewards/forensics_tiou_reward": 0.3761803209781647,
      "step": 442
    },
    {
      "completion_length": 22.65625,
      "epoch": 1.8535564853556485,
      "grad_norm": 4.384087562561035,
      "kl": 0.33203125,
      "learning_rate": 5.376569037656904e-07,
      "loss": 0.0872,
      "reward": 0.7491435408592224,
      "reward_std": 0.06578771770000458,
      "reward_std_all": 0.06337147206068039,
      "rewards-max/forensics_sinkhorn_reward": 0.9646528959274292,
      "rewards-max/forensics_strict_edge_reward": 0.14728037267923355,
      "rewards-max/forensics_tiou_reward": 0.9364263713359833,
      "rewards/forensics_sinkhorn_reward": 0.6716030836105347,
      "rewards/forensics_strict_edge_reward": 0.008917439728975296,
      "rewards/forensics_tiou_reward": 0.4349019229412079,
      "step": 443
    },
    {
      "completion_length": 20.9375,
      "epoch": 1.8577405857740585,
      "grad_norm": 5.977875232696533,
      "kl": 0.3388671875,
      "learning_rate": 5.366108786610879e-07,
      "loss": -0.0807,
      "reward": 0.661528617143631,
      "reward_std": 0.06666721031069756,
      "reward_std_all": 0.10532507300376892,
      "rewards-max/forensics_sinkhorn_reward": 0.9857012033462524,
      "rewards-max/forensics_strict_edge_reward": 0.4991503059864044,
      "rewards-max/forensics_tiou_reward": 0.9907495677471161,
      "rewards/forensics_sinkhorn_reward": 0.6147383749485016,
      "rewards/forensics_strict_edge_reward": 0.05374870449304581,
      "rewards/forensics_tiou_reward": 0.3849962502717972,
      "step": 444
    },
    {
      "completion_length": 20.640625,
      "epoch": 1.8619246861924688,
      "grad_norm": 12.691442489624023,
      "kl": 1.125,
      "learning_rate": 5.355648535564853e-07,
      "loss": 0.0524,
      "reward": 0.747626394033432,
      "reward_std": 0.07090195268392563,
      "reward_std_all": 0.06781614199280739,
      "rewards-max/forensics_sinkhorn_reward": 0.9617370665073395,
      "rewards-max/forensics_strict_edge_reward": 0.3055209591984749,
      "rewards-max/forensics_tiou_reward": 0.9319975078105927,
      "rewards/forensics_sinkhorn_reward": 0.6757144331932068,
      "rewards/forensics_strict_edge_reward": 0.03560714889317751,
      "rewards/forensics_tiou_reward": 0.38359084725379944,
      "step": 445
    },
    {
      "completion_length": 20.828125,
      "epoch": 1.8661087866108788,
      "grad_norm": 8.014305114746094,
      "kl": 0.3828125,
      "learning_rate": 5.345188284518828e-07,
      "loss": 0.1317,
      "reward": 0.7031209766864777,
      "reward_std": 0.09037108346819878,
      "reward_std_all": 0.09377722069621086,
      "rewards-max/forensics_sinkhorn_reward": 0.9225395917892456,
      "rewards-max/forensics_strict_edge_reward": 0.055307429283857346,
      "rewards-max/forensics_tiou_reward": 0.8365030586719513,
      "rewards/forensics_sinkhorn_reward": 0.6368383169174194,
      "rewards/forensics_strict_edge_reward": 0.005368647398427129,
      "rewards/forensics_tiou_reward": 0.3880458027124405,
      "step": 446
    },
    {
      "completion_length": 19.828125,
      "epoch": 1.8702928870292888,
      "grad_norm": 5.498876094818115,
      "kl": 0.259765625,
      "learning_rate": 5.334728033472803e-07,
      "loss": 0.008,
      "reward": 0.7871727347373962,
      "reward_std": 0.029886456206440926,
      "reward_std_all": 0.028035522205755115,
      "rewards-max/forensics_sinkhorn_reward": 0.9862222969532013,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9923284947872162,
      "rewards/forensics_sinkhorn_reward": 0.7902349233627319,
      "rewards/forensics_strict_edge_reward": 0.0563915204256773,
      "rewards/forensics_tiou_reward": 0.574347734451294,
      "step": 447
    },
    {
      "completion_length": 21.015625,
      "epoch": 1.8744769874476988,
      "grad_norm": 6.647215366363525,
      "kl": 0.3759765625,
      "learning_rate": 5.324267782426778e-07,
      "loss": -0.0135,
      "reward": 0.7235544919967651,
      "reward_std": 0.05759909562766552,
      "reward_std_all": 0.07082310877740383,
      "rewards-max/forensics_sinkhorn_reward": 0.9524634182453156,
      "rewards-max/forensics_strict_edge_reward": 0.1665339386672713,
      "rewards-max/forensics_tiou_reward": 0.8830944895744324,
      "rewards/forensics_sinkhorn_reward": 0.6769288182258606,
      "rewards/forensics_strict_edge_reward": 0.007518196796809207,
      "rewards/forensics_tiou_reward": 0.4155212342739105,
      "step": 448
    },
    {
      "completion_length": 21.953125,
      "epoch": 1.8786610878661087,
      "grad_norm": 4.99420166015625,
      "kl": 0.3994140625,
      "learning_rate": 5.313807531380753e-07,
      "loss": 0.1265,
      "reward": 0.7409491539001465,
      "reward_std": 0.08193589001893997,
      "reward_std_all": 0.07669999450445175,
      "rewards-max/forensics_sinkhorn_reward": 0.9858828485012054,
      "rewards-max/forensics_strict_edge_reward": 0.37195122241973877,
      "rewards-max/forensics_tiou_reward": 0.9921501576900482,
      "rewards/forensics_sinkhorn_reward": 0.6993051171302795,
      "rewards/forensics_strict_edge_reward": 0.04403379186987877,
      "rewards/forensics_tiou_reward": 0.4886385649442673,
      "step": 449
    },
    {
      "completion_length": 20.546875,
      "epoch": 1.8828451882845187,
      "grad_norm": 5.301395416259766,
      "kl": 0.310546875,
      "learning_rate": 5.303347280334727e-07,
      "loss": 0.1115,
      "reward": 0.7783399224281311,
      "reward_std": 0.02693430334329605,
      "reward_std_all": 0.030503767542541027,
      "rewards-max/forensics_sinkhorn_reward": 0.9727223515510559,
      "rewards-max/forensics_strict_edge_reward": 0.44892527163028717,
      "rewards-max/forensics_tiou_reward": 0.9564975202083588,
      "rewards/forensics_sinkhorn_reward": 0.7600750029087067,
      "rewards/forensics_strict_edge_reward": 0.043656984344124794,
      "rewards/forensics_tiou_reward": 0.5883914232254028,
      "step": 450
    },
    {
      "completion_length": 22.234375,
      "epoch": 1.8870292887029287,
      "grad_norm": 5.276987552642822,
      "kl": 0.34375,
      "learning_rate": 5.292887029288703e-07,
      "loss": 0.0481,
      "reward": 0.7559427618980408,
      "reward_std": 0.0755676981061697,
      "reward_std_all": 0.06883447617292404,
      "rewards-max/forensics_sinkhorn_reward": 0.9595631062984467,
      "rewards-max/forensics_strict_edge_reward": 0.30970802903175354,
      "rewards-max/forensics_tiou_reward": 0.9240466952323914,
      "rewards/forensics_sinkhorn_reward": 0.7039682865142822,
      "rewards/forensics_strict_edge_reward": 0.028775731101632118,
      "rewards/forensics_tiou_reward": 0.4654434770345688,
      "step": 451
    },
    {
      "completion_length": 21.40625,
      "epoch": 1.891213389121339,
      "grad_norm": 5.630591869354248,
      "kl": 0.28173828125,
      "learning_rate": 5.282426778242678e-07,
      "loss": 0.1642,
      "reward": 0.7187274098396301,
      "reward_std": 0.07899919152259827,
      "reward_std_all": 0.08276226185262203,
      "rewards-max/forensics_sinkhorn_reward": 0.9803001284599304,
      "rewards-max/forensics_strict_edge_reward": 0.5347417257726192,
      "rewards-max/forensics_tiou_reward": 0.9680155813694,
      "rewards/forensics_sinkhorn_reward": 0.6740018427371979,
      "rewards/forensics_strict_edge_reward": 0.036413877503946424,
      "rewards/forensics_tiou_reward": 0.4799390137195587,
      "step": 452
    },
    {
      "completion_length": 21.328125,
      "epoch": 1.895397489539749,
      "grad_norm": 4.201014995574951,
      "kl": 0.44140625,
      "learning_rate": 5.271966527196653e-07,
      "loss": 0.156,
      "reward": 0.7402606010437012,
      "reward_std": 0.06317829340696335,
      "reward_std_all": 0.06663721986114979,
      "rewards-max/forensics_sinkhorn_reward": 0.9569255411624908,
      "rewards-max/forensics_strict_edge_reward": 0.3480503261089325,
      "rewards-max/forensics_tiou_reward": 0.9173626005649567,
      "rewards/forensics_sinkhorn_reward": 0.681544691324234,
      "rewards/forensics_strict_edge_reward": 0.019858059473335743,
      "rewards/forensics_tiou_reward": 0.4245496839284897,
      "step": 453
    },
    {
      "completion_length": 21.140625,
      "epoch": 1.899581589958159,
      "grad_norm": 4.211976051330566,
      "kl": 0.2626953125,
      "learning_rate": 5.261506276150628e-07,
      "loss": 0.0714,
      "reward": 0.7253749370574951,
      "reward_std": 0.0692231897264719,
      "reward_std_all": 0.07393097877502441,
      "rewards-max/forensics_sinkhorn_reward": 0.95315220952034,
      "rewards-max/forensics_strict_edge_reward": 0.3788442946970463,
      "rewards-max/forensics_tiou_reward": 0.9047722518444061,
      "rewards/forensics_sinkhorn_reward": 0.6525472104549408,
      "rewards/forensics_strict_edge_reward": 0.038542820839211345,
      "rewards/forensics_tiou_reward": 0.42887625098228455,
      "step": 454
    },
    {
      "completion_length": 20.875,
      "epoch": 1.9037656903765692,
      "grad_norm": 5.20354700088501,
      "kl": 0.3349609375,
      "learning_rate": 5.251046025104602e-07,
      "loss": 0.1772,
      "reward": 0.7414305508136749,
      "reward_std": 0.08485645987093449,
      "reward_std_all": 0.07895162515342236,
      "rewards-max/forensics_sinkhorn_reward": 0.9225400388240814,
      "rewards-max/forensics_strict_edge_reward": 0.038233269937336445,
      "rewards-max/forensics_tiou_reward": 0.8274078965187073,
      "rewards/forensics_sinkhorn_reward": 0.6880348920822144,
      "rewards/forensics_strict_edge_reward": 0.0033855962101370096,
      "rewards/forensics_tiou_reward": 0.4297402948141098,
      "step": 455
    },
    {
      "completion_length": 20.890625,
      "epoch": 1.9079497907949792,
      "grad_norm": 5.544978141784668,
      "kl": 0.3212890625,
      "learning_rate": 5.240585774058577e-07,
      "loss": 0.2131,
      "reward": 0.7314082682132721,
      "reward_std": 0.06857435032725334,
      "reward_std_all": 0.07083902880549431,
      "rewards-max/forensics_sinkhorn_reward": 0.98075732588768,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9723107516765594,
      "rewards/forensics_sinkhorn_reward": 0.6258952915668488,
      "rewards/forensics_strict_edge_reward": 0.024760075844824314,
      "rewards/forensics_tiou_reward": 0.3864828646183014,
      "step": 456
    },
    {
      "completion_length": 20.21875,
      "epoch": 1.9121338912133892,
      "grad_norm": 5.82332706451416,
      "kl": 0.3486328125,
      "learning_rate": 5.230125523012552e-07,
      "loss": -0.0166,
      "reward": 0.7661001086235046,
      "reward_std": 0.02243304718285799,
      "reward_std_all": 0.031862217001616955,
      "rewards-max/forensics_sinkhorn_reward": 0.9696182906627655,
      "rewards-max/forensics_strict_edge_reward": 0.3340276926755905,
      "rewards-max/forensics_tiou_reward": 0.9544276595115662,
      "rewards/forensics_sinkhorn_reward": 0.7448920011520386,
      "rewards/forensics_strict_edge_reward": 0.01471689622849226,
      "rewards/forensics_tiou_reward": 0.4924495220184326,
      "step": 457
    },
    {
      "completion_length": 22.71875,
      "epoch": 1.9163179916317992,
      "grad_norm": 6.230825901031494,
      "kl": 0.322265625,
      "learning_rate": 5.219665271966526e-07,
      "loss": 0.0234,
      "reward": 0.7108881175518036,
      "reward_std": 0.09111040085554123,
      "reward_std_all": 0.09199612960219383,
      "rewards-max/forensics_sinkhorn_reward": 0.9612981975078583,
      "rewards-max/forensics_strict_edge_reward": 0.18036524578928947,
      "rewards-max/forensics_tiou_reward": 0.9323085844516754,
      "rewards/forensics_sinkhorn_reward": 0.6194021701812744,
      "rewards/forensics_strict_edge_reward": 0.010019023437052965,
      "rewards/forensics_tiou_reward": 0.38742272555828094,
      "step": 458
    },
    {
      "completion_length": 20.8125,
      "epoch": 1.9205020920502092,
      "grad_norm": 5.278504848480225,
      "kl": 0.22705078125,
      "learning_rate": 5.209205020920501e-07,
      "loss": 0.1069,
      "reward": 0.6997349262237549,
      "reward_std": 0.06109362468123436,
      "reward_std_all": 0.08039593696594238,
      "rewards-max/forensics_sinkhorn_reward": 0.9555002748966217,
      "rewards-max/forensics_strict_edge_reward": 0.1572493463754654,
      "rewards-max/forensics_tiou_reward": 0.9035914540290833,
      "rewards/forensics_sinkhorn_reward": 0.635039746761322,
      "rewards/forensics_strict_edge_reward": 0.007920751115307212,
      "rewards/forensics_tiou_reward": 0.3568096458911896,
      "step": 459
    },
    {
      "completion_length": 20.96875,
      "epoch": 1.9246861924686192,
      "grad_norm": 5.390693187713623,
      "kl": 0.3251953125,
      "learning_rate": 5.198744769874477e-07,
      "loss": 0.1134,
      "reward": 0.7670577168464661,
      "reward_std": 0.04605993069708347,
      "reward_std_all": 0.04468523524701595,
      "rewards-max/forensics_sinkhorn_reward": 0.9635559320449829,
      "rewards-max/forensics_strict_edge_reward": 0.48336396366357803,
      "rewards-max/forensics_tiou_reward": 0.9160311818122864,
      "rewards/forensics_sinkhorn_reward": 0.6946019530296326,
      "rewards/forensics_strict_edge_reward": 0.04540796508081257,
      "rewards/forensics_tiou_reward": 0.4707763195037842,
      "step": 460
    },
    {
      "completion_length": 21.375,
      "epoch": 1.9288702928870292,
      "grad_norm": 5.29003381729126,
      "kl": 0.3359375,
      "learning_rate": 5.188284518828452e-07,
      "loss": 0.1324,
      "reward": 0.6964645981788635,
      "reward_std": 0.09039533883333206,
      "reward_std_all": 0.10282416641712189,
      "rewards-max/forensics_sinkhorn_reward": 0.9678497910499573,
      "rewards-max/forensics_strict_edge_reward": 0.270952470600605,
      "rewards-max/forensics_tiou_reward": 0.9437710344791412,
      "rewards/forensics_sinkhorn_reward": 0.6524396538734436,
      "rewards/forensics_strict_edge_reward": 0.012850911356508732,
      "rewards/forensics_tiou_reward": 0.42826326191425323,
      "step": 461
    },
    {
      "completion_length": 21.34375,
      "epoch": 1.9330543933054394,
      "grad_norm": 4.897796630859375,
      "kl": 0.27099609375,
      "learning_rate": 5.177824267782427e-07,
      "loss": 0.1026,
      "reward": 0.7114116251468658,
      "reward_std": 0.0927608348429203,
      "reward_std_all": 0.09288395941257477,
      "rewards-max/forensics_sinkhorn_reward": 0.9685583114624023,
      "rewards-max/forensics_strict_edge_reward": 0.3157382905483246,
      "rewards-max/forensics_tiou_reward": 0.9407476186752319,
      "rewards/forensics_sinkhorn_reward": 0.6661068499088287,
      "rewards/forensics_strict_edge_reward": 0.01380711980164051,
      "rewards/forensics_tiou_reward": 0.4183274656534195,
      "step": 462
    },
    {
      "completion_length": 20.5625,
      "epoch": 1.9372384937238494,
      "grad_norm": 4.278543949127197,
      "kl": 0.326171875,
      "learning_rate": 5.167364016736402e-07,
      "loss": 0.1176,
      "reward": 0.7351942956447601,
      "reward_std": 0.056880329735577106,
      "reward_std_all": 0.06527858413755894,
      "rewards-max/forensics_sinkhorn_reward": 0.9737828969955444,
      "rewards-max/forensics_strict_edge_reward": 0.4208007901906967,
      "rewards-max/forensics_tiou_reward": 0.9561460912227631,
      "rewards/forensics_sinkhorn_reward": 0.6778168380260468,
      "rewards/forensics_strict_edge_reward": 0.028847452253103256,
      "rewards/forensics_tiou_reward": 0.42833133041858673,
      "step": 463
    },
    {
      "completion_length": 20.625,
      "epoch": 1.9414225941422594,
      "grad_norm": 5.201605796813965,
      "kl": 0.31005859375,
      "learning_rate": 5.156903765690377e-07,
      "loss": 0.1205,
      "reward": 0.7550710439682007,
      "reward_std": 0.0648167822510004,
      "reward_std_all": 0.061052076518535614,
      "rewards-max/forensics_sinkhorn_reward": 0.9773297607898712,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9722064137458801,
      "rewards/forensics_sinkhorn_reward": 0.7167166769504547,
      "rewards/forensics_strict_edge_reward": 0.05029717832803726,
      "rewards/forensics_tiou_reward": 0.49134373664855957,
      "step": 464
    },
    {
      "completion_length": 21.109375,
      "epoch": 1.9456066945606696,
      "grad_norm": 4.700014114379883,
      "kl": 0.26806640625,
      "learning_rate": 5.146443514644351e-07,
      "loss": 0.0983,
      "reward": 0.7089016735553741,
      "reward_std": 0.048081956803798676,
      "reward_std_all": 0.07251258194446564,
      "rewards-max/forensics_sinkhorn_reward": 0.9569127559661865,
      "rewards-max/forensics_strict_edge_reward": 0.09971306100487709,
      "rewards-max/forensics_tiou_reward": 0.92702916264534,
      "rewards/forensics_sinkhorn_reward": 0.6497891545295715,
      "rewards/forensics_strict_edge_reward": 0.004423317965120077,
      "rewards/forensics_tiou_reward": 0.36678798496723175,
      "step": 465
    },
    {
      "completion_length": 20.6875,
      "epoch": 1.9497907949790796,
      "grad_norm": 5.301154136657715,
      "kl": 0.494140625,
      "learning_rate": 5.135983263598326e-07,
      "loss": 0.1126,
      "reward": 0.6788020133972168,
      "reward_std": 0.10796771198511124,
      "reward_std_all": 0.11977415531873703,
      "rewards-max/forensics_sinkhorn_reward": 0.9706214368343353,
      "rewards-max/forensics_strict_edge_reward": 0.32841892540454865,
      "rewards-max/forensics_tiou_reward": 0.9477759897708893,
      "rewards/forensics_sinkhorn_reward": 0.6074216961860657,
      "rewards/forensics_strict_edge_reward": 0.023129846900701523,
      "rewards/forensics_tiou_reward": 0.33224722743034363,
      "step": 466
    },
    {
      "completion_length": 20.765625,
      "epoch": 1.9539748953974896,
      "grad_norm": 6.117879867553711,
      "kl": 0.3154296875,
      "learning_rate": 5.1255230125523e-07,
      "loss": 0.1793,
      "reward": 0.7399483919143677,
      "reward_std": 0.07829587534070015,
      "reward_std_all": 0.07413376867771149,
      "rewards-max/forensics_sinkhorn_reward": 0.9793424606323242,
      "rewards-max/forensics_strict_edge_reward": 0.3362295627593994,
      "rewards-max/forensics_tiou_reward": 0.9718473553657532,
      "rewards/forensics_sinkhorn_reward": 0.7080129683017731,
      "rewards/forensics_strict_edge_reward": 0.0171940540894866,
      "rewards/forensics_tiou_reward": 0.4500987380743027,
      "step": 467
    },
    {
      "completion_length": 23.046875,
      "epoch": 1.9581589958158996,
      "grad_norm": 5.180855751037598,
      "kl": 0.2978515625,
      "learning_rate": 5.115062761506275e-07,
      "loss": -0.0709,
      "reward": 0.7540523409843445,
      "reward_std": 0.07901521772146225,
      "reward_std_all": 0.07154357805848122,
      "rewards-max/forensics_sinkhorn_reward": 0.9297895431518555,
      "rewards-max/forensics_strict_edge_reward": 0.10525796032743528,
      "rewards-max/forensics_tiou_reward": 0.8487860560417175,
      "rewards/forensics_sinkhorn_reward": 0.6481786370277405,
      "rewards/forensics_strict_edge_reward": 0.012797958013834432,
      "rewards/forensics_tiou_reward": 0.35824018716812134,
      "step": 468
    },
    {
      "completion_length": 20.265625,
      "epoch": 1.9623430962343096,
      "grad_norm": 5.160233497619629,
      "kl": 0.2861328125,
      "learning_rate": 5.104602510460251e-07,
      "loss": 0.018,
      "reward": 0.7701991498470306,
      "reward_std": 0.04460931569337845,
      "reward_std_all": 0.04320855438709259,
      "rewards-max/forensics_sinkhorn_reward": 0.9846104383468628,
      "rewards-max/forensics_strict_edge_reward": 0.5984183251857758,
      "rewards-max/forensics_tiou_reward": 0.9887230694293976,
      "rewards/forensics_sinkhorn_reward": 0.7299156486988068,
      "rewards/forensics_strict_edge_reward": 0.02984857652336359,
      "rewards/forensics_tiou_reward": 0.46476370096206665,
      "step": 469
    },
    {
      "completion_length": 21.65625,
      "epoch": 1.9665271966527196,
      "grad_norm": 4.608546733856201,
      "kl": 0.3896484375,
      "learning_rate": 5.094142259414226e-07,
      "loss": -0.0026,
      "reward": 0.7728832364082336,
      "reward_std": 0.04464715742506087,
      "reward_std_all": 0.04231690801680088,
      "rewards-max/forensics_sinkhorn_reward": 0.9753392934799194,
      "rewards-max/forensics_strict_edge_reward": 0.2201457917690277,
      "rewards-max/forensics_tiou_reward": 0.9641893804073334,
      "rewards/forensics_sinkhorn_reward": 0.7103329598903656,
      "rewards/forensics_strict_edge_reward": 0.02369566261768341,
      "rewards/forensics_tiou_reward": 0.5193036198616028,
      "step": 470
    },
    {
      "completion_length": 19.65625,
      "epoch": 1.9707112970711296,
      "grad_norm": 5.5759992599487305,
      "kl": 0.380859375,
      "learning_rate": 5.083682008368201e-07,
      "loss": -0.0518,
      "reward": 0.7327156960964203,
      "reward_std": 0.02199563942849636,
      "reward_std_all": 0.047745914198458195,
      "rewards-max/forensics_sinkhorn_reward": 0.9492157399654388,
      "rewards-max/forensics_strict_edge_reward": 0.3187941014766693,
      "rewards-max/forensics_tiou_reward": 0.9060819149017334,
      "rewards/forensics_sinkhorn_reward": 0.6839282214641571,
      "rewards/forensics_strict_edge_reward": 0.014055822975933552,
      "rewards/forensics_tiou_reward": 0.43538698554039,
      "step": 471
    },
    {
      "completion_length": 19.921875,
      "epoch": 1.9748953974895398,
      "grad_norm": 5.35008430480957,
      "kl": 0.47998046875,
      "learning_rate": 5.073221757322176e-07,
      "loss": -0.0734,
      "reward": 0.764178603887558,
      "reward_std": 0.02170026651583612,
      "reward_std_all": 0.03334823600016534,
      "rewards-max/forensics_sinkhorn_reward": 0.9326137006282806,
      "rewards-max/forensics_strict_edge_reward": 0.36780498921871185,
      "rewards-max/forensics_tiou_reward": 0.8523896336555481,
      "rewards/forensics_sinkhorn_reward": 0.7129857242107391,
      "rewards/forensics_strict_edge_reward": 0.02373436721973121,
      "rewards/forensics_tiou_reward": 0.4616251587867737,
      "step": 472
    },
    {
      "completion_length": 20.5,
      "epoch": 1.9790794979079498,
      "grad_norm": 4.559858798980713,
      "kl": 0.337890625,
      "learning_rate": 5.062761506276151e-07,
      "loss": 0.0876,
      "reward": 0.7353086471557617,
      "reward_std": 0.05538800731301308,
      "reward_std_all": 0.06389500573277473,
      "rewards-max/forensics_sinkhorn_reward": 0.9215219616889954,
      "rewards-max/forensics_strict_edge_reward": 0.39852967113256454,
      "rewards-max/forensics_tiou_reward": 0.8398425281047821,
      "rewards/forensics_sinkhorn_reward": 0.6564807593822479,
      "rewards/forensics_strict_edge_reward": 0.01892636390402913,
      "rewards/forensics_tiou_reward": 0.37343829870224,
      "step": 473
    },
    {
      "completion_length": 20.953125,
      "epoch": 1.9832635983263598,
      "grad_norm": 5.765857219696045,
      "kl": 0.392578125,
      "learning_rate": 5.052301255230126e-07,
      "loss": 0.0298,
      "reward": 0.755614697933197,
      "reward_std": 0.06783555448055267,
      "reward_std_all": 0.06298803165555,
      "rewards-max/forensics_sinkhorn_reward": 0.9805985391139984,
      "rewards-max/forensics_strict_edge_reward": 0.31910227984189987,
      "rewards-max/forensics_tiou_reward": 0.9718476831912994,
      "rewards/forensics_sinkhorn_reward": 0.7056438624858856,
      "rewards/forensics_strict_edge_reward": 0.028376799076795578,
      "rewards/forensics_tiou_reward": 0.44226041436195374,
      "step": 474
    },
    {
      "completion_length": 20.3125,
      "epoch": 1.98744769874477,
      "grad_norm": 5.380107879638672,
      "kl": 0.40625,
      "learning_rate": 5.041841004184099e-07,
      "loss": 0.1163,
      "reward": 0.7708988189697266,
      "reward_std": 0.03472813218832016,
      "reward_std_all": 0.03774428740143776,
      "rewards-max/forensics_sinkhorn_reward": 0.916386216878891,
      "rewards-max/forensics_strict_edge_reward": 0.35830988164889277,
      "rewards-max/forensics_tiou_reward": 0.8070248663425446,
      "rewards/forensics_sinkhorn_reward": 0.7065650820732117,
      "rewards/forensics_strict_edge_reward": 0.044994926990511885,
      "rewards/forensics_tiou_reward": 0.4163246154785156,
      "step": 475
    },
    {
      "completion_length": 19.390625,
      "epoch": 1.99163179916318,
      "grad_norm": 6.292788028717041,
      "kl": 0.3583984375,
      "learning_rate": 5.031380753138074e-07,
      "loss": -0.0227,
      "reward": 0.7508713006973267,
      "reward_std": 0.05482403747737408,
      "reward_std_all": 0.05593627505004406,
      "rewards-max/forensics_sinkhorn_reward": 0.9798420965671539,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.9768048822879791,
      "rewards/forensics_sinkhorn_reward": 0.7232242524623871,
      "rewards/forensics_strict_edge_reward": 0.055814968422055244,
      "rewards/forensics_tiou_reward": 0.45892293751239777,
      "step": 476
    },
    {
      "completion_length": 21.296875,
      "epoch": 1.99581589958159,
      "grad_norm": 6.31141471862793,
      "kl": 0.3779296875,
      "learning_rate": 5.02092050209205e-07,
      "loss": 0.1132,
      "reward": 0.7221712470054626,
      "reward_std": 0.055260345339775085,
      "reward_std_all": 0.06993898749351501,
      "rewards-max/forensics_sinkhorn_reward": 0.9748754501342773,
      "rewards-max/forensics_strict_edge_reward": 0.18409663897182327,
      "rewards-max/forensics_tiou_reward": 0.9588279128074646,
      "rewards/forensics_sinkhorn_reward": 0.6376425623893738,
      "rewards/forensics_strict_edge_reward": 0.010631086832290748,
      "rewards/forensics_tiou_reward": 0.3894307315349579,
      "step": 477
    },
    {
      "completion_length": 19.642858505249023,
      "epoch": 2.0,
      "grad_norm": 5.381106376647949,
      "kl": 0.3544921875,
      "learning_rate": 5.010460251046025e-07,
      "loss": -0.0214,
      "reward": 0.7752267122268677,
      "reward_std": 0.05180600844323635,
      "reward_std_all": 0.045692505314946175,
      "rewards-max/forensics_sinkhorn_reward": 0.9149640202522278,
      "rewards-max/forensics_strict_edge_reward": 0.030864019878208637,
      "rewards-max/forensics_tiou_reward": 0.8100830912590027,
      "rewards/forensics_sinkhorn_reward": 0.7132641077041626,
      "rewards/forensics_strict_edge_reward": 0.004567211610265076,
      "rewards/forensics_tiou_reward": 0.4557513892650604,
      "step": 478
    },
    {
      "completion_length": 21.65625,
      "epoch": 2.00418410041841,
      "grad_norm": 7.179206848144531,
      "kl": 0.31640625,
      "learning_rate": 5e-07,
      "loss": 0.1092,
      "reward": 0.6791412532329559,
      "reward_std": 0.06715320982038975,
      "reward_std_all": 0.09458186104893684,
      "rewards-max/forensics_sinkhorn_reward": 0.9703925549983978,
      "rewards-max/forensics_strict_edge_reward": 0.568440854549408,
      "rewards-max/forensics_tiou_reward": 0.9552623331546783,
      "rewards/forensics_sinkhorn_reward": 0.6291174292564392,
      "rewards/forensics_strict_edge_reward": 0.02983782347291708,
      "rewards/forensics_tiou_reward": 0.4098418354988098,
      "step": 479
    },
    {
      "completion_length": 20.234375,
      "epoch": 2.00836820083682,
      "grad_norm": 8.237691879272461,
      "kl": 0.3271484375,
      "learning_rate": 4.989539748953975e-07,
      "loss": 0.0698,
      "reward": 0.7304572463035583,
      "reward_std": 0.06982330046594143,
      "reward_std_all": 0.07107491604983807,
      "rewards-max/forensics_sinkhorn_reward": 0.949927031993866,
      "rewards-max/forensics_strict_edge_reward": 0.1611594371497631,
      "rewards-max/forensics_tiou_reward": 0.900643527507782,
      "rewards/forensics_sinkhorn_reward": 0.685434103012085,
      "rewards/forensics_strict_edge_reward": 0.010701427236199379,
      "rewards/forensics_tiou_reward": 0.3916795551776886,
      "step": 480
    },
    {
      "completion_length": 21.34375,
      "epoch": 2.01255230125523,
      "grad_norm": 6.063350677490234,
      "kl": 0.326171875,
      "learning_rate": 4.97907949790795e-07,
      "loss": -0.044,
      "reward": 0.7688780128955841,
      "reward_std": 0.043247777968645096,
      "reward_std_all": 0.04247278533875942,
      "rewards-max/forensics_sinkhorn_reward": 0.9644419848918915,
      "rewards-max/forensics_strict_edge_reward": 0.41551393270492554,
      "rewards-max/forensics_tiou_reward": 0.9352799355983734,
      "rewards/forensics_sinkhorn_reward": 0.7127331793308258,
      "rewards/forensics_strict_edge_reward": 0.042753725312650204,
      "rewards/forensics_tiou_reward": 0.4995281398296356,
      "step": 481
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.01673640167364,
      "grad_norm": 4.746125221252441,
      "kl": 0.24609375,
      "learning_rate": 4.968619246861924e-07,
      "loss": 0.0072,
      "reward": 0.7922167479991913,
      "reward_std": 0.010150309186428785,
      "reward_std_all": 0.014240376651287079,
      "rewards-max/forensics_sinkhorn_reward": 0.9670760631561279,
      "rewards-max/forensics_strict_edge_reward": 0.40273305773735046,
      "rewards-max/forensics_tiou_reward": 0.9470076560974121,
      "rewards/forensics_sinkhorn_reward": 0.7198789715766907,
      "rewards/forensics_strict_edge_reward": 0.04276757687330246,
      "rewards/forensics_tiou_reward": 0.4986726939678192,
      "step": 482
    },
    {
      "completion_length": 21.171875,
      "epoch": 2.02092050209205,
      "grad_norm": 9.541452407836914,
      "kl": 0.708984375,
      "learning_rate": 4.958158995815899e-07,
      "loss": 0.115,
      "reward": 0.7691188454627991,
      "reward_std": 0.031168019399046898,
      "reward_std_all": 0.03773779980838299,
      "rewards-max/forensics_sinkhorn_reward": 0.9734992682933807,
      "rewards-max/forensics_strict_edge_reward": 0.3503752648830414,
      "rewards-max/forensics_tiou_reward": 0.9615620374679565,
      "rewards/forensics_sinkhorn_reward": 0.7246451377868652,
      "rewards/forensics_strict_edge_reward": 0.03354176506400108,
      "rewards/forensics_tiou_reward": 0.4730944186449051,
      "step": 483
    },
    {
      "completion_length": 23.359375,
      "epoch": 2.0251046025104604,
      "grad_norm": 4.4577717781066895,
      "kl": 0.4052734375,
      "learning_rate": 4.947698744769874e-07,
      "loss": 0.146,
      "reward": 0.7299139201641083,
      "reward_std": 0.06066523492336273,
      "reward_std_all": 0.06654616817831993,
      "rewards-max/forensics_sinkhorn_reward": 0.9614147841930389,
      "rewards-max/forensics_strict_edge_reward": 0.24823588877916336,
      "rewards-max/forensics_tiou_reward": 0.9257369935512543,
      "rewards/forensics_sinkhorn_reward": 0.6051261126995087,
      "rewards/forensics_strict_edge_reward": 0.015394895803183317,
      "rewards/forensics_tiou_reward": 0.4244857579469681,
      "step": 484
    },
    {
      "completion_length": 22.28125,
      "epoch": 2.0292887029288704,
      "grad_norm": 5.56451940536499,
      "kl": 0.30078125,
      "learning_rate": 4.93723849372385e-07,
      "loss": 0.2132,
      "reward": 0.7559432089328766,
      "reward_std": 0.05655086413025856,
      "reward_std_all": 0.05486222356557846,
      "rewards-max/forensics_sinkhorn_reward": 0.9811746776103973,
      "rewards-max/forensics_strict_edge_reward": 0.24298179522156715,
      "rewards-max/forensics_tiou_reward": 0.9759746491909027,
      "rewards/forensics_sinkhorn_reward": 0.6591641306877136,
      "rewards/forensics_strict_edge_reward": 0.018283448880538344,
      "rewards/forensics_tiou_reward": 0.49036814272403717,
      "step": 485
    },
    {
      "completion_length": 20.265625,
      "epoch": 2.0334728033472804,
      "grad_norm": 5.254391670227051,
      "kl": 0.296875,
      "learning_rate": 4.926778242677825e-07,
      "loss": 0.1609,
      "reward": 0.7193782925605774,
      "reward_std": 0.07338497415184975,
      "reward_std_all": 0.08008981123566628,
      "rewards-max/forensics_sinkhorn_reward": 0.974193662405014,
      "rewards-max/forensics_strict_edge_reward": 0.4240923821926117,
      "rewards-max/forensics_tiou_reward": 0.9622718691825867,
      "rewards/forensics_sinkhorn_reward": 0.6932234168052673,
      "rewards/forensics_strict_edge_reward": 0.04882177338004112,
      "rewards/forensics_tiou_reward": 0.4537588655948639,
      "step": 486
    },
    {
      "completion_length": 20.734375,
      "epoch": 2.0376569037656904,
      "grad_norm": 5.181321620941162,
      "kl": 0.435546875,
      "learning_rate": 4.916317991631799e-07,
      "loss": 0.0649,
      "reward": 0.7424506545066833,
      "reward_std": 0.06095500290393829,
      "reward_std_all": 0.0652596727013588,
      "rewards-max/forensics_sinkhorn_reward": 0.9687223136425018,
      "rewards-max/forensics_strict_edge_reward": 0.26387959718704224,
      "rewards-max/forensics_tiou_reward": 0.947577714920044,
      "rewards/forensics_sinkhorn_reward": 0.6808677613735199,
      "rewards/forensics_strict_edge_reward": 0.015691895969212055,
      "rewards/forensics_tiou_reward": 0.44064393639564514,
      "step": 487
    },
    {
      "completion_length": 21.09375,
      "epoch": 2.0418410041841004,
      "grad_norm": 5.151922702789307,
      "kl": 0.29833984375,
      "learning_rate": 4.905857740585774e-07,
      "loss": 0.0759,
      "reward": 0.7840109467506409,
      "reward_std": 0.026525244116783142,
      "reward_std_all": 0.027207467705011368,
      "rewards-max/forensics_sinkhorn_reward": 0.9675212800502777,
      "rewards-max/forensics_strict_edge_reward": 0.47400766611099243,
      "rewards-max/forensics_tiou_reward": 0.9408344328403473,
      "rewards/forensics_sinkhorn_reward": 0.7659254372119904,
      "rewards/forensics_strict_edge_reward": 0.04588644579052925,
      "rewards/forensics_tiou_reward": 0.5954129993915558,
      "step": 488
    },
    {
      "completion_length": 21.3125,
      "epoch": 2.0460251046025104,
      "grad_norm": 4.795408725738525,
      "kl": 0.298828125,
      "learning_rate": 4.895397489539749e-07,
      "loss": 0.1943,
      "reward": 0.7347447276115417,
      "reward_std": 0.05636784993112087,
      "reward_std_all": 0.06472824141383171,
      "rewards-max/forensics_sinkhorn_reward": 0.9831012487411499,
      "rewards-max/forensics_strict_edge_reward": 0.4496074318885803,
      "rewards-max/forensics_tiou_reward": 0.985521525144577,
      "rewards/forensics_sinkhorn_reward": 0.6698004603385925,
      "rewards/forensics_strict_edge_reward": 0.05357362702488899,
      "rewards/forensics_tiou_reward": 0.49056534469127655,
      "step": 489
    },
    {
      "completion_length": 20.46875,
      "epoch": 2.0502092050209204,
      "grad_norm": 6.392441272735596,
      "kl": 0.3994140625,
      "learning_rate": 4.884937238493724e-07,
      "loss": -0.0335,
      "reward": 0.7506491541862488,
      "reward_std": 0.05273710563778877,
      "reward_std_all": 0.05345885083079338,
      "rewards-max/forensics_sinkhorn_reward": 0.9693812429904938,
      "rewards-max/forensics_strict_edge_reward": 0.5599738955497742,
      "rewards-max/forensics_tiou_reward": 0.9530565440654755,
      "rewards/forensics_sinkhorn_reward": 0.6919406056404114,
      "rewards/forensics_strict_edge_reward": 0.03883352503180504,
      "rewards/forensics_tiou_reward": 0.4276962876319885,
      "step": 490
    },
    {
      "completion_length": 22.0625,
      "epoch": 2.0543933054393304,
      "grad_norm": 6.034440517425537,
      "kl": 0.3955078125,
      "learning_rate": 4.874476987447699e-07,
      "loss": 0.0529,
      "reward": 0.693629115819931,
      "reward_std": 0.07917571812868118,
      "reward_std_all": 0.09422168880701065,
      "rewards-max/forensics_sinkhorn_reward": 0.9747587442398071,
      "rewards-max/forensics_strict_edge_reward": 0.5995520651340485,
      "rewards-max/forensics_tiou_reward": 0.9624325335025787,
      "rewards/forensics_sinkhorn_reward": 0.6135664582252502,
      "rewards/forensics_strict_edge_reward": 0.03723141085356474,
      "rewards/forensics_tiou_reward": 0.36124998331069946,
      "step": 491
    },
    {
      "completion_length": 21.015625,
      "epoch": 2.0585774058577404,
      "grad_norm": 6.012566089630127,
      "kl": 0.30859375,
      "learning_rate": 4.864016736401673e-07,
      "loss": 0.0461,
      "reward": 0.7685425877571106,
      "reward_std": 0.06317717483034357,
      "reward_std_all": 0.057056419318541884,
      "rewards-max/forensics_sinkhorn_reward": 0.9783093631267548,
      "rewards-max/forensics_strict_edge_reward": 0.4141879454255104,
      "rewards-max/forensics_tiou_reward": 0.9720630347728729,
      "rewards/forensics_sinkhorn_reward": 0.7539833784103394,
      "rewards/forensics_strict_edge_reward": 0.03245204035192728,
      "rewards/forensics_tiou_reward": 0.5577722787857056,
      "step": 492
    },
    {
      "completion_length": 20.796875,
      "epoch": 2.062761506276151,
      "grad_norm": 5.567983627319336,
      "kl": 0.294921875,
      "learning_rate": 4.853556485355648e-07,
      "loss": 0.1064,
      "reward": 0.7670404314994812,
      "reward_std": 0.04645001236349344,
      "reward_std_all": 0.044814265333116055,
      "rewards-max/forensics_sinkhorn_reward": 0.9631796479225159,
      "rewards-max/forensics_strict_edge_reward": 0.36461610347032547,
      "rewards-max/forensics_tiou_reward": 0.9394750893115997,
      "rewards/forensics_sinkhorn_reward": 0.7374753355979919,
      "rewards/forensics_strict_edge_reward": 0.052543070167303085,
      "rewards/forensics_tiou_reward": 0.5284758508205414,
      "step": 493
    },
    {
      "completion_length": 20.140625,
      "epoch": 2.066945606694561,
      "grad_norm": 5.370961666107178,
      "kl": 0.3271484375,
      "learning_rate": 4.843096234309624e-07,
      "loss": -0.0164,
      "reward": 0.7717893123626709,
      "reward_std": 0.046755289658904076,
      "reward_std_all": 0.04450840596109629,
      "rewards-max/forensics_sinkhorn_reward": 0.9567363560199738,
      "rewards-max/forensics_strict_edge_reward": 0.43068649992346764,
      "rewards-max/forensics_tiou_reward": 0.9242690801620483,
      "rewards/forensics_sinkhorn_reward": 0.7339513599872589,
      "rewards/forensics_strict_edge_reward": 0.026695251232013106,
      "rewards/forensics_tiou_reward": 0.4855440855026245,
      "step": 494
    },
    {
      "completion_length": 21.515625,
      "epoch": 2.071129707112971,
      "grad_norm": 6.315979480743408,
      "kl": 0.30078125,
      "learning_rate": 4.832635983263599e-07,
      "loss": 0.0057,
      "reward": 0.7723011672496796,
      "reward_std": 0.029957685619592667,
      "reward_std_all": 0.0318108294159174,
      "rewards-max/forensics_sinkhorn_reward": 0.975146621465683,
      "rewards-max/forensics_strict_edge_reward": 0.5465418845415115,
      "rewards-max/forensics_tiou_reward": 0.9585334658622742,
      "rewards/forensics_sinkhorn_reward": 0.7052096128463745,
      "rewards/forensics_strict_edge_reward": 0.041596151888370514,
      "rewards/forensics_tiou_reward": 0.48271504044532776,
      "step": 495
    },
    {
      "completion_length": 21.796875,
      "epoch": 2.075313807531381,
      "grad_norm": 6.974424362182617,
      "kl": 0.33984375,
      "learning_rate": 4.822175732217573e-07,
      "loss": 0.1161,
      "reward": 0.7030099332332611,
      "reward_std": 0.052436452358961105,
      "reward_std_all": 0.07814646139740944,
      "rewards-max/forensics_sinkhorn_reward": 0.9437254667282104,
      "rewards-max/forensics_strict_edge_reward": 0.09607945382595062,
      "rewards-max/forensics_tiou_reward": 0.8799692392349243,
      "rewards/forensics_sinkhorn_reward": 0.6573789715766907,
      "rewards/forensics_strict_edge_reward": 0.008492698660120368,
      "rewards/forensics_tiou_reward": 0.4666029065847397,
      "step": 496
    },
    {
      "completion_length": 21.671875,
      "epoch": 2.079497907949791,
      "grad_norm": 5.636169910430908,
      "kl": 0.3583984375,
      "learning_rate": 4.811715481171548e-07,
      "loss": 0.0735,
      "reward": 0.7190687656402588,
      "reward_std": 0.059074483811855316,
      "reward_std_all": 0.07509832829236984,
      "rewards-max/forensics_sinkhorn_reward": 0.9769355952739716,
      "rewards-max/forensics_strict_edge_reward": 0.2832247018814087,
      "rewards-max/forensics_tiou_reward": 0.9680771827697754,
      "rewards/forensics_sinkhorn_reward": 0.6599850654602051,
      "rewards/forensics_strict_edge_reward": 0.01955643855035305,
      "rewards/forensics_tiou_reward": 0.46368347108364105,
      "step": 497
    },
    {
      "completion_length": 21.953125,
      "epoch": 2.083682008368201,
      "grad_norm": 5.891427993774414,
      "kl": 0.2646484375,
      "learning_rate": 4.801255230125523e-07,
      "loss": 0.0491,
      "reward": 0.7748954594135284,
      "reward_std": 0.040049998089671135,
      "reward_std_all": 0.03834572061896324,
      "rewards-max/forensics_sinkhorn_reward": 0.9786108434200287,
      "rewards-max/forensics_strict_edge_reward": 0.4533289223909378,
      "rewards-max/forensics_tiou_reward": 0.9677492678165436,
      "rewards/forensics_sinkhorn_reward": 0.7532595992088318,
      "rewards/forensics_strict_edge_reward": 0.0380100030452013,
      "rewards/forensics_tiou_reward": 0.5612748563289642,
      "step": 498
    },
    {
      "completion_length": 20.375,
      "epoch": 2.087866108786611,
      "grad_norm": 5.721343517303467,
      "kl": 0.330078125,
      "learning_rate": 4.790794979079498e-07,
      "loss": -0.0298,
      "reward": 0.7587188482284546,
      "reward_std": 0.038490625098347664,
      "reward_std_all": 0.04564761742949486,
      "rewards-max/forensics_sinkhorn_reward": 0.9769717156887054,
      "rewards-max/forensics_strict_edge_reward": 0.3657475709915161,
      "rewards-max/forensics_tiou_reward": 0.9666001796722412,
      "rewards/forensics_sinkhorn_reward": 0.751720130443573,
      "rewards/forensics_strict_edge_reward": 0.04510718863457441,
      "rewards/forensics_tiou_reward": 0.5495745092630386,
      "step": 499
    },
    {
      "completion_length": 23.453125,
      "epoch": 2.092050209205021,
      "grad_norm": 4.9395751953125,
      "kl": 0.310546875,
      "learning_rate": 4.780334728033473e-07,
      "loss": 0.0441,
      "reward": 0.7431129813194275,
      "reward_std": 0.017926277592778206,
      "reward_std_all": 0.04035908915102482,
      "rewards-max/forensics_sinkhorn_reward": 0.967346727848053,
      "rewards-max/forensics_strict_edge_reward": 0.18399372764542932,
      "rewards-max/forensics_tiou_reward": 0.943963497877121,
      "rewards/forensics_sinkhorn_reward": 0.7027260661125183,
      "rewards/forensics_strict_edge_reward": 0.019678043094245368,
      "rewards/forensics_tiou_reward": 0.48820897936820984,
      "step": 500
    },
    {
      "completion_length": 20.390625,
      "epoch": 2.096234309623431,
      "grad_norm": 6.028220176696777,
      "kl": 0.3310546875,
      "learning_rate": 4.769874476987447e-07,
      "loss": 0.0565,
      "reward": 0.7291177213191986,
      "reward_std": 0.04563793167471886,
      "reward_std_all": 0.05959399603307247,
      "rewards-max/forensics_sinkhorn_reward": 0.9763977229595184,
      "rewards-max/forensics_strict_edge_reward": 0.48575712740421295,
      "rewards-max/forensics_tiou_reward": 0.9641073644161224,
      "rewards/forensics_sinkhorn_reward": 0.701496958732605,
      "rewards/forensics_strict_edge_reward": 0.03357022535055876,
      "rewards/forensics_tiou_reward": 0.47118863463401794,
      "step": 501
    },
    {
      "completion_length": 20.203125,
      "epoch": 2.100418410041841,
      "grad_norm": 6.611440658569336,
      "kl": 0.3876953125,
      "learning_rate": 4.759414225941422e-07,
      "loss": 0.0753,
      "reward": 0.7357408106327057,
      "reward_std": 0.06741535104811192,
      "reward_std_all": 0.06772027909755707,
      "rewards-max/forensics_sinkhorn_reward": 0.9547930359840393,
      "rewards-max/forensics_strict_edge_reward": 0.3159602954983711,
      "rewards-max/forensics_tiou_reward": 0.9199918806552887,
      "rewards/forensics_sinkhorn_reward": 0.6725686490535736,
      "rewards/forensics_strict_edge_reward": 0.024245738983154297,
      "rewards/forensics_tiou_reward": 0.4189341515302658,
      "step": 502
    },
    {
      "completion_length": 19.875,
      "epoch": 2.104602510460251,
      "grad_norm": 5.410585880279541,
      "kl": 0.2861328125,
      "learning_rate": 4.748953974895397e-07,
      "loss": 0.0181,
      "reward": 0.7589962780475616,
      "reward_std": 0.046923222951591015,
      "reward_std_all": 0.047653766348958015,
      "rewards-max/forensics_sinkhorn_reward": 0.9563888609409332,
      "rewards-max/forensics_strict_edge_reward": 0.2657850868999958,
      "rewards-max/forensics_tiou_reward": 0.9146722257137299,
      "rewards/forensics_sinkhorn_reward": 0.6843410730361938,
      "rewards/forensics_strict_edge_reward": 0.0162038195412606,
      "rewards/forensics_tiou_reward": 0.37626461684703827,
      "step": 503
    },
    {
      "completion_length": 20.53125,
      "epoch": 2.1087866108786613,
      "grad_norm": 6.35353422164917,
      "kl": 0.4140625,
      "learning_rate": 4.738493723849372e-07,
      "loss": -0.0276,
      "reward": 0.7649461030960083,
      "reward_std": 0.025668052141554654,
      "reward_std_all": 0.035482382168993354,
      "rewards-max/forensics_sinkhorn_reward": 0.9548965692520142,
      "rewards-max/forensics_strict_edge_reward": 0.28496674448251724,
      "rewards-max/forensics_tiou_reward": 0.9107859432697296,
      "rewards/forensics_sinkhorn_reward": 0.7329635620117188,
      "rewards/forensics_strict_edge_reward": 0.03505043871700764,
      "rewards/forensics_tiou_reward": 0.4997579753398895,
      "step": 504
    },
    {
      "completion_length": 22.671875,
      "epoch": 2.1129707112970713,
      "grad_norm": 6.350840091705322,
      "kl": 0.287109375,
      "learning_rate": 4.7280334728033473e-07,
      "loss": 0.1451,
      "reward": 0.7412730753421783,
      "reward_std": 0.07756553217768669,
      "reward_std_all": 0.07394465059041977,
      "rewards-max/forensics_sinkhorn_reward": 0.9761519432067871,
      "rewards-max/forensics_strict_edge_reward": 0.5042535066604614,
      "rewards-max/forensics_tiou_reward": 0.9675332307815552,
      "rewards/forensics_sinkhorn_reward": 0.6901875138282776,
      "rewards/forensics_strict_edge_reward": 0.03228754736483097,
      "rewards/forensics_tiou_reward": 0.4542723000049591,
      "step": 505
    },
    {
      "completion_length": 23.96875,
      "epoch": 2.1171548117154813,
      "grad_norm": 5.383124351501465,
      "kl": 0.494140625,
      "learning_rate": 4.7175732217573224e-07,
      "loss": 0.0162,
      "reward": 0.7253367304801941,
      "reward_std": 0.047059543430805206,
      "reward_std_all": 0.0642552450299263,
      "rewards-max/forensics_sinkhorn_reward": 0.9558390974998474,
      "rewards-max/forensics_strict_edge_reward": 0.21902879246044904,
      "rewards-max/forensics_tiou_reward": 0.9137154817581177,
      "rewards/forensics_sinkhorn_reward": 0.6398943960666656,
      "rewards/forensics_strict_edge_reward": 0.019087237378698774,
      "rewards/forensics_tiou_reward": 0.37335461378097534,
      "step": 506
    },
    {
      "completion_length": 20.53125,
      "epoch": 2.1213389121338913,
      "grad_norm": 5.107645034790039,
      "kl": 0.384765625,
      "learning_rate": 4.7071129707112965e-07,
      "loss": -0.0669,
      "reward": 0.7598741352558136,
      "reward_std": 0.03645052667707205,
      "reward_std_all": 0.04169652983546257,
      "rewards-max/forensics_sinkhorn_reward": 0.9746622145175934,
      "rewards-max/forensics_strict_edge_reward": 0.39366208761930466,
      "rewards-max/forensics_tiou_reward": 0.9654021561145782,
      "rewards/forensics_sinkhorn_reward": 0.752016007900238,
      "rewards/forensics_strict_edge_reward": 0.03623213246464729,
      "rewards/forensics_tiou_reward": 0.5260193645954132,
      "step": 507
    },
    {
      "completion_length": 25.109375,
      "epoch": 2.1255230125523012,
      "grad_norm": 4.720836639404297,
      "kl": 0.3505859375,
      "learning_rate": 4.6966527196652716e-07,
      "loss": -0.0416,
      "reward": 0.6914059817790985,
      "reward_std": 0.0408727815374732,
      "reward_std_all": 0.07626299187541008,
      "rewards-max/forensics_sinkhorn_reward": 0.9657108187675476,
      "rewards-max/forensics_strict_edge_reward": 0.327541142469272,
      "rewards-max/forensics_tiou_reward": 0.9371993243694305,
      "rewards/forensics_sinkhorn_reward": 0.6017718613147736,
      "rewards/forensics_strict_edge_reward": 0.015383446385385469,
      "rewards/forensics_tiou_reward": 0.3919884264469147,
      "step": 508
    },
    {
      "completion_length": 19.71875,
      "epoch": 2.1297071129707112,
      "grad_norm": 7.121277332305908,
      "kl": 0.3291015625,
      "learning_rate": 4.686192468619247e-07,
      "loss": -0.0038,
      "reward": 0.7274768948554993,
      "reward_std": 0.023677038960158825,
      "reward_std_all": 0.051228756085038185,
      "rewards-max/forensics_sinkhorn_reward": 0.9547612965106964,
      "rewards-max/forensics_strict_edge_reward": 0.250081405043602,
      "rewards-max/forensics_tiou_reward": 0.8970467746257782,
      "rewards/forensics_sinkhorn_reward": 0.703779548406601,
      "rewards/forensics_strict_edge_reward": 0.017543285619467497,
      "rewards/forensics_tiou_reward": 0.45918242633342743,
      "step": 509
    },
    {
      "completion_length": 20.015625,
      "epoch": 2.1338912133891212,
      "grad_norm": 6.516739845275879,
      "kl": 0.408203125,
      "learning_rate": 4.675732217573222e-07,
      "loss": 0.1331,
      "reward": 0.695791095495224,
      "reward_std": 0.09292484540492296,
      "reward_std_all": 0.10090041719377041,
      "rewards-max/forensics_sinkhorn_reward": 0.963360458612442,
      "rewards-max/forensics_strict_edge_reward": 0.45875632762908936,
      "rewards-max/forensics_tiou_reward": 0.9254361391067505,
      "rewards/forensics_sinkhorn_reward": 0.6430473029613495,
      "rewards/forensics_strict_edge_reward": 0.03106633946299553,
      "rewards/forensics_tiou_reward": 0.403695672750473,
      "step": 510
    },
    {
      "completion_length": 20.390625,
      "epoch": 2.1380753138075312,
      "grad_norm": 5.499398708343506,
      "kl": 0.42578125,
      "learning_rate": 4.6652719665271965e-07,
      "loss": -0.0526,
      "reward": 0.7777186334133148,
      "reward_std": 0.040575109887868166,
      "reward_std_all": 0.038238337030634284,
      "rewards-max/forensics_sinkhorn_reward": 0.9781762361526489,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9753544926643372,
      "rewards/forensics_sinkhorn_reward": 0.7444254755973816,
      "rewards/forensics_strict_edge_reward": 0.05173537693917751,
      "rewards/forensics_tiou_reward": 0.5379616618156433,
      "step": 511
    },
    {
      "completion_length": 21.421875,
      "epoch": 2.1422594142259412,
      "grad_norm": 5.544994831085205,
      "kl": 0.3642578125,
      "learning_rate": 4.654811715481171e-07,
      "loss": 0.0618,
      "reward": 0.7496817708015442,
      "reward_std": 0.03164525702595711,
      "reward_std_all": 0.042995525524020195,
      "rewards-max/forensics_sinkhorn_reward": 0.9602930247783661,
      "rewards-max/forensics_strict_edge_reward": 0.3017394542694092,
      "rewards-max/forensics_tiou_reward": 0.9294363856315613,
      "rewards/forensics_sinkhorn_reward": 0.7010483741760254,
      "rewards/forensics_strict_edge_reward": 0.038829753175377846,
      "rewards/forensics_tiou_reward": 0.47116564214229584,
      "step": 512
    },
    {
      "completion_length": 23.09375,
      "epoch": 2.1464435146443517,
      "grad_norm": 5.006314754486084,
      "kl": 0.2802734375,
      "learning_rate": 4.644351464435146e-07,
      "loss": -0.0228,
      "reward": 0.7489407658576965,
      "reward_std": 0.0649071903899312,
      "reward_std_all": 0.06364958733320236,
      "rewards-max/forensics_sinkhorn_reward": 0.9807257354259491,
      "rewards-max/forensics_strict_edge_reward": 0.5650643557310104,
      "rewards-max/forensics_tiou_reward": 0.9734720885753632,
      "rewards/forensics_sinkhorn_reward": 0.6852312088012695,
      "rewards/forensics_strict_edge_reward": 0.042290820740163326,
      "rewards/forensics_tiou_reward": 0.4577408730983734,
      "step": 513
    },
    {
      "completion_length": 23.625,
      "epoch": 2.1506276150627617,
      "grad_norm": 5.123887062072754,
      "kl": 0.3837890625,
      "learning_rate": 4.6338912133891213e-07,
      "loss": 0.0258,
      "reward": 0.7444247603416443,
      "reward_std": 0.051792245358228683,
      "reward_std_all": 0.05581716448068619,
      "rewards-max/forensics_sinkhorn_reward": 0.9695333242416382,
      "rewards-max/forensics_strict_edge_reward": 0.4685998484492302,
      "rewards-max/forensics_tiou_reward": 0.9438371956348419,
      "rewards/forensics_sinkhorn_reward": 0.6999676823616028,
      "rewards/forensics_strict_edge_reward": 0.044498531613498926,
      "rewards/forensics_tiou_reward": 0.489242821931839,
      "step": 514
    },
    {
      "completion_length": 18.984375,
      "epoch": 2.1548117154811717,
      "grad_norm": 5.969424724578857,
      "kl": 0.33203125,
      "learning_rate": 4.623430962343096e-07,
      "loss": -0.085,
      "reward": 0.7494279742240906,
      "reward_std": 0.03971659578382969,
      "reward_std_all": 0.047768861055374146,
      "rewards-max/forensics_sinkhorn_reward": 0.9665744304656982,
      "rewards-max/forensics_strict_edge_reward": 0.5311143696308136,
      "rewards-max/forensics_tiou_reward": 0.9492850303649902,
      "rewards/forensics_sinkhorn_reward": 0.7336131632328033,
      "rewards/forensics_strict_edge_reward": 0.03503322973847389,
      "rewards/forensics_tiou_reward": 0.49225132167339325,
      "step": 515
    },
    {
      "completion_length": 22.875,
      "epoch": 2.1589958158995817,
      "grad_norm": 5.789949893951416,
      "kl": 0.408203125,
      "learning_rate": 4.612970711297071e-07,
      "loss": 0.0495,
      "reward": 0.6764950156211853,
      "reward_std": 0.07661367021501064,
      "reward_std_all": 0.10368558019399643,
      "rewards-max/forensics_sinkhorn_reward": 0.9361260831356049,
      "rewards-max/forensics_strict_edge_reward": 0.08322770893573761,
      "rewards-max/forensics_tiou_reward": 0.8622733652591705,
      "rewards/forensics_sinkhorn_reward": 0.6027010679244995,
      "rewards/forensics_strict_edge_reward": 0.0070326137356460094,
      "rewards/forensics_tiou_reward": 0.3939143717288971,
      "step": 516
    },
    {
      "completion_length": 24.765625,
      "epoch": 2.1631799163179917,
      "grad_norm": 5.271854400634766,
      "kl": 0.328125,
      "learning_rate": 4.6025104602510457e-07,
      "loss": 0.1078,
      "reward": 0.7575666904449463,
      "reward_std": 0.02884680964052677,
      "reward_std_all": 0.04052962735295296,
      "rewards-max/forensics_sinkhorn_reward": 0.9671913385391235,
      "rewards-max/forensics_strict_edge_reward": 0.4533289223909378,
      "rewards-max/forensics_tiou_reward": 0.94721719622612,
      "rewards/forensics_sinkhorn_reward": 0.7270198464393616,
      "rewards/forensics_strict_edge_reward": 0.033772002905607224,
      "rewards/forensics_tiou_reward": 0.5284322500228882,
      "step": 517
    },
    {
      "completion_length": 23.96875,
      "epoch": 2.1673640167364017,
      "grad_norm": 5.2686567306518555,
      "kl": 0.2568359375,
      "learning_rate": 4.592050209205021e-07,
      "loss": 0.1069,
      "reward": 0.7399524450302124,
      "reward_std": 0.07234354875981808,
      "reward_std_all": 0.0715028066188097,
      "rewards-max/forensics_sinkhorn_reward": 0.9744146466255188,
      "rewards-max/forensics_strict_edge_reward": 0.5361751317977905,
      "rewards-max/forensics_tiou_reward": 0.9625757932662964,
      "rewards/forensics_sinkhorn_reward": 0.6942535638809204,
      "rewards/forensics_strict_edge_reward": 0.052355099469423294,
      "rewards/forensics_tiou_reward": 0.4719303846359253,
      "step": 518
    },
    {
      "completion_length": 21.03125,
      "epoch": 2.1715481171548117,
      "grad_norm": 5.018321990966797,
      "kl": 0.337890625,
      "learning_rate": 4.5815899581589954e-07,
      "loss": 0.1888,
      "reward": 0.7296029329299927,
      "reward_std": 0.07703709974884987,
      "reward_std_all": 0.08237702772021294,
      "rewards-max/forensics_sinkhorn_reward": 0.9885928928852081,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9990125298500061,
      "rewards/forensics_sinkhorn_reward": 0.6867437064647675,
      "rewards/forensics_strict_edge_reward": 0.08917794190347195,
      "rewards/forensics_tiou_reward": 0.49243396520614624,
      "step": 519
    },
    {
      "completion_length": 23.875,
      "epoch": 2.1757322175732217,
      "grad_norm": 5.634328842163086,
      "kl": 0.30078125,
      "learning_rate": 4.5711297071129705e-07,
      "loss": -0.0901,
      "reward": 0.752691924571991,
      "reward_std": 0.043584538623690605,
      "reward_std_all": 0.04666898027062416,
      "rewards-max/forensics_sinkhorn_reward": 0.9737770855426788,
      "rewards-max/forensics_strict_edge_reward": 0.561550498008728,
      "rewards-max/forensics_tiou_reward": 0.9626891911029816,
      "rewards/forensics_sinkhorn_reward": 0.7266184389591217,
      "rewards/forensics_strict_edge_reward": 0.07767772674560547,
      "rewards/forensics_tiou_reward": 0.5562663376331329,
      "step": 520
    },
    {
      "completion_length": 23.65625,
      "epoch": 2.1799163179916317,
      "grad_norm": 5.107607364654541,
      "kl": 0.302734375,
      "learning_rate": 4.5606694560669456e-07,
      "loss": 0.0673,
      "reward": 0.709159255027771,
      "reward_std": 0.034968405961990356,
      "reward_std_all": 0.06524854153394699,
      "rewards-max/forensics_sinkhorn_reward": 0.9809110760688782,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9795007109642029,
      "rewards/forensics_sinkhorn_reward": 0.6557026207447052,
      "rewards/forensics_strict_edge_reward": 0.04677686467766762,
      "rewards/forensics_tiou_reward": 0.4513293653726578,
      "step": 521
    },
    {
      "completion_length": 22.515625,
      "epoch": 2.1841004184100417,
      "grad_norm": 4.979752540588379,
      "kl": 0.28759765625,
      "learning_rate": 4.55020920502092e-07,
      "loss": 0.0418,
      "reward": 0.7102883458137512,
      "reward_std": 0.05498878285288811,
      "reward_std_all": 0.07683329842984676,
      "rewards-max/forensics_sinkhorn_reward": 0.9856867790222168,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9917379021644592,
      "rewards/forensics_sinkhorn_reward": 0.6609249711036682,
      "rewards/forensics_strict_edge_reward": 0.06900433450937271,
      "rewards/forensics_tiou_reward": 0.447647824883461,
      "step": 522
    },
    {
      "completion_length": 22.4375,
      "epoch": 2.1882845188284517,
      "grad_norm": 6.302180767059326,
      "kl": 0.34765625,
      "learning_rate": 4.539748953974895e-07,
      "loss": -0.0359,
      "reward": 0.7477871775627136,
      "reward_std": 0.043520040810108185,
      "reward_std_all": 0.05385456793010235,
      "rewards-max/forensics_sinkhorn_reward": 0.9761154353618622,
      "rewards-max/forensics_strict_edge_reward": 0.7361255586147308,
      "rewards-max/forensics_tiou_reward": 0.9698688685894012,
      "rewards/forensics_sinkhorn_reward": 0.6712071299552917,
      "rewards/forensics_strict_edge_reward": 0.05522548034787178,
      "rewards/forensics_tiou_reward": 0.4756298214197159,
      "step": 523
    },
    {
      "completion_length": 20.78125,
      "epoch": 2.192468619246862,
      "grad_norm": 5.327425956726074,
      "kl": 0.234375,
      "learning_rate": 4.52928870292887e-07,
      "loss": 0.0781,
      "reward": 0.7569288909435272,
      "reward_std": 0.05205366015434265,
      "reward_std_all": 0.051527537405490875,
      "rewards-max/forensics_sinkhorn_reward": 0.9712466597557068,
      "rewards-max/forensics_strict_edge_reward": 0.6284831166267395,
      "rewards-max/forensics_tiou_reward": 0.9583317041397095,
      "rewards/forensics_sinkhorn_reward": 0.6841581463813782,
      "rewards/forensics_strict_edge_reward": 0.03620360232889652,
      "rewards/forensics_tiou_reward": 0.4580410420894623,
      "step": 524
    },
    {
      "completion_length": 22.015625,
      "epoch": 2.196652719665272,
      "grad_norm": 5.957977294921875,
      "kl": 0.400390625,
      "learning_rate": 4.518828451882845e-07,
      "loss": 0.0775,
      "reward": 0.7177655398845673,
      "reward_std": 0.10769626125693321,
      "reward_std_all": 0.10377368703484535,
      "rewards-max/forensics_sinkhorn_reward": 0.9631764590740204,
      "rewards-max/forensics_strict_edge_reward": 0.5256901979446411,
      "rewards-max/forensics_tiou_reward": 0.9376353621482849,
      "rewards/forensics_sinkhorn_reward": 0.6470312476158142,
      "rewards/forensics_strict_edge_reward": 0.05482916347682476,
      "rewards/forensics_tiou_reward": 0.44290101528167725,
      "step": 525
    },
    {
      "completion_length": 20.015625,
      "epoch": 2.200836820083682,
      "grad_norm": 5.9238433837890625,
      "kl": 0.3115234375,
      "learning_rate": 4.50836820083682e-07,
      "loss": 0.0619,
      "reward": 0.7841934263706207,
      "reward_std": 0.026532036252319813,
      "reward_std_all": 0.026836233213543892,
      "rewards-max/forensics_sinkhorn_reward": 0.9722516238689423,
      "rewards-max/forensics_strict_edge_reward": 0.28324673511087894,
      "rewards-max/forensics_tiou_reward": 0.9362694323062897,
      "rewards/forensics_sinkhorn_reward": 0.7338747680187225,
      "rewards/forensics_strict_edge_reward": 0.032169971615076065,
      "rewards/forensics_tiou_reward": 0.48594218492507935,
      "step": 526
    },
    {
      "completion_length": 23.921875,
      "epoch": 2.205020920502092,
      "grad_norm": 5.280872344970703,
      "kl": 0.26904296875,
      "learning_rate": 4.497907949790795e-07,
      "loss": 0.0293,
      "reward": 0.7104398310184479,
      "reward_std": 0.0577419176697731,
      "reward_std_all": 0.07599225267767906,
      "rewards-max/forensics_sinkhorn_reward": 0.984889030456543,
      "rewards-max/forensics_strict_edge_reward": 0.6505971103906631,
      "rewards-max/forensics_tiou_reward": 0.988495796918869,
      "rewards/forensics_sinkhorn_reward": 0.6608200669288635,
      "rewards/forensics_strict_edge_reward": 0.06605817191302776,
      "rewards/forensics_tiou_reward": 0.4639871418476105,
      "step": 527
    },
    {
      "completion_length": 22.65625,
      "epoch": 2.209205020920502,
      "grad_norm": 7.808990001678467,
      "kl": 0.34375,
      "learning_rate": 4.4874476987447694e-07,
      "loss": -0.0037,
      "reward": 0.7530241310596466,
      "reward_std": 0.05481417942792177,
      "reward_std_all": 0.05523594608530402,
      "rewards-max/forensics_sinkhorn_reward": 0.9787160754203796,
      "rewards-max/forensics_strict_edge_reward": 0.2977510616183281,
      "rewards-max/forensics_tiou_reward": 0.9747969210147858,
      "rewards/forensics_sinkhorn_reward": 0.7118105590343475,
      "rewards/forensics_strict_edge_reward": 0.04210788058117032,
      "rewards/forensics_tiou_reward": 0.5084881782531738,
      "step": 528
    },
    {
      "completion_length": 22.109375,
      "epoch": 2.213389121338912,
      "grad_norm": 7.215502738952637,
      "kl": 0.357421875,
      "learning_rate": 4.4769874476987445e-07,
      "loss": 0.0691,
      "reward": 0.7467401325702667,
      "reward_std": 0.05084287375211716,
      "reward_std_all": 0.0555605236440897,
      "rewards-max/forensics_sinkhorn_reward": 0.9410011768341064,
      "rewards-max/forensics_strict_edge_reward": 0.05315634282305837,
      "rewards-max/forensics_tiou_reward": 0.8703267574310303,
      "rewards/forensics_sinkhorn_reward": 0.7050576508045197,
      "rewards/forensics_strict_edge_reward": 0.007199823798146099,
      "rewards/forensics_tiou_reward": 0.4864360988140106,
      "step": 529
    },
    {
      "completion_length": 21.734375,
      "epoch": 2.217573221757322,
      "grad_norm": 4.7765960693359375,
      "kl": 0.3447265625,
      "learning_rate": 4.4665271966527197e-07,
      "loss": 0.0112,
      "reward": 0.7749122083187103,
      "reward_std": 0.04545323085039854,
      "reward_std_all": 0.04209904931485653,
      "rewards-max/forensics_sinkhorn_reward": 0.9853724539279938,
      "rewards-max/forensics_strict_edge_reward": 0.5343473553657532,
      "rewards-max/forensics_tiou_reward": 0.9902356564998627,
      "rewards/forensics_sinkhorn_reward": 0.7898814678192139,
      "rewards/forensics_strict_edge_reward": 0.06032688729465008,
      "rewards/forensics_tiou_reward": 0.6580716967582703,
      "step": 530
    },
    {
      "completion_length": 22.359375,
      "epoch": 2.221757322175732,
      "grad_norm": 5.5961127281188965,
      "kl": 0.330078125,
      "learning_rate": 4.456066945606695e-07,
      "loss": 0.0451,
      "reward": 0.7546485662460327,
      "reward_std": 0.055015914142131805,
      "reward_std_all": 0.05442176200449467,
      "rewards-max/forensics_sinkhorn_reward": 0.9763966500759125,
      "rewards-max/forensics_strict_edge_reward": 0.48137304186820984,
      "rewards-max/forensics_tiou_reward": 0.9642209708690643,
      "rewards/forensics_sinkhorn_reward": 0.7258455455303192,
      "rewards/forensics_strict_edge_reward": 0.03417875710874796,
      "rewards/forensics_tiou_reward": 0.529145210981369,
      "step": 531
    },
    {
      "completion_length": 23.75,
      "epoch": 2.225941422594142,
      "grad_norm": 6.676736354827881,
      "kl": 0.49609375,
      "learning_rate": 4.445606694560669e-07,
      "loss": 0.0931,
      "reward": 0.73084756731987,
      "reward_std": 0.062487099319696426,
      "reward_std_all": 0.06767560727894306,
      "rewards-max/forensics_sinkhorn_reward": 0.9769497811794281,
      "rewards-max/forensics_strict_edge_reward": 0.22929870337247849,
      "rewards-max/forensics_tiou_reward": 0.9620125591754913,
      "rewards/forensics_sinkhorn_reward": 0.6384852230548859,
      "rewards/forensics_strict_edge_reward": 0.022513657808303833,
      "rewards/forensics_tiou_reward": 0.44792014360427856,
      "step": 532
    },
    {
      "completion_length": 22.828125,
      "epoch": 2.2301255230125525,
      "grad_norm": 6.121236324310303,
      "kl": 0.3046875,
      "learning_rate": 4.435146443514644e-07,
      "loss": 0.0721,
      "reward": 0.7058646976947784,
      "reward_std": 0.08010389655828476,
      "reward_std_all": 0.0914805456995964,
      "rewards-max/forensics_sinkhorn_reward": 0.9141372442245483,
      "rewards-max/forensics_strict_edge_reward": 0.16937285661697388,
      "rewards-max/forensics_tiou_reward": 0.8252792060375214,
      "rewards/forensics_sinkhorn_reward": 0.61241215467453,
      "rewards/forensics_strict_edge_reward": 0.013101087883114815,
      "rewards/forensics_tiou_reward": 0.38668395578861237,
      "step": 533
    },
    {
      "completion_length": 20.546875,
      "epoch": 2.2343096234309625,
      "grad_norm": 5.990094184875488,
      "kl": 0.4619140625,
      "learning_rate": 4.424686192468619e-07,
      "loss": 0.0353,
      "reward": 0.7728227376937866,
      "reward_std": 0.05247142072767019,
      "reward_std_all": 0.04802865441888571,
      "rewards-max/forensics_sinkhorn_reward": 0.9869913458824158,
      "rewards-max/forensics_strict_edge_reward": 0.6966203600168228,
      "rewards-max/forensics_tiou_reward": 0.992585301399231,
      "rewards/forensics_sinkhorn_reward": 0.7342154383659363,
      "rewards/forensics_strict_edge_reward": 0.04405607841908932,
      "rewards/forensics_tiou_reward": 0.5135372877120972,
      "step": 534
    },
    {
      "completion_length": 22.15625,
      "epoch": 2.2384937238493725,
      "grad_norm": 7.552720546722412,
      "kl": 0.2890625,
      "learning_rate": 4.414225941422594e-07,
      "loss": 0.0516,
      "reward": 0.7052906453609467,
      "reward_std": 0.0766146332025528,
      "reward_std_all": 0.08920122683048248,
      "rewards-max/forensics_sinkhorn_reward": 0.9530175924301147,
      "rewards-max/forensics_strict_edge_reward": 0.41129170358181,
      "rewards-max/forensics_tiou_reward": 0.8973096311092377,
      "rewards/forensics_sinkhorn_reward": 0.6737342774868011,
      "rewards/forensics_strict_edge_reward": 0.01777351228520274,
      "rewards/forensics_tiou_reward": 0.45308272540569305,
      "step": 535
    },
    {
      "completion_length": 22.9375,
      "epoch": 2.2426778242677825,
      "grad_norm": 8.079556465148926,
      "kl": 0.3173828125,
      "learning_rate": 4.403765690376569e-07,
      "loss": 0.0018,
      "reward": 0.7599369287490845,
      "reward_std": 0.05397089570760727,
      "reward_std_all": 0.05257314629852772,
      "rewards-max/forensics_sinkhorn_reward": 0.967177152633667,
      "rewards-max/forensics_strict_edge_reward": 0.41780946403741837,
      "rewards-max/forensics_tiou_reward": 0.9417897164821625,
      "rewards/forensics_sinkhorn_reward": 0.7103196978569031,
      "rewards/forensics_strict_edge_reward": 0.03653226140886545,
      "rewards/forensics_tiou_reward": 0.516249030828476,
      "step": 536
    },
    {
      "completion_length": 21.21875,
      "epoch": 2.2468619246861925,
      "grad_norm": 8.230122566223145,
      "kl": 0.2939453125,
      "learning_rate": 4.3933054393305435e-07,
      "loss": 0.0333,
      "reward": 0.7520603537559509,
      "reward_std": 0.04098569322377443,
      "reward_std_all": 0.04820871911942959,
      "rewards-max/forensics_sinkhorn_reward": 0.9804017543792725,
      "rewards-max/forensics_strict_edge_reward": 0.7797744870185852,
      "rewards-max/forensics_tiou_reward": 0.9757785797119141,
      "rewards/forensics_sinkhorn_reward": 0.7170892953872681,
      "rewards/forensics_strict_edge_reward": 0.04490376450121403,
      "rewards/forensics_tiou_reward": 0.48518499732017517,
      "step": 537
    },
    {
      "completion_length": 21.0625,
      "epoch": 2.2510460251046025,
      "grad_norm": 6.554934024810791,
      "kl": 0.2822265625,
      "learning_rate": 4.3828451882845186e-07,
      "loss": 0.0558,
      "reward": 0.7736446261405945,
      "reward_std": 0.03716297075152397,
      "reward_std_all": 0.03676586598157883,
      "rewards-max/forensics_sinkhorn_reward": 0.9726031720638275,
      "rewards-max/forensics_strict_edge_reward": 0.42878545820713043,
      "rewards-max/forensics_tiou_reward": 0.9552377760410309,
      "rewards/forensics_sinkhorn_reward": 0.7141989767551422,
      "rewards/forensics_strict_edge_reward": 0.041515142656862736,
      "rewards/forensics_tiou_reward": 0.48142997920513153,
      "step": 538
    },
    {
      "completion_length": 20.5,
      "epoch": 2.2552301255230125,
      "grad_norm": 7.459859371185303,
      "kl": 0.421875,
      "learning_rate": 4.3723849372384937e-07,
      "loss": 0.1542,
      "reward": 0.7300178408622742,
      "reward_std": 0.055414775386452675,
      "reward_std_all": 0.06356049329042435,
      "rewards-max/forensics_sinkhorn_reward": 0.9607014954090118,
      "rewards-max/forensics_strict_edge_reward": 0.3119669482111931,
      "rewards-max/forensics_tiou_reward": 0.9301281273365021,
      "rewards/forensics_sinkhorn_reward": 0.6796320378780365,
      "rewards/forensics_strict_edge_reward": 0.016597357112914324,
      "rewards/forensics_tiou_reward": 0.4501732438802719,
      "step": 539
    },
    {
      "completion_length": 24.140625,
      "epoch": 2.2594142259414225,
      "grad_norm": 6.163105010986328,
      "kl": 0.3623046875,
      "learning_rate": 4.3619246861924683e-07,
      "loss": 0.0826,
      "reward": 0.704031765460968,
      "reward_std": 0.0806976892054081,
      "reward_std_all": 0.0879563856869936,
      "rewards-max/forensics_sinkhorn_reward": 0.9706829786300659,
      "rewards-max/forensics_strict_edge_reward": 0.34477679431438446,
      "rewards-max/forensics_tiou_reward": 0.9503747224807739,
      "rewards/forensics_sinkhorn_reward": 0.6434873342514038,
      "rewards/forensics_strict_edge_reward": 0.033750975504517555,
      "rewards/forensics_tiou_reward": 0.4192223846912384,
      "step": 540
    },
    {
      "completion_length": 20.890625,
      "epoch": 2.2635983263598325,
      "grad_norm": 8.360313415527344,
      "kl": 0.2744140625,
      "learning_rate": 4.3514644351464434e-07,
      "loss": -0.0261,
      "reward": 0.7664194703102112,
      "reward_std": 0.03859551181085408,
      "reward_std_all": 0.03944535041227937,
      "rewards-max/forensics_sinkhorn_reward": 0.9258929193019867,
      "rewards-max/forensics_strict_edge_reward": 0.6077391505241394,
      "rewards-max/forensics_tiou_reward": 0.8437467217445374,
      "rewards/forensics_sinkhorn_reward": 0.7406983077526093,
      "rewards/forensics_strict_edge_reward": 0.03407333791255951,
      "rewards/forensics_tiou_reward": 0.519870787858963,
      "step": 541
    },
    {
      "completion_length": 21.25,
      "epoch": 2.2677824267782425,
      "grad_norm": 5.1699323654174805,
      "kl": 0.3603515625,
      "learning_rate": 4.341004184100418e-07,
      "loss": 0.0171,
      "reward": 0.7690790295600891,
      "reward_std": 0.04580058064311743,
      "reward_std_all": 0.044255830347537994,
      "rewards-max/forensics_sinkhorn_reward": 0.957034558057785,
      "rewards-max/forensics_strict_edge_reward": 0.4200657457113266,
      "rewards-max/forensics_tiou_reward": 0.9205863177776337,
      "rewards/forensics_sinkhorn_reward": 0.7324811220169067,
      "rewards/forensics_strict_edge_reward": 0.04690703563392162,
      "rewards/forensics_tiou_reward": 0.5477247536182404,
      "step": 542
    },
    {
      "completion_length": 21.53125,
      "epoch": 2.2719665271966525,
      "grad_norm": 5.176781177520752,
      "kl": 0.3369140625,
      "learning_rate": 4.330543933054393e-07,
      "loss": 0.1499,
      "reward": 0.769065797328949,
      "reward_std": 0.04267500527203083,
      "reward_std_all": 0.041926782578229904,
      "rewards-max/forensics_sinkhorn_reward": 0.9782216250896454,
      "rewards-max/forensics_strict_edge_reward": 0.218681450933218,
      "rewards-max/forensics_tiou_reward": 0.9673289954662323,
      "rewards/forensics_sinkhorn_reward": 0.7083698511123657,
      "rewards/forensics_strict_edge_reward": 0.015782953589223325,
      "rewards/forensics_tiou_reward": 0.4839823395013809,
      "step": 543
    },
    {
      "completion_length": 20.765625,
      "epoch": 2.276150627615063,
      "grad_norm": 5.569972991943359,
      "kl": 0.33984375,
      "learning_rate": 4.3200836820083683e-07,
      "loss": -0.0207,
      "reward": 0.7803143560886383,
      "reward_std": 0.03921561874449253,
      "reward_std_all": 0.03660288080573082,
      "rewards-max/forensics_sinkhorn_reward": 0.9609420597553253,
      "rewards-max/forensics_strict_edge_reward": 0.36461610347032547,
      "rewards-max/forensics_tiou_reward": 0.9314334392547607,
      "rewards/forensics_sinkhorn_reward": 0.7769317328929901,
      "rewards/forensics_strict_edge_reward": 0.03387465886771679,
      "rewards/forensics_tiou_reward": 0.5640237033367157,
      "step": 544
    },
    {
      "completion_length": 20.25,
      "epoch": 2.280334728033473,
      "grad_norm": 6.876011371612549,
      "kl": 0.3310546875,
      "learning_rate": 4.309623430962343e-07,
      "loss": -0.0018,
      "reward": 0.7845838665962219,
      "reward_std": 0.03333703428506851,
      "reward_std_all": 0.03147376608103514,
      "rewards-max/forensics_sinkhorn_reward": 0.9499436914920807,
      "rewards-max/forensics_strict_edge_reward": 0.257596614130307,
      "rewards-max/forensics_tiou_reward": 0.8688383102416992,
      "rewards/forensics_sinkhorn_reward": 0.7646774649620056,
      "rewards/forensics_strict_edge_reward": 0.04948653527389979,
      "rewards/forensics_tiou_reward": 0.5144681036472321,
      "step": 545
    },
    {
      "completion_length": 20.75,
      "epoch": 2.284518828451883,
      "grad_norm": 6.166548728942871,
      "kl": 0.388671875,
      "learning_rate": 4.299163179916318e-07,
      "loss": 0.0749,
      "reward": 0.7606906294822693,
      "reward_std": 0.0452542919665575,
      "reward_std_all": 0.046663958579301834,
      "rewards-max/forensics_sinkhorn_reward": 0.9829438626766205,
      "rewards-max/forensics_strict_edge_reward": 0.5870097801089287,
      "rewards-max/forensics_tiou_reward": 0.9835634231567383,
      "rewards/forensics_sinkhorn_reward": 0.709060400724411,
      "rewards/forensics_strict_edge_reward": 0.056885555386543274,
      "rewards/forensics_tiou_reward": 0.48069503903388977,
      "step": 546
    },
    {
      "completion_length": 23.125,
      "epoch": 2.288702928870293,
      "grad_norm": 6.252879619598389,
      "kl": 0.3779296875,
      "learning_rate": 4.2887029288702926e-07,
      "loss": 0.0311,
      "reward": 0.6749285161495209,
      "reward_std": 0.049489385448396206,
      "reward_std_all": 0.08635985851287842,
      "rewards-max/forensics_sinkhorn_reward": 0.9627875983715057,
      "rewards-max/forensics_strict_edge_reward": 0.303529292345047,
      "rewards-max/forensics_tiou_reward": 0.9343119561672211,
      "rewards/forensics_sinkhorn_reward": 0.5953208208084106,
      "rewards/forensics_strict_edge_reward": 0.02975131943821907,
      "rewards/forensics_tiou_reward": 0.3826346695423126,
      "step": 547
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.292887029288703,
      "grad_norm": 4.859904766082764,
      "kl": 0.501953125,
      "learning_rate": 4.278242677824268e-07,
      "loss": 0.1571,
      "reward": 0.7490333914756775,
      "reward_std": 0.08923264965415001,
      "reward_std_all": 0.08013684302568436,
      "rewards-max/forensics_sinkhorn_reward": 0.9769763946533203,
      "rewards-max/forensics_strict_edge_reward": 0.5015955567359924,
      "rewards-max/forensics_tiou_reward": 0.9673238694667816,
      "rewards/forensics_sinkhorn_reward": 0.6974615156650543,
      "rewards/forensics_strict_edge_reward": 0.05825253576040268,
      "rewards/forensics_tiou_reward": 0.513943076133728,
      "step": 548
    },
    {
      "completion_length": 19.796875,
      "epoch": 2.297071129707113,
      "grad_norm": 4.916119575500488,
      "kl": 0.37109375,
      "learning_rate": 4.2677824267782424e-07,
      "loss": -0.0734,
      "reward": 0.7802632451057434,
      "reward_std": 0.042528790421783924,
      "reward_std_all": 0.039229921996593475,
      "rewards-max/forensics_sinkhorn_reward": 0.975535124540329,
      "rewards-max/forensics_strict_edge_reward": 0.25927625596523285,
      "rewards-max/forensics_tiou_reward": 0.9658862054347992,
      "rewards/forensics_sinkhorn_reward": 0.7537071704864502,
      "rewards/forensics_strict_edge_reward": 0.023394706659018993,
      "rewards/forensics_tiou_reward": 0.5362512469291687,
      "step": 549
    },
    {
      "completion_length": 23.03125,
      "epoch": 2.301255230125523,
      "grad_norm": 5.875505447387695,
      "kl": 0.3134765625,
      "learning_rate": 4.2573221757322175e-07,
      "loss": 0.0804,
      "reward": 0.7655788064002991,
      "reward_std": 0.04842782020568848,
      "reward_std_all": 0.04681665450334549,
      "rewards-max/forensics_sinkhorn_reward": 0.9580846726894379,
      "rewards-max/forensics_strict_edge_reward": 0.2686737570911646,
      "rewards-max/forensics_tiou_reward": 0.9149530827999115,
      "rewards/forensics_sinkhorn_reward": 0.6910441517829895,
      "rewards/forensics_strict_edge_reward": 0.019928644876927137,
      "rewards/forensics_tiou_reward": 0.5001556724309921,
      "step": 550
    },
    {
      "completion_length": 22.796875,
      "epoch": 2.305439330543933,
      "grad_norm": 6.006573677062988,
      "kl": 0.3603515625,
      "learning_rate": 4.2468619246861926e-07,
      "loss": 0.0681,
      "reward": 0.7451353371143341,
      "reward_std": 0.0401482880115509,
      "reward_std_all": 0.052333906292915344,
      "rewards-max/forensics_sinkhorn_reward": 0.8948642313480377,
      "rewards-max/forensics_strict_edge_reward": 0.14000378921627998,
      "rewards-max/forensics_tiou_reward": 0.7676441669464111,
      "rewards/forensics_sinkhorn_reward": 0.6815985441207886,
      "rewards/forensics_strict_edge_reward": 0.01206212304532528,
      "rewards/forensics_tiou_reward": 0.4228862375020981,
      "step": 551
    },
    {
      "completion_length": 20.625,
      "epoch": 2.309623430962343,
      "grad_norm": 5.273921012878418,
      "kl": 0.2294921875,
      "learning_rate": 4.236401673640167e-07,
      "loss": 0.001,
      "reward": 0.7868504524230957,
      "reward_std": 0.016753327916376293,
      "reward_std_all": 0.0191330355592072,
      "rewards-max/forensics_sinkhorn_reward": 0.9362254738807678,
      "rewards-max/forensics_strict_edge_reward": 0.3196670934557915,
      "rewards-max/forensics_tiou_reward": 0.8722323775291443,
      "rewards/forensics_sinkhorn_reward": 0.7407753169536591,
      "rewards/forensics_strict_edge_reward": 0.019996498711407185,
      "rewards/forensics_tiou_reward": 0.5017451196908951,
      "step": 552
    },
    {
      "completion_length": 21.15625,
      "epoch": 2.3138075313807533,
      "grad_norm": 5.686520576477051,
      "kl": 0.3125,
      "learning_rate": 4.225941422594142e-07,
      "loss": -0.0418,
      "reward": 0.7847346663475037,
      "reward_std": 0.03246236313134432,
      "reward_std_all": 0.030839945189654827,
      "rewards-max/forensics_sinkhorn_reward": 0.9773309826850891,
      "rewards-max/forensics_strict_edge_reward": 0.2743282364681363,
      "rewards-max/forensics_tiou_reward": 0.9577438235282898,
      "rewards/forensics_sinkhorn_reward": 0.7488189339637756,
      "rewards/forensics_strict_edge_reward": 0.019825437222607434,
      "rewards/forensics_tiou_reward": 0.501383364200592,
      "step": 553
    },
    {
      "completion_length": 21.234375,
      "epoch": 2.3179916317991633,
      "grad_norm": 6.085212707519531,
      "kl": 0.4248046875,
      "learning_rate": 4.215481171548117e-07,
      "loss": 0.0532,
      "reward": 0.7845423221588135,
      "reward_std": 0.02983382623642683,
      "reward_std_all": 0.028996167704463005,
      "rewards-max/forensics_sinkhorn_reward": 0.9537221491336823,
      "rewards-max/forensics_strict_edge_reward": 0.27923316322267056,
      "rewards-max/forensics_tiou_reward": 0.9048877656459808,
      "rewards/forensics_sinkhorn_reward": 0.7310264408588409,
      "rewards/forensics_strict_edge_reward": 0.031040977919474244,
      "rewards/forensics_tiou_reward": 0.484200581908226,
      "step": 554
    },
    {
      "completion_length": 21.0,
      "epoch": 2.3221757322175733,
      "grad_norm": 7.573530197143555,
      "kl": 0.330078125,
      "learning_rate": 4.205020920502092e-07,
      "loss": -0.0053,
      "reward": 0.7495649755001068,
      "reward_std": 0.048371415585279465,
      "reward_std_all": 0.0507006011903286,
      "rewards-max/forensics_sinkhorn_reward": 0.9660416841506958,
      "rewards-max/forensics_strict_edge_reward": 0.5493997037410736,
      "rewards-max/forensics_tiou_reward": 0.9453352093696594,
      "rewards/forensics_sinkhorn_reward": 0.7326563894748688,
      "rewards/forensics_strict_edge_reward": 0.037509193643927574,
      "rewards/forensics_tiou_reward": 0.5153211504220963,
      "step": 555
    },
    {
      "completion_length": 20.796875,
      "epoch": 2.3263598326359833,
      "grad_norm": 7.053199291229248,
      "kl": 0.462890625,
      "learning_rate": 4.194560669456067e-07,
      "loss": 0.1042,
      "reward": 0.755298376083374,
      "reward_std": 0.054808950051665306,
      "reward_std_all": 0.0532638244330883,
      "rewards-max/forensics_sinkhorn_reward": 0.9632160663604736,
      "rewards-max/forensics_strict_edge_reward": 0.3934844136238098,
      "rewards-max/forensics_tiou_reward": 0.9360815286636353,
      "rewards/forensics_sinkhorn_reward": 0.7252266705036163,
      "rewards/forensics_strict_edge_reward": 0.030275178141891956,
      "rewards/forensics_tiou_reward": 0.4850057363510132,
      "step": 556
    },
    {
      "completion_length": 21.40625,
      "epoch": 2.3305439330543933,
      "grad_norm": 5.458123683929443,
      "kl": 0.34765625,
      "learning_rate": 4.184100418410041e-07,
      "loss": 0.0274,
      "reward": 0.7883680164813995,
      "reward_std": 0.023614749312400818,
      "reward_std_all": 0.023409110493957996,
      "rewards-max/forensics_sinkhorn_reward": 0.97338005900383,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9653326570987701,
      "rewards/forensics_sinkhorn_reward": 0.7284335792064667,
      "rewards/forensics_strict_edge_reward": 0.05524369142949581,
      "rewards/forensics_tiou_reward": 0.5311345458030701,
      "step": 557
    },
    {
      "completion_length": 20.765625,
      "epoch": 2.3347280334728033,
      "grad_norm": 5.407232284545898,
      "kl": 0.33203125,
      "learning_rate": 4.1736401673640164e-07,
      "loss": 0.0643,
      "reward": 0.7533629834651947,
      "reward_std": 0.04703969834372401,
      "reward_std_all": 0.05015408433973789,
      "rewards-max/forensics_sinkhorn_reward": 0.9882792234420776,
      "rewards-max/forensics_strict_edge_reward": 0.952418714761734,
      "rewards-max/forensics_tiou_reward": 0.9981286525726318,
      "rewards/forensics_sinkhorn_reward": 0.720967173576355,
      "rewards/forensics_strict_edge_reward": 0.0865778960287571,
      "rewards/forensics_tiou_reward": 0.5369281023740768,
      "step": 558
    },
    {
      "completion_length": 19.4375,
      "epoch": 2.3389121338912133,
      "grad_norm": 6.46774959564209,
      "kl": 0.3134765625,
      "learning_rate": 4.1631799163179915e-07,
      "loss": -0.1219,
      "reward": 0.7958614230155945,
      "reward_std": 0.017602015985175967,
      "reward_std_all": 0.017086173174902797,
      "rewards-max/forensics_sinkhorn_reward": 0.9829216301441193,
      "rewards-max/forensics_strict_edge_reward": 0.7654097378253937,
      "rewards-max/forensics_tiou_reward": 0.9819678962230682,
      "rewards/forensics_sinkhorn_reward": 0.8050283789634705,
      "rewards/forensics_strict_edge_reward": 0.08594805374741554,
      "rewards/forensics_tiou_reward": 0.5970264375209808,
      "step": 559
    },
    {
      "completion_length": 20.921875,
      "epoch": 2.3430962343096233,
      "grad_norm": 31.739713668823242,
      "kl": 1.74609375,
      "learning_rate": 4.1527196652719666e-07,
      "loss": 0.1201,
      "reward": 0.7708094120025635,
      "reward_std": 0.03839240409433842,
      "reward_std_all": 0.03818396292626858,
      "rewards-max/forensics_sinkhorn_reward": 0.9593334197998047,
      "rewards-max/forensics_strict_edge_reward": 0.29184726625680923,
      "rewards-max/forensics_tiou_reward": 0.9276761114597321,
      "rewards/forensics_sinkhorn_reward": 0.7550023794174194,
      "rewards/forensics_strict_edge_reward": 0.028730168472975492,
      "rewards/forensics_tiou_reward": 0.5508022308349609,
      "step": 560
    },
    {
      "completion_length": 20.640625,
      "epoch": 2.3472803347280333,
      "grad_norm": 6.240365028381348,
      "kl": 0.3525390625,
      "learning_rate": 4.142259414225941e-07,
      "loss": 0.0352,
      "reward": 0.7848774790763855,
      "reward_std": 0.03146578557789326,
      "reward_std_all": 0.02985553629696369,
      "rewards-max/forensics_sinkhorn_reward": 0.9653988480567932,
      "rewards-max/forensics_strict_edge_reward": 0.3584480918943882,
      "rewards-max/forensics_tiou_reward": 0.9394569993019104,
      "rewards/forensics_sinkhorn_reward": 0.7555091977119446,
      "rewards/forensics_strict_edge_reward": 0.025822060415521264,
      "rewards/forensics_tiou_reward": 0.530600368976593,
      "step": 561
    },
    {
      "completion_length": 21.15625,
      "epoch": 2.3514644351464433,
      "grad_norm": 6.596576690673828,
      "kl": 0.3447265625,
      "learning_rate": 4.131799163179916e-07,
      "loss": 0.0669,
      "reward": 0.7477467656135559,
      "reward_std": 0.06711270287632942,
      "reward_std_all": 0.06456545740365982,
      "rewards-max/forensics_sinkhorn_reward": 0.9613618850708008,
      "rewards-max/forensics_strict_edge_reward": 0.4548458904027939,
      "rewards-max/forensics_tiou_reward": 0.9265267252922058,
      "rewards/forensics_sinkhorn_reward": 0.6910860240459442,
      "rewards/forensics_strict_edge_reward": 0.032654667273163795,
      "rewards/forensics_tiou_reward": 0.46322259306907654,
      "step": 562
    },
    {
      "completion_length": 22.28125,
      "epoch": 2.3556485355648533,
      "grad_norm": 6.156699180603027,
      "kl": 0.4052734375,
      "learning_rate": 4.121338912133891e-07,
      "loss": 0.0834,
      "reward": 0.7726151049137115,
      "reward_std": 0.05801522172987461,
      "reward_std_all": 0.05195144563913345,
      "rewards-max/forensics_sinkhorn_reward": 0.9861723482608795,
      "rewards-max/forensics_strict_edge_reward": 0.7367854118347168,
      "rewards-max/forensics_tiou_reward": 0.9924668967723846,
      "rewards/forensics_sinkhorn_reward": 0.7214900851249695,
      "rewards/forensics_strict_edge_reward": 0.0840938463807106,
      "rewards/forensics_tiou_reward": 0.5832332670688629,
      "step": 563
    },
    {
      "completion_length": 20.703125,
      "epoch": 2.3598326359832638,
      "grad_norm": 4.34193229675293,
      "kl": 0.41796875,
      "learning_rate": 4.110878661087866e-07,
      "loss": 0.0947,
      "reward": 0.7815786004066467,
      "reward_std": 0.03464415483176708,
      "reward_std_all": 0.0330157857388258,
      "rewards-max/forensics_sinkhorn_reward": 0.9697867333889008,
      "rewards-max/forensics_strict_edge_reward": 0.27234726399183273,
      "rewards-max/forensics_tiou_reward": 0.9438251852989197,
      "rewards/forensics_sinkhorn_reward": 0.7511564493179321,
      "rewards/forensics_strict_edge_reward": 0.01943969028070569,
      "rewards/forensics_tiou_reward": 0.561275839805603,
      "step": 564
    },
    {
      "completion_length": 22.265625,
      "epoch": 2.3640167364016738,
      "grad_norm": 5.8197712898254395,
      "kl": 0.35546875,
      "learning_rate": 4.100418410041841e-07,
      "loss": -0.0021,
      "reward": 0.779317319393158,
      "reward_std": 0.015392812434583902,
      "reward_std_all": 0.022821961902081966,
      "rewards-max/forensics_sinkhorn_reward": 0.9802094399929047,
      "rewards-max/forensics_strict_edge_reward": 0.42878545820713043,
      "rewards-max/forensics_tiou_reward": 0.9749733209609985,
      "rewards/forensics_sinkhorn_reward": 0.737059623003006,
      "rewards/forensics_strict_edge_reward": 0.05294438824057579,
      "rewards/forensics_tiou_reward": 0.5331471562385559,
      "step": 565
    },
    {
      "completion_length": 20.3125,
      "epoch": 2.3682008368200838,
      "grad_norm": 9.985532760620117,
      "kl": 0.306640625,
      "learning_rate": 4.089958158995816e-07,
      "loss": 0.0235,
      "reward": 0.7727871835231781,
      "reward_std": 0.032879894599318504,
      "reward_std_all": 0.034358917735517025,
      "rewards-max/forensics_sinkhorn_reward": 0.9814810752868652,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.979020893573761,
      "rewards/forensics_sinkhorn_reward": 0.7496477663516998,
      "rewards/forensics_strict_edge_reward": 0.07402149215340614,
      "rewards/forensics_tiou_reward": 0.5196117460727692,
      "step": 566
    },
    {
      "completion_length": 21.828125,
      "epoch": 2.3723849372384938,
      "grad_norm": 6.226570129394531,
      "kl": 0.4736328125,
      "learning_rate": 4.0794979079497904e-07,
      "loss": 0.0982,
      "reward": 0.7714107930660248,
      "reward_std": 0.043555399402976036,
      "reward_std_all": 0.042099323123693466,
      "rewards-max/forensics_sinkhorn_reward": 0.9479502141475677,
      "rewards-max/forensics_strict_edge_reward": 0.273585494607687,
      "rewards-max/forensics_tiou_reward": 0.8829293847084045,
      "rewards/forensics_sinkhorn_reward": 0.7291375398635864,
      "rewards/forensics_strict_edge_reward": 0.019732341403141618,
      "rewards/forensics_tiou_reward": 0.5074219405651093,
      "step": 567
    },
    {
      "completion_length": 21.453125,
      "epoch": 2.3765690376569037,
      "grad_norm": 6.666354656219482,
      "kl": 0.283203125,
      "learning_rate": 4.0690376569037656e-07,
      "loss": -0.0868,
      "reward": 0.7290050685405731,
      "reward_std": 0.03030753880739212,
      "reward_std_all": 0.054643915966153145,
      "rewards-max/forensics_sinkhorn_reward": 0.9707790017127991,
      "rewards-max/forensics_strict_edge_reward": 0.6087096929550171,
      "rewards-max/forensics_tiou_reward": 0.9576899111270905,
      "rewards/forensics_sinkhorn_reward": 0.7105794250965118,
      "rewards/forensics_strict_edge_reward": 0.03837031312286854,
      "rewards/forensics_tiou_reward": 0.5029108822345734,
      "step": 568
    },
    {
      "completion_length": 21.703125,
      "epoch": 2.3807531380753137,
      "grad_norm": 5.240201473236084,
      "kl": 0.3076171875,
      "learning_rate": 4.0585774058577407e-07,
      "loss": -0.0244,
      "reward": 0.763378918170929,
      "reward_std": 0.04547612660098821,
      "reward_std_all": 0.04522803227882832,
      "rewards-max/forensics_sinkhorn_reward": 0.9810831546783447,
      "rewards-max/forensics_strict_edge_reward": 0.4295876109972596,
      "rewards-max/forensics_tiou_reward": 0.977663516998291,
      "rewards/forensics_sinkhorn_reward": 0.7370702922344208,
      "rewards/forensics_strict_edge_reward": 0.03263171701109968,
      "rewards/forensics_tiou_reward": 0.512683093547821,
      "step": 569
    },
    {
      "completion_length": 21.75,
      "epoch": 2.3849372384937237,
      "grad_norm": 5.556040287017822,
      "kl": 0.5146484375,
      "learning_rate": 4.0481171548117153e-07,
      "loss": -0.0015,
      "reward": 0.7756960391998291,
      "reward_std": 0.038925934582948685,
      "reward_std_all": 0.037529220804572105,
      "rewards-max/forensics_sinkhorn_reward": 0.9849941432476044,
      "rewards-max/forensics_strict_edge_reward": 0.7201169431209564,
      "rewards-max/forensics_tiou_reward": 0.986854076385498,
      "rewards/forensics_sinkhorn_reward": 0.761326253414154,
      "rewards/forensics_strict_edge_reward": 0.03935043513774872,
      "rewards/forensics_tiou_reward": 0.5565151572227478,
      "step": 570
    },
    {
      "completion_length": 20.78125,
      "epoch": 2.3891213389121337,
      "grad_norm": 5.334150314331055,
      "kl": 0.39453125,
      "learning_rate": 4.0376569037656904e-07,
      "loss": 0.065,
      "reward": 0.7340821027755737,
      "reward_std": 0.05936156399548054,
      "reward_std_all": 0.06366891227662563,
      "rewards-max/forensics_sinkhorn_reward": 0.983809232711792,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9849041998386383,
      "rewards/forensics_sinkhorn_reward": 0.6862479150295258,
      "rewards/forensics_strict_edge_reward": 0.06983180530369282,
      "rewards/forensics_tiou_reward": 0.5610696375370026,
      "step": 571
    },
    {
      "completion_length": 21.90625,
      "epoch": 2.393305439330544,
      "grad_norm": 6.842750549316406,
      "kl": 0.376953125,
      "learning_rate": 4.027196652719665e-07,
      "loss": 0.0961,
      "reward": 0.7433229386806488,
      "reward_std": 0.05219368077814579,
      "reward_std_all": 0.061012400314211845,
      "rewards-max/forensics_sinkhorn_reward": 0.9786140620708466,
      "rewards-max/forensics_strict_edge_reward": 0.5311143696308136,
      "rewards-max/forensics_tiou_reward": 0.9734722077846527,
      "rewards/forensics_sinkhorn_reward": 0.6881047785282135,
      "rewards/forensics_strict_edge_reward": 0.04499385319650173,
      "rewards/forensics_tiou_reward": 0.46537891030311584,
      "step": 572
    },
    {
      "completion_length": 23.9375,
      "epoch": 2.397489539748954,
      "grad_norm": 5.737908363342285,
      "kl": 0.3212890625,
      "learning_rate": 4.01673640167364e-07,
      "loss": 0.146,
      "reward": 0.731876939535141,
      "reward_std": 0.0596969872713089,
      "reward_std_all": 0.0672269482165575,
      "rewards-max/forensics_sinkhorn_reward": 0.9677876234054565,
      "rewards-max/forensics_strict_edge_reward": 0.20031130127608776,
      "rewards-max/forensics_tiou_reward": 0.9330083429813385,
      "rewards/forensics_sinkhorn_reward": 0.6207310855388641,
      "rewards/forensics_strict_edge_reward": 0.01470986194908619,
      "rewards/forensics_tiou_reward": 0.4523197412490845,
      "step": 573
    },
    {
      "completion_length": 20.78125,
      "epoch": 2.401673640167364,
      "grad_norm": 6.030047416687012,
      "kl": 0.4345703125,
      "learning_rate": 4.006276150627615e-07,
      "loss": 0.0338,
      "reward": 0.7874506413936615,
      "reward_std": 0.02744890283793211,
      "reward_std_all": 0.026626826729625463,
      "rewards-max/forensics_sinkhorn_reward": 0.9742847084999084,
      "rewards-max/forensics_strict_edge_reward": 0.5834526717662811,
      "rewards-max/forensics_tiou_reward": 0.9641818702220917,
      "rewards/forensics_sinkhorn_reward": 0.7468599379062653,
      "rewards/forensics_strict_edge_reward": 0.039216590113937855,
      "rewards/forensics_tiou_reward": 0.4921513944864273,
      "step": 574
    },
    {
      "completion_length": 21.96875,
      "epoch": 2.405857740585774,
      "grad_norm": 6.5489068031311035,
      "kl": 0.3193359375,
      "learning_rate": 3.99581589958159e-07,
      "loss": 0.0279,
      "reward": 0.7831229567527771,
      "reward_std": 0.037434873171150684,
      "reward_std_all": 0.03492602240294218,
      "rewards-max/forensics_sinkhorn_reward": 0.9444615244865417,
      "rewards-max/forensics_strict_edge_reward": 0.10067093744874,
      "rewards-max/forensics_tiou_reward": 0.8853059411048889,
      "rewards/forensics_sinkhorn_reward": 0.736893355846405,
      "rewards/forensics_strict_edge_reward": 0.007515231845900416,
      "rewards/forensics_tiou_reward": 0.5225401222705841,
      "step": 575
    },
    {
      "completion_length": 19.875,
      "epoch": 2.410041841004184,
      "grad_norm": 6.366136074066162,
      "kl": 0.3291015625,
      "learning_rate": 3.985355648535565e-07,
      "loss": 0.063,
      "reward": 0.7763157188892365,
      "reward_std": 0.0399844478815794,
      "reward_std_all": 0.03796793520450592,
      "rewards-max/forensics_sinkhorn_reward": 0.9750275611877441,
      "rewards-max/forensics_strict_edge_reward": 0.6231168210506439,
      "rewards-max/forensics_tiou_reward": 0.9635329246520996,
      "rewards/forensics_sinkhorn_reward": 0.7845514118671417,
      "rewards/forensics_strict_edge_reward": 0.0453585647046566,
      "rewards/forensics_tiou_reward": 0.5754630863666534,
      "step": 576
    },
    {
      "completion_length": 22.890625,
      "epoch": 2.414225941422594,
      "grad_norm": 7.857662200927734,
      "kl": 0.2626953125,
      "learning_rate": 3.9748953974895396e-07,
      "loss": 0.1304,
      "reward": 0.7482703030109406,
      "reward_std": 0.0553492596372962,
      "reward_std_all": 0.059230461716651917,
      "rewards-max/forensics_sinkhorn_reward": 0.9811971187591553,
      "rewards-max/forensics_strict_edge_reward": 0.4231441020965576,
      "rewards-max/forensics_tiou_reward": 0.9707581400871277,
      "rewards/forensics_sinkhorn_reward": 0.6725329756736755,
      "rewards/forensics_strict_edge_reward": 0.03141352254897356,
      "rewards/forensics_tiou_reward": 0.4757254123687744,
      "step": 577
    },
    {
      "completion_length": 20.859375,
      "epoch": 2.418410041841004,
      "grad_norm": 7.143680095672607,
      "kl": 0.3251953125,
      "learning_rate": 3.964435146443514e-07,
      "loss": 0.1057,
      "reward": 0.74436354637146,
      "reward_std": 0.05272084288299084,
      "reward_std_all": 0.05748499557375908,
      "rewards-max/forensics_sinkhorn_reward": 0.9817923009395599,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9802006185054779,
      "rewards/forensics_sinkhorn_reward": 0.6845565736293793,
      "rewards/forensics_strict_edge_reward": 0.028174255043268204,
      "rewards/forensics_tiou_reward": 0.4940173327922821,
      "step": 578
    },
    {
      "completion_length": 21.765625,
      "epoch": 2.422594142259414,
      "grad_norm": 4.686521053314209,
      "kl": 0.4140625,
      "learning_rate": 3.9539748953974893e-07,
      "loss": 0.0103,
      "reward": 0.7336481511592865,
      "reward_std": 0.07181436568498611,
      "reward_std_all": 0.07201912999153137,
      "rewards-max/forensics_sinkhorn_reward": 0.9695665240287781,
      "rewards-max/forensics_strict_edge_reward": 0.3350406885147095,
      "rewards-max/forensics_tiou_reward": 0.9362369477748871,
      "rewards/forensics_sinkhorn_reward": 0.6872188448905945,
      "rewards/forensics_strict_edge_reward": 0.030027256812900305,
      "rewards/forensics_tiou_reward": 0.4646592438220978,
      "step": 579
    },
    {
      "completion_length": 22.765625,
      "epoch": 2.426778242677824,
      "grad_norm": 7.325338363647461,
      "kl": 0.5849609375,
      "learning_rate": 3.9435146443514645e-07,
      "loss": 0.0075,
      "reward": 0.7679519355297089,
      "reward_std": 0.04084074683487415,
      "reward_std_all": 0.04156918451189995,
      "rewards-max/forensics_sinkhorn_reward": 0.9431577920913696,
      "rewards-max/forensics_strict_edge_reward": 0.17100762948393822,
      "rewards-max/forensics_tiou_reward": 0.8865666687488556,
      "rewards/forensics_sinkhorn_reward": 0.7043980658054352,
      "rewards/forensics_strict_edge_reward": 0.0202791765332222,
      "rewards/forensics_tiou_reward": 0.5011756122112274,
      "step": 580
    },
    {
      "completion_length": 20.625,
      "epoch": 2.430962343096234,
      "grad_norm": 7.363012313842773,
      "kl": 0.4111328125,
      "learning_rate": 3.9330543933054396e-07,
      "loss": -0.0386,
      "reward": 0.7688292562961578,
      "reward_std": 0.020679651759564877,
      "reward_std_all": 0.02900865487754345,
      "rewards-max/forensics_sinkhorn_reward": 0.9753446280956268,
      "rewards-max/forensics_strict_edge_reward": 0.5274569988250732,
      "rewards-max/forensics_tiou_reward": 0.9634754061698914,
      "rewards/forensics_sinkhorn_reward": 0.7249014377593994,
      "rewards/forensics_strict_edge_reward": 0.03235720843076706,
      "rewards/forensics_tiou_reward": 0.46588048338890076,
      "step": 581
    },
    {
      "completion_length": 20.34375,
      "epoch": 2.435146443514644,
      "grad_norm": 7.814530372619629,
      "kl": 0.314453125,
      "learning_rate": 3.9225941422594137e-07,
      "loss": -0.0069,
      "reward": 0.7284992635250092,
      "reward_std": 0.0581876989454031,
      "reward_std_all": 0.06934488192200661,
      "rewards-max/forensics_sinkhorn_reward": 0.981071949005127,
      "rewards-max/forensics_strict_edge_reward": 0.7584008872509003,
      "rewards-max/forensics_tiou_reward": 0.9806590676307678,
      "rewards/forensics_sinkhorn_reward": 0.7142841219902039,
      "rewards/forensics_strict_edge_reward": 0.07066502422094345,
      "rewards/forensics_tiou_reward": 0.4980498254299164,
      "step": 582
    },
    {
      "completion_length": 20.234375,
      "epoch": 2.439330543933054,
      "grad_norm": 7.311187267303467,
      "kl": 0.4091796875,
      "learning_rate": 3.912133891213389e-07,
      "loss": -0.0228,
      "reward": 0.7883035242557526,
      "reward_std": 0.02283809194341302,
      "reward_std_all": 0.023083426407538354,
      "rewards-max/forensics_sinkhorn_reward": 0.96694216132164,
      "rewards-max/forensics_strict_edge_reward": 0.34059758484363556,
      "rewards-max/forensics_tiou_reward": 0.9399478435516357,
      "rewards/forensics_sinkhorn_reward": 0.762650728225708,
      "rewards/forensics_strict_edge_reward": 0.04328951984643936,
      "rewards/forensics_tiou_reward": 0.5346298068761826,
      "step": 583
    },
    {
      "completion_length": 21.8125,
      "epoch": 2.4435146443514646,
      "grad_norm": 6.282877445220947,
      "kl": 0.4091796875,
      "learning_rate": 3.901673640167364e-07,
      "loss": 0.064,
      "reward": 0.776572048664093,
      "reward_std": 0.03944743983447552,
      "reward_std_all": 0.03761564288288355,
      "rewards-max/forensics_sinkhorn_reward": 0.9815087616443634,
      "rewards-max/forensics_strict_edge_reward": 0.6660739481449127,
      "rewards-max/forensics_tiou_reward": 0.9821124076843262,
      "rewards/forensics_sinkhorn_reward": 0.7271766066551208,
      "rewards/forensics_strict_edge_reward": 0.05164782702922821,
      "rewards/forensics_tiou_reward": 0.5126066356897354,
      "step": 584
    },
    {
      "completion_length": 21.375,
      "epoch": 2.4476987447698746,
      "grad_norm": 4.995599269866943,
      "kl": 0.375,
      "learning_rate": 3.891213389121339e-07,
      "loss": -0.0561,
      "reward": 0.7601503431797028,
      "reward_std": 0.034875212237238884,
      "reward_std_all": 0.04036402329802513,
      "rewards-max/forensics_sinkhorn_reward": 0.978638082742691,
      "rewards-max/forensics_strict_edge_reward": 0.5159863531589508,
      "rewards-max/forensics_tiou_reward": 0.9726166725158691,
      "rewards/forensics_sinkhorn_reward": 0.7416793704032898,
      "rewards/forensics_strict_edge_reward": 0.06399045884609222,
      "rewards/forensics_tiou_reward": 0.5509640276432037,
      "step": 585
    },
    {
      "completion_length": 21.453125,
      "epoch": 2.4518828451882846,
      "grad_norm": 7.335512638092041,
      "kl": 0.34765625,
      "learning_rate": 3.880753138075314e-07,
      "loss": 0.0561,
      "reward": 0.7395953834056854,
      "reward_std": 0.06946376897394657,
      "reward_std_all": 0.0683269239962101,
      "rewards-max/forensics_sinkhorn_reward": 0.9637340605258942,
      "rewards-max/forensics_strict_edge_reward": 0.3449188619852066,
      "rewards-max/forensics_tiou_reward": 0.9351596832275391,
      "rewards/forensics_sinkhorn_reward": 0.6720137894153595,
      "rewards/forensics_strict_edge_reward": 0.021778536029160023,
      "rewards/forensics_tiou_reward": 0.4372776597738266,
      "step": 586
    },
    {
      "completion_length": 22.765625,
      "epoch": 2.4560669456066946,
      "grad_norm": 7.530570030212402,
      "kl": 0.3564453125,
      "learning_rate": 3.870292887029288e-07,
      "loss": 0.0735,
      "reward": 0.725606232881546,
      "reward_std": 0.06102430820465088,
      "reward_std_all": 0.06904383562505245,
      "rewards-max/forensics_sinkhorn_reward": 0.9544230997562408,
      "rewards-max/forensics_strict_edge_reward": 0.15453355759382248,
      "rewards-max/forensics_tiou_reward": 0.9203377068042755,
      "rewards/forensics_sinkhorn_reward": 0.6659373342990875,
      "rewards/forensics_strict_edge_reward": 0.015284152468666434,
      "rewards/forensics_tiou_reward": 0.43647435307502747,
      "step": 587
    },
    {
      "completion_length": 22.75,
      "epoch": 2.4602510460251046,
      "grad_norm": 4.629430770874023,
      "kl": 0.373046875,
      "learning_rate": 3.8598326359832634e-07,
      "loss": 0.0401,
      "reward": 0.7886288166046143,
      "reward_std": 0.030091707361862063,
      "reward_std_all": 0.027978110127151012,
      "rewards-max/forensics_sinkhorn_reward": 0.9845983684062958,
      "rewards-max/forensics_strict_edge_reward": 0.37596484273672104,
      "rewards-max/forensics_tiou_reward": 0.9858384132385254,
      "rewards/forensics_sinkhorn_reward": 0.7836732268333435,
      "rewards/forensics_strict_edge_reward": 0.03540014661848545,
      "rewards/forensics_tiou_reward": 0.6542019546031952,
      "step": 588
    },
    {
      "completion_length": 19.984375,
      "epoch": 2.4644351464435146,
      "grad_norm": 6.0668230056762695,
      "kl": 0.5244140625,
      "learning_rate": 3.8493723849372385e-07,
      "loss": -0.0469,
      "reward": 0.7914779484272003,
      "reward_std": 0.023991727619431913,
      "reward_std_all": 0.02320765657350421,
      "rewards-max/forensics_sinkhorn_reward": 0.9788135886192322,
      "rewards-max/forensics_strict_edge_reward": 0.4712325632572174,
      "rewards-max/forensics_tiou_reward": 0.9717918634414673,
      "rewards/forensics_sinkhorn_reward": 0.7725204229354858,
      "rewards/forensics_strict_edge_reward": 0.045790284872055054,
      "rewards/forensics_tiou_reward": 0.5562600195407867,
      "step": 589
    },
    {
      "completion_length": 22.359375,
      "epoch": 2.4686192468619246,
      "grad_norm": 6.2418084144592285,
      "kl": 0.361328125,
      "learning_rate": 3.8389121338912136e-07,
      "loss": 0.0527,
      "reward": 0.7818967998027802,
      "reward_std": 0.025978206656873226,
      "reward_std_all": 0.026874479837715626,
      "rewards-max/forensics_sinkhorn_reward": 0.9643560647964478,
      "rewards-max/forensics_strict_edge_reward": 0.3400110676884651,
      "rewards-max/forensics_tiou_reward": 0.9355396330356598,
      "rewards/forensics_sinkhorn_reward": 0.7155098617076874,
      "rewards/forensics_strict_edge_reward": 0.025989562273025513,
      "rewards/forensics_tiou_reward": 0.4842225909233093,
      "step": 590
    },
    {
      "completion_length": 20.203125,
      "epoch": 2.4728033472803346,
      "grad_norm": 8.702220916748047,
      "kl": 0.53515625,
      "learning_rate": 3.828451882845188e-07,
      "loss": 0.0198,
      "reward": 0.7655745148658752,
      "reward_std": 0.03987490385770798,
      "reward_std_all": 0.04124786891043186,
      "rewards-max/forensics_sinkhorn_reward": 0.985835999250412,
      "rewards-max/forensics_strict_edge_reward": 0.6664355397224426,
      "rewards-max/forensics_tiou_reward": 0.991701066493988,
      "rewards/forensics_sinkhorn_reward": 0.7300769686698914,
      "rewards/forensics_strict_edge_reward": 0.07795061357319355,
      "rewards/forensics_tiou_reward": 0.5279337167739868,
      "step": 591
    },
    {
      "completion_length": 20.046875,
      "epoch": 2.476987447698745,
      "grad_norm": 10.636459350585938,
      "kl": 0.3876953125,
      "learning_rate": 3.817991631799163e-07,
      "loss": -0.0409,
      "reward": 0.7427354454994202,
      "reward_std": 0.051863616332411766,
      "reward_std_all": 0.05994441919028759,
      "rewards-max/forensics_sinkhorn_reward": 0.9732015430927277,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9551556408405304,
      "rewards/forensics_sinkhorn_reward": 0.728901594877243,
      "rewards/forensics_strict_edge_reward": 0.05775525327771902,
      "rewards/forensics_tiou_reward": 0.517715647816658,
      "step": 592
    },
    {
      "completion_length": 23.671875,
      "epoch": 2.481171548117155,
      "grad_norm": 4.754887104034424,
      "kl": 0.4150390625,
      "learning_rate": 3.807531380753138e-07,
      "loss": 0.0831,
      "reward": 0.7278524041175842,
      "reward_std": 0.06853343546390533,
      "reward_std_all": 0.07118546962738037,
      "rewards-max/forensics_sinkhorn_reward": 0.9665742218494415,
      "rewards-max/forensics_strict_edge_reward": 0.12756267562508583,
      "rewards-max/forensics_tiou_reward": 0.9468056261539459,
      "rewards/forensics_sinkhorn_reward": 0.6868389248847961,
      "rewards/forensics_strict_edge_reward": 0.009853971190750599,
      "rewards/forensics_tiou_reward": 0.45842401683330536,
      "step": 593
    },
    {
      "completion_length": 20.28125,
      "epoch": 2.485355648535565,
      "grad_norm": 6.4004716873168945,
      "kl": 0.3291015625,
      "learning_rate": 3.797071129707113e-07,
      "loss": 0.0479,
      "reward": 0.7602353692054749,
      "reward_std": 0.03741297638043761,
      "reward_std_all": 0.041972795501351357,
      "rewards-max/forensics_sinkhorn_reward": 0.9869579076766968,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9928353428840637,
      "rewards/forensics_sinkhorn_reward": 0.7329484820365906,
      "rewards/forensics_strict_edge_reward": 0.0490613617002964,
      "rewards/forensics_tiou_reward": 0.47900815308094025,
      "step": 594
    },
    {
      "completion_length": 22.234375,
      "epoch": 2.489539748953975,
      "grad_norm": 4.884827613830566,
      "kl": 0.3642578125,
      "learning_rate": 3.7866108786610877e-07,
      "loss": 0.0797,
      "reward": 0.7529641091823578,
      "reward_std": 0.06794680655002594,
      "reward_std_all": 0.06449558772146702,
      "rewards-max/forensics_sinkhorn_reward": 0.9735345542430878,
      "rewards-max/forensics_strict_edge_reward": 0.6711390465497971,
      "rewards-max/forensics_tiou_reward": 0.9558545053005219,
      "rewards/forensics_sinkhorn_reward": 0.7285886108875275,
      "rewards/forensics_strict_edge_reward": 0.059329524636268616,
      "rewards/forensics_tiou_reward": 0.5830845832824707,
      "step": 595
    },
    {
      "completion_length": 19.375,
      "epoch": 2.493723849372385,
      "grad_norm": 6.219348430633545,
      "kl": 0.416015625,
      "learning_rate": 3.776150627615063e-07,
      "loss": 0.0753,
      "reward": 0.7577251493930817,
      "reward_std": 0.05287950485944748,
      "reward_std_all": 0.05632000416517258,
      "rewards-max/forensics_sinkhorn_reward": 0.9805143475532532,
      "rewards-max/forensics_strict_edge_reward": 0.3657475709915161,
      "rewards-max/forensics_tiou_reward": 0.9725654423236847,
      "rewards/forensics_sinkhorn_reward": 0.7815310657024384,
      "rewards/forensics_strict_edge_reward": 0.07098218984901905,
      "rewards/forensics_tiou_reward": 0.5831527858972549,
      "step": 596
    },
    {
      "completion_length": 19.171875,
      "epoch": 2.497907949790795,
      "grad_norm": 6.93222188949585,
      "kl": 0.291015625,
      "learning_rate": 3.7656903765690374e-07,
      "loss": -0.0877,
      "reward": 0.7888315618038177,
      "reward_std": 0.029536041663959622,
      "reward_std_all": 0.027808955870568752,
      "rewards-max/forensics_sinkhorn_reward": 0.9789204299449921,
      "rewards-max/forensics_strict_edge_reward": 0.6448149085044861,
      "rewards-max/forensics_tiou_reward": 0.9711501896381378,
      "rewards/forensics_sinkhorn_reward": 0.7959207594394684,
      "rewards/forensics_strict_edge_reward": 0.07223663479089737,
      "rewards/forensics_tiou_reward": 0.5932707786560059,
      "step": 597
    },
    {
      "completion_length": 20.453125,
      "epoch": 2.502092050209205,
      "grad_norm": 6.073071479797363,
      "kl": 0.4296875,
      "learning_rate": 3.7552301255230125e-07,
      "loss": 0.0642,
      "reward": 0.7532937824726105,
      "reward_std": 0.042630051262676716,
      "reward_std_all": 0.0501436423510313,
      "rewards-max/forensics_sinkhorn_reward": 0.9846788644790649,
      "rewards-max/forensics_strict_edge_reward": 0.8228278160095215,
      "rewards-max/forensics_tiou_reward": 0.986944317817688,
      "rewards/forensics_sinkhorn_reward": 0.7034916579723358,
      "rewards/forensics_strict_edge_reward": 0.05331154726445675,
      "rewards/forensics_tiou_reward": 0.5011675655841827,
      "step": 598
    },
    {
      "completion_length": 20.359375,
      "epoch": 2.506276150627615,
      "grad_norm": 5.998929023742676,
      "kl": 0.2861328125,
      "learning_rate": 3.744769874476987e-07,
      "loss": 0.0925,
      "reward": 0.7544230818748474,
      "reward_std": 0.028579693287611008,
      "reward_std_all": 0.04102796129882336,
      "rewards-max/forensics_sinkhorn_reward": 0.9825544655323029,
      "rewards-max/forensics_strict_edge_reward": 0.31112344563007355,
      "rewards-max/forensics_tiou_reward": 0.9798858165740967,
      "rewards/forensics_sinkhorn_reward": 0.7042470276355743,
      "rewards/forensics_strict_edge_reward": 0.03442957531660795,
      "rewards/forensics_tiou_reward": 0.5063752979040146,
      "step": 599
    },
    {
      "completion_length": 20.5,
      "epoch": 2.510460251046025,
      "grad_norm": 6.889093399047852,
      "kl": 0.3291015625,
      "learning_rate": 3.734309623430962e-07,
      "loss": 0.0243,
      "reward": 0.734873503446579,
      "reward_std": 0.06828301027417183,
      "reward_std_all": 0.07301788777112961,
      "rewards-max/forensics_sinkhorn_reward": 0.9863193333148956,
      "rewards-max/forensics_strict_edge_reward": 0.904837429523468,
      "rewards-max/forensics_tiou_reward": 0.9932402968406677,
      "rewards/forensics_sinkhorn_reward": 0.7312785387039185,
      "rewards/forensics_strict_edge_reward": 0.08174520544707775,
      "rewards/forensics_tiou_reward": 0.5536405444145203,
      "step": 600
    },
    {
      "completion_length": 21.46875,
      "epoch": 2.514644351464435,
      "grad_norm": 5.599733352661133,
      "kl": 0.390625,
      "learning_rate": 3.723849372384937e-07,
      "loss": -0.0083,
      "reward": 0.7446297109127045,
      "reward_std": 0.02324056252837181,
      "reward_std_all": 0.044426592998206615,
      "rewards-max/forensics_sinkhorn_reward": 0.9772909283638,
      "rewards-max/forensics_strict_edge_reward": 0.3188343495130539,
      "rewards-max/forensics_tiou_reward": 0.9658633470535278,
      "rewards/forensics_sinkhorn_reward": 0.7034904360771179,
      "rewards/forensics_strict_edge_reward": 0.024813862517476082,
      "rewards/forensics_tiou_reward": 0.4728930592536926,
      "step": 601
    },
    {
      "completion_length": 21.078125,
      "epoch": 2.518828451882845,
      "grad_norm": 5.732152938842773,
      "kl": 0.2802734375,
      "learning_rate": 3.713389121338912e-07,
      "loss": -0.0124,
      "reward": 0.7644478678703308,
      "reward_std": 0.046728821471333504,
      "reward_std_all": 0.04589558765292168,
      "rewards-max/forensics_sinkhorn_reward": 0.9850244224071503,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9891660809516907,
      "rewards/forensics_sinkhorn_reward": 0.7859160304069519,
      "rewards/forensics_strict_edge_reward": 0.09256884455680847,
      "rewards/forensics_tiou_reward": 0.6223388910293579,
      "step": 602
    },
    {
      "completion_length": 20.84375,
      "epoch": 2.523012552301255,
      "grad_norm": 5.893537998199463,
      "kl": 0.3564453125,
      "learning_rate": 3.7029288702928866e-07,
      "loss": 0.1018,
      "reward": 0.7161239981651306,
      "reward_std": 0.04405519738793373,
      "reward_std_all": 0.0701056458055973,
      "rewards-max/forensics_sinkhorn_reward": 0.947257399559021,
      "rewards-max/forensics_strict_edge_reward": 0.1253628209233284,
      "rewards-max/forensics_tiou_reward": 0.8996148109436035,
      "rewards/forensics_sinkhorn_reward": 0.6591629981994629,
      "rewards/forensics_strict_edge_reward": 0.012144466862082481,
      "rewards/forensics_tiou_reward": 0.4245954602956772,
      "step": 603
    },
    {
      "completion_length": 23.484375,
      "epoch": 2.5271966527196654,
      "grad_norm": 6.595123291015625,
      "kl": 0.5009765625,
      "learning_rate": 3.6924686192468617e-07,
      "loss": -0.0189,
      "reward": 0.7120737433433533,
      "reward_std": 0.04652221128344536,
      "reward_std_all": 0.06940098851919174,
      "rewards-max/forensics_sinkhorn_reward": 0.9751812815666199,
      "rewards-max/forensics_strict_edge_reward": 0.4495786130428314,
      "rewards-max/forensics_tiou_reward": 0.9606951773166656,
      "rewards/forensics_sinkhorn_reward": 0.7026991844177246,
      "rewards/forensics_strict_edge_reward": 0.04318326711654663,
      "rewards/forensics_tiou_reward": 0.5300911962985992,
      "step": 604
    },
    {
      "completion_length": 20.328125,
      "epoch": 2.5313807531380754,
      "grad_norm": 6.611294746398926,
      "kl": 0.4951171875,
      "learning_rate": 3.682008368200837e-07,
      "loss": 0.024,
      "reward": 0.7485214471817017,
      "reward_std": 0.023612720426172018,
      "reward_std_all": 0.03953682631254196,
      "rewards-max/forensics_sinkhorn_reward": 0.9761500060558319,
      "rewards-max/forensics_strict_edge_reward": 0.13239384430926293,
      "rewards-max/forensics_tiou_reward": 0.9545296132564545,
      "rewards/forensics_sinkhorn_reward": 0.7144781351089478,
      "rewards/forensics_strict_edge_reward": 0.006461868018959649,
      "rewards/forensics_tiou_reward": 0.4603392034769058,
      "step": 605
    },
    {
      "completion_length": 21.671875,
      "epoch": 2.5355648535564854,
      "grad_norm": 6.655762672424316,
      "kl": 0.3955078125,
      "learning_rate": 3.6715481171548114e-07,
      "loss": 0.0496,
      "reward": 0.7896873652935028,
      "reward_std": 0.023685557767748833,
      "reward_std_all": 0.02317730523645878,
      "rewards-max/forensics_sinkhorn_reward": 0.975648432970047,
      "rewards-max/forensics_strict_edge_reward": 0.6052497625350952,
      "rewards-max/forensics_tiou_reward": 0.9664154350757599,
      "rewards/forensics_sinkhorn_reward": 0.7514499127864838,
      "rewards/forensics_strict_edge_reward": 0.04919355548918247,
      "rewards/forensics_tiou_reward": 0.5121455043554306,
      "step": 606
    },
    {
      "completion_length": 21.828125,
      "epoch": 2.5397489539748954,
      "grad_norm": 4.183310031890869,
      "kl": 0.34375,
      "learning_rate": 3.661087866108786e-07,
      "loss": 0.0962,
      "reward": 0.7653594017028809,
      "reward_std": 0.03960139770060778,
      "reward_std_all": 0.04265381023287773,
      "rewards-max/forensics_sinkhorn_reward": 0.9707864820957184,
      "rewards-max/forensics_strict_edge_reward": 0.4430994242429733,
      "rewards-max/forensics_tiou_reward": 0.9468111991882324,
      "rewards/forensics_sinkhorn_reward": 0.7187924981117249,
      "rewards/forensics_strict_edge_reward": 0.04165233299136162,
      "rewards/forensics_tiou_reward": 0.5358521640300751,
      "step": 607
    },
    {
      "completion_length": 20.75,
      "epoch": 2.5439330543933054,
      "grad_norm": 5.472422122955322,
      "kl": 0.3115234375,
      "learning_rate": 3.650627615062761e-07,
      "loss": -0.0262,
      "reward": 0.7829910814762115,
      "reward_std": 0.025582627567928284,
      "reward_std_all": 0.026377359870821238,
      "rewards-max/forensics_sinkhorn_reward": 0.9776717722415924,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9706521034240723,
      "rewards/forensics_sinkhorn_reward": 0.7723467648029327,
      "rewards/forensics_strict_edge_reward": 0.044389981776475906,
      "rewards/forensics_tiou_reward": 0.5759753882884979,
      "step": 608
    },
    {
      "completion_length": 19.546875,
      "epoch": 2.5481171548117154,
      "grad_norm": 6.082596778869629,
      "kl": 0.3818359375,
      "learning_rate": 3.6401673640167363e-07,
      "loss": 0.0592,
      "reward": 0.7515498995780945,
      "reward_std": 0.05032403580844402,
      "reward_std_all": 0.05540760047733784,
      "rewards-max/forensics_sinkhorn_reward": 0.9751176536083221,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9663901329040527,
      "rewards/forensics_sinkhorn_reward": 0.7304194569587708,
      "rewards/forensics_strict_edge_reward": 0.05163956619799137,
      "rewards/forensics_tiou_reward": 0.5080266892910004,
      "step": 609
    },
    {
      "completion_length": 21.90625,
      "epoch": 2.5523012552301254,
      "grad_norm": 5.74039363861084,
      "kl": 0.3310546875,
      "learning_rate": 3.6297071129707114e-07,
      "loss": 0.0035,
      "reward": 0.7840551733970642,
      "reward_std": 0.021767455153167248,
      "reward_std_all": 0.02309324173256755,
      "rewards-max/forensics_sinkhorn_reward": 0.9646945595741272,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9416705965995789,
      "rewards/forensics_sinkhorn_reward": 0.7487689554691315,
      "rewards/forensics_strict_edge_reward": 0.03944344399496913,
      "rewards/forensics_tiou_reward": 0.5393404066562653,
      "step": 610
    },
    {
      "completion_length": 20.640625,
      "epoch": 2.5564853556485354,
      "grad_norm": 7.265350341796875,
      "kl": 0.4140625,
      "learning_rate": 3.6192468619246855e-07,
      "loss": -0.0341,
      "reward": 0.7449750304222107,
      "reward_std": 0.08711748942732811,
      "reward_std_all": 0.07991443760693073,
      "rewards-max/forensics_sinkhorn_reward": 0.9820851683616638,
      "rewards-max/forensics_strict_edge_reward": 0.6770831942558289,
      "rewards-max/forensics_tiou_reward": 0.9801069498062134,
      "rewards/forensics_sinkhorn_reward": 0.7115130126476288,
      "rewards/forensics_strict_edge_reward": 0.06607694551348686,
      "rewards/forensics_tiou_reward": 0.534901887178421,
      "step": 611
    },
    {
      "completion_length": 21.640625,
      "epoch": 2.560669456066946,
      "grad_norm": 7.3426618576049805,
      "kl": 0.3427734375,
      "learning_rate": 3.6087866108786606e-07,
      "loss": -0.0352,
      "reward": 0.7770026624202728,
      "reward_std": 0.025346746551804245,
      "reward_std_all": 0.029991278424859047,
      "rewards-max/forensics_sinkhorn_reward": 0.9664867520332336,
      "rewards-max/forensics_strict_edge_reward": 0.46539999544620514,
      "rewards-max/forensics_tiou_reward": 0.9443540871143341,
      "rewards/forensics_sinkhorn_reward": 0.7327403128147125,
      "rewards/forensics_strict_edge_reward": 0.045957300812006,
      "rewards/forensics_tiou_reward": 0.48430152237415314,
      "step": 612
    },
    {
      "completion_length": 23.03125,
      "epoch": 2.564853556485356,
      "grad_norm": 5.41560173034668,
      "kl": 0.390625,
      "learning_rate": 3.598326359832636e-07,
      "loss": 0.1034,
      "reward": 0.7833127677440643,
      "reward_std": 0.02430588286370039,
      "reward_std_all": 0.02653861977159977,
      "rewards-max/forensics_sinkhorn_reward": 0.9777522087097168,
      "rewards-max/forensics_strict_edge_reward": 0.2088332585990429,
      "rewards-max/forensics_tiou_reward": 0.9625949859619141,
      "rewards/forensics_sinkhorn_reward": 0.7213671207427979,
      "rewards/forensics_strict_edge_reward": 0.01765295211225748,
      "rewards/forensics_tiou_reward": 0.4852035492658615,
      "step": 613
    },
    {
      "completion_length": 22.609375,
      "epoch": 2.569037656903766,
      "grad_norm": 6.430539608001709,
      "kl": 0.400390625,
      "learning_rate": 3.587866108786611e-07,
      "loss": 0.0182,
      "reward": 0.7494467198848724,
      "reward_std": 0.04062812216579914,
      "reward_std_all": 0.050179196521639824,
      "rewards-max/forensics_sinkhorn_reward": 0.957577109336853,
      "rewards-max/forensics_strict_edge_reward": 0.5608851462602615,
      "rewards-max/forensics_tiou_reward": 0.9205210506916046,
      "rewards/forensics_sinkhorn_reward": 0.6591280400753021,
      "rewards/forensics_strict_edge_reward": 0.05880642868578434,
      "rewards/forensics_tiou_reward": 0.4628358483314514,
      "step": 614
    },
    {
      "completion_length": 20.875,
      "epoch": 2.573221757322176,
      "grad_norm": 5.971790790557861,
      "kl": 0.556640625,
      "learning_rate": 3.577405857740586e-07,
      "loss": -0.036,
      "reward": 0.7504089772701263,
      "reward_std": 0.05827227886766195,
      "reward_std_all": 0.05734612978994846,
      "rewards-max/forensics_sinkhorn_reward": 0.9766650199890137,
      "rewards-max/forensics_strict_edge_reward": 0.5483185648918152,
      "rewards-max/forensics_tiou_reward": 0.9704303741455078,
      "rewards/forensics_sinkhorn_reward": 0.7502927482128143,
      "rewards/forensics_strict_edge_reward": 0.04834608267992735,
      "rewards/forensics_tiou_reward": 0.5449743866920471,
      "step": 615
    },
    {
      "completion_length": 22.203125,
      "epoch": 2.577405857740586,
      "grad_norm": 9.109413146972656,
      "kl": 0.33935546875,
      "learning_rate": 3.56694560669456e-07,
      "loss": 0.0867,
      "reward": 0.7333050668239594,
      "reward_std": 0.06711319461464882,
      "reward_std_all": 0.06790285184979439,
      "rewards-max/forensics_sinkhorn_reward": 0.9717864096164703,
      "rewards-max/forensics_strict_edge_reward": 0.4651319086551666,
      "rewards-max/forensics_tiou_reward": 0.9536147713661194,
      "rewards/forensics_sinkhorn_reward": 0.6630479395389557,
      "rewards/forensics_strict_edge_reward": 0.044655690900981426,
      "rewards/forensics_tiou_reward": 0.45806871354579926,
      "step": 616
    },
    {
      "completion_length": 20.78125,
      "epoch": 2.581589958158996,
      "grad_norm": 5.7271952629089355,
      "kl": 0.345703125,
      "learning_rate": 3.556485355648535e-07,
      "loss": 0.0289,
      "reward": 0.7657254040241241,
      "reward_std": 0.035600846633315086,
      "reward_std_all": 0.041889509186148643,
      "rewards-max/forensics_sinkhorn_reward": 0.9545750021934509,
      "rewards-max/forensics_strict_edge_reward": 0.0452746357768774,
      "rewards-max/forensics_tiou_reward": 0.8888571858406067,
      "rewards/forensics_sinkhorn_reward": 0.7407250702381134,
      "rewards/forensics_strict_edge_reward": 0.0024678835179656744,
      "rewards/forensics_tiou_reward": 0.5269922912120819,
      "step": 617
    },
    {
      "completion_length": 20.53125,
      "epoch": 2.585774058577406,
      "grad_norm": 7.050764083862305,
      "kl": 0.380859375,
      "learning_rate": 3.5460251046025103e-07,
      "loss": 0.0502,
      "reward": 0.7366276383399963,
      "reward_std": 0.08235440403223038,
      "reward_std_all": 0.07897720858454704,
      "rewards-max/forensics_sinkhorn_reward": 0.9564995169639587,
      "rewards-max/forensics_strict_edge_reward": 0.25757846236228943,
      "rewards-max/forensics_tiou_reward": 0.9089570343494415,
      "rewards/forensics_sinkhorn_reward": 0.6645708978176117,
      "rewards/forensics_strict_edge_reward": 0.04218104691244662,
      "rewards/forensics_tiou_reward": 0.43685439229011536,
      "step": 618
    },
    {
      "completion_length": 20.21875,
      "epoch": 2.589958158995816,
      "grad_norm": 5.605987071990967,
      "kl": 0.314453125,
      "learning_rate": 3.5355648535564855e-07,
      "loss": -0.0388,
      "reward": 0.8038383424282074,
      "reward_std": 0.0034659639932215214,
      "reward_std_all": 0.005213505821302533,
      "rewards-max/forensics_sinkhorn_reward": 0.9563886225223541,
      "rewards-max/forensics_strict_edge_reward": 0.44148343801498413,
      "rewards-max/forensics_tiou_reward": 0.9192770719528198,
      "rewards/forensics_sinkhorn_reward": 0.8022659718990326,
      "rewards/forensics_strict_edge_reward": 0.06803877046331763,
      "rewards/forensics_tiou_reward": 0.6077511608600616,
      "step": 619
    },
    {
      "completion_length": 21.46875,
      "epoch": 2.594142259414226,
      "grad_norm": 6.466946601867676,
      "kl": 0.4033203125,
      "learning_rate": 3.52510460251046e-07,
      "loss": 0.0159,
      "reward": 0.7858346998691559,
      "reward_std": 0.026494039222598076,
      "reward_std_all": 0.02584242820739746,
      "rewards-max/forensics_sinkhorn_reward": 0.9718104898929596,
      "rewards-max/forensics_strict_edge_reward": 0.43846702575683594,
      "rewards-max/forensics_tiou_reward": 0.9541296064853668,
      "rewards/forensics_sinkhorn_reward": 0.7847190499305725,
      "rewards/forensics_strict_edge_reward": 0.0408161785453558,
      "rewards/forensics_tiou_reward": 0.6228030920028687,
      "step": 620
    },
    {
      "completion_length": 24.15625,
      "epoch": 2.598326359832636,
      "grad_norm": 4.990799427032471,
      "kl": 0.4404296875,
      "learning_rate": 3.5146443514644347e-07,
      "loss": 0.0265,
      "reward": 0.7488555312156677,
      "reward_std": 0.044690662529319525,
      "reward_std_all": 0.04880718211643398,
      "rewards-max/forensics_sinkhorn_reward": 0.9802053272724152,
      "rewards-max/forensics_strict_edge_reward": 0.6360974311828613,
      "rewards-max/forensics_tiou_reward": 0.9774231612682343,
      "rewards/forensics_sinkhorn_reward": 0.7224697172641754,
      "rewards/forensics_strict_edge_reward": 0.09725645929574966,
      "rewards/forensics_tiou_reward": 0.558881089091301,
      "step": 621
    },
    {
      "completion_length": 21.390625,
      "epoch": 2.602510460251046,
      "grad_norm": 6.058696269989014,
      "kl": 0.3955078125,
      "learning_rate": 3.50418410041841e-07,
      "loss": -0.0633,
      "reward": 0.7581726610660553,
      "reward_std": 0.03439293056726456,
      "reward_std_all": 0.0405244929715991,
      "rewards-max/forensics_sinkhorn_reward": 0.9787078201770782,
      "rewards-max/forensics_strict_edge_reward": 0.6026886105537415,
      "rewards-max/forensics_tiou_reward": 0.9740451574325562,
      "rewards/forensics_sinkhorn_reward": 0.737664133310318,
      "rewards/forensics_strict_edge_reward": 0.06633642222732306,
      "rewards/forensics_tiou_reward": 0.565353274345398,
      "step": 622
    },
    {
      "completion_length": 21.53125,
      "epoch": 2.606694560669456,
      "grad_norm": 6.758168697357178,
      "kl": 0.4130859375,
      "learning_rate": 3.493723849372385e-07,
      "loss": 0.0873,
      "reward": 0.7726192474365234,
      "reward_std": 0.03401458263397217,
      "reward_std_all": 0.03502768278121948,
      "rewards-max/forensics_sinkhorn_reward": 0.945947527885437,
      "rewards-max/forensics_strict_edge_reward": 0.24270229786634445,
      "rewards-max/forensics_tiou_reward": 0.867527186870575,
      "rewards/forensics_sinkhorn_reward": 0.7005541026592255,
      "rewards/forensics_strict_edge_reward": 0.019009833689779043,
      "rewards/forensics_tiou_reward": 0.4307048171758652,
      "step": 623
    },
    {
      "completion_length": 22.296875,
      "epoch": 2.6108786610878663,
      "grad_norm": 6.39846134185791,
      "kl": 0.416015625,
      "learning_rate": 3.4832635983263595e-07,
      "loss": -0.0608,
      "reward": 0.7690555453300476,
      "reward_std": 0.04826617054641247,
      "reward_std_all": 0.04613986145704985,
      "rewards-max/forensics_sinkhorn_reward": 0.9348273873329163,
      "rewards-max/forensics_strict_edge_reward": 0.25777730345726013,
      "rewards-max/forensics_tiou_reward": 0.8664689064025879,
      "rewards/forensics_sinkhorn_reward": 0.7279624938964844,
      "rewards/forensics_strict_edge_reward": 0.025408015586435795,
      "rewards/forensics_tiou_reward": 0.5079587697982788,
      "step": 624
    },
    {
      "completion_length": 20.734375,
      "epoch": 2.6150627615062763,
      "grad_norm": 6.262167930603027,
      "kl": 0.4296875,
      "learning_rate": 3.4728033472803346e-07,
      "loss": -0.0884,
      "reward": 0.7465334236621857,
      "reward_std": 0.06589158065617085,
      "reward_std_all": 0.06404509022831917,
      "rewards-max/forensics_sinkhorn_reward": 0.9682027697563171,
      "rewards-max/forensics_strict_edge_reward": 0.5842172801494598,
      "rewards-max/forensics_tiou_reward": 0.9498212039470673,
      "rewards/forensics_sinkhorn_reward": 0.7143082916736603,
      "rewards/forensics_strict_edge_reward": 0.061160001903772354,
      "rewards/forensics_tiou_reward": 0.49488696455955505,
      "step": 625
    },
    {
      "completion_length": 21.828125,
      "epoch": 2.6192468619246863,
      "grad_norm": 6.776823043823242,
      "kl": 0.34765625,
      "learning_rate": 3.462343096234309e-07,
      "loss": 0.0122,
      "reward": 0.6886842250823975,
      "reward_std": 0.04369048960506916,
      "reward_std_all": 0.07929184846580029,
      "rewards-max/forensics_sinkhorn_reward": 0.9503701329231262,
      "rewards-max/forensics_strict_edge_reward": 0.1573963314294815,
      "rewards-max/forensics_tiou_reward": 0.8921264708042145,
      "rewards/forensics_sinkhorn_reward": 0.628728061914444,
      "rewards/forensics_strict_edge_reward": 0.010612123645842075,
      "rewards/forensics_tiou_reward": 0.4062145948410034,
      "step": 626
    },
    {
      "completion_length": 23.171875,
      "epoch": 2.6234309623430963,
      "grad_norm": 6.018642425537109,
      "kl": 0.32421875,
      "learning_rate": 3.4518828451882844e-07,
      "loss": 0.1585,
      "reward": 0.7043982148170471,
      "reward_std": 0.08680610358715057,
      "reward_std_all": 0.09239542484283447,
      "rewards-max/forensics_sinkhorn_reward": 0.9294307231903076,
      "rewards-max/forensics_strict_edge_reward": 0.42542512505315244,
      "rewards-max/forensics_tiou_reward": 0.852316826581955,
      "rewards/forensics_sinkhorn_reward": 0.6572585701942444,
      "rewards/forensics_strict_edge_reward": 0.028390806837705895,
      "rewards/forensics_tiou_reward": 0.39647363126277924,
      "step": 627
    },
    {
      "completion_length": 22.390625,
      "epoch": 2.6276150627615062,
      "grad_norm": 10.188720703125,
      "kl": 0.61328125,
      "learning_rate": 3.441422594142259e-07,
      "loss": 0.1069,
      "reward": 0.7408422529697418,
      "reward_std": 0.06338995136320591,
      "reward_std_all": 0.06866609305143356,
      "rewards-max/forensics_sinkhorn_reward": 0.9828186929225922,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9810445606708527,
      "rewards/forensics_sinkhorn_reward": 0.7333829700946808,
      "rewards/forensics_strict_edge_reward": 0.06609240174293518,
      "rewards/forensics_tiou_reward": 0.549629420042038,
      "step": 628
    },
    {
      "completion_length": 22.484375,
      "epoch": 2.6317991631799162,
      "grad_norm": 4.70028829574585,
      "kl": 0.375,
      "learning_rate": 3.430962343096234e-07,
      "loss": 0.1688,
      "reward": 0.7383419871330261,
      "reward_std": 0.07772603631019592,
      "reward_std_all": 0.07738963514566422,
      "rewards-max/forensics_sinkhorn_reward": 0.9621945321559906,
      "rewards-max/forensics_strict_edge_reward": 0.299948550760746,
      "rewards-max/forensics_tiou_reward": 0.9353647828102112,
      "rewards/forensics_sinkhorn_reward": 0.6964085102081299,
      "rewards/forensics_strict_edge_reward": 0.0226389872841537,
      "rewards/forensics_tiou_reward": 0.5171794295310974,
      "step": 629
    },
    {
      "completion_length": 20.25,
      "epoch": 2.6359832635983262,
      "grad_norm": 9.43249797821045,
      "kl": 0.486328125,
      "learning_rate": 3.420502092050209e-07,
      "loss": 0.0446,
      "reward": 0.7220035493373871,
      "reward_std": 0.05309552326798439,
      "reward_std_all": 0.0723166223615408,
      "rewards-max/forensics_sinkhorn_reward": 0.9758312404155731,
      "rewards-max/forensics_strict_edge_reward": 0.3975932002067566,
      "rewards-max/forensics_tiou_reward": 0.9604610502719879,
      "rewards/forensics_sinkhorn_reward": 0.69642773270607,
      "rewards/forensics_strict_edge_reward": 0.061883650720119476,
      "rewards/forensics_tiou_reward": 0.50200155377388,
      "step": 630
    },
    {
      "completion_length": 20.015625,
      "epoch": 2.6401673640167362,
      "grad_norm": 6.103680610656738,
      "kl": 0.3798828125,
      "learning_rate": 3.410041841004184e-07,
      "loss": -0.0478,
      "reward": 0.7817833721637726,
      "reward_std": 0.033054836094379425,
      "reward_std_all": 0.03198780119419098,
      "rewards-max/forensics_sinkhorn_reward": 0.9528439342975616,
      "rewards-max/forensics_strict_edge_reward": 0.19178982637822628,
      "rewards-max/forensics_tiou_reward": 0.9176374077796936,
      "rewards/forensics_sinkhorn_reward": 0.7474197745323181,
      "rewards/forensics_strict_edge_reward": 0.014400660991668701,
      "rewards/forensics_tiou_reward": 0.5182994455099106,
      "step": 631
    },
    {
      "completion_length": 23.390625,
      "epoch": 2.6443514644351467,
      "grad_norm": 6.24610710144043,
      "kl": 0.29833984375,
      "learning_rate": 3.3995815899581584e-07,
      "loss": 0.1068,
      "reward": 0.7538719475269318,
      "reward_std": 0.06156355142593384,
      "reward_std_all": 0.05993320420384407,
      "rewards-max/forensics_sinkhorn_reward": 0.9872146844863892,
      "rewards-max/forensics_strict_edge_reward": 0.5724648237228394,
      "rewards-max/forensics_tiou_reward": 0.9944153726100922,
      "rewards/forensics_sinkhorn_reward": 0.7140731513500214,
      "rewards/forensics_strict_edge_reward": 0.052898820489645004,
      "rewards/forensics_tiou_reward": 0.5258273482322693,
      "step": 632
    },
    {
      "completion_length": 22.015625,
      "epoch": 2.6485355648535567,
      "grad_norm": 6.415136337280273,
      "kl": 0.396484375,
      "learning_rate": 3.3891213389121336e-07,
      "loss": -0.0158,
      "reward": 0.6969576776027679,
      "reward_std": 0.08603718504309654,
      "reward_std_all": 0.09535450488328934,
      "rewards-max/forensics_sinkhorn_reward": 0.9697538316249847,
      "rewards-max/forensics_strict_edge_reward": 0.5770299434661865,
      "rewards-max/forensics_tiou_reward": 0.9542205333709717,
      "rewards/forensics_sinkhorn_reward": 0.6645958125591278,
      "rewards/forensics_strict_edge_reward": 0.04808631166815758,
      "rewards/forensics_tiou_reward": 0.4966288059949875,
      "step": 633
    },
    {
      "completion_length": 22.734375,
      "epoch": 2.6527196652719667,
      "grad_norm": 5.368725299835205,
      "kl": 0.3896484375,
      "learning_rate": 3.3786610878661087e-07,
      "loss": -0.0099,
      "reward": 0.7588444352149963,
      "reward_std": 0.04432301037013531,
      "reward_std_all": 0.04606596287339926,
      "rewards-max/forensics_sinkhorn_reward": 0.9809917509555817,
      "rewards-max/forensics_strict_edge_reward": 0.3065662086009979,
      "rewards-max/forensics_tiou_reward": 0.9745804071426392,
      "rewards/forensics_sinkhorn_reward": 0.6832091510295868,
      "rewards/forensics_strict_edge_reward": 0.04123690817505121,
      "rewards/forensics_tiou_reward": 0.5226039886474609,
      "step": 634
    },
    {
      "completion_length": 21.671875,
      "epoch": 2.6569037656903767,
      "grad_norm": 6.985921859741211,
      "kl": 0.51953125,
      "learning_rate": 3.368200836820084e-07,
      "loss": 0.0832,
      "reward": 0.7462861835956573,
      "reward_std": 0.07396403886377811,
      "reward_std_all": 0.06949920020997524,
      "rewards-max/forensics_sinkhorn_reward": 0.9639788866043091,
      "rewards-max/forensics_strict_edge_reward": 0.4293667823076248,
      "rewards-max/forensics_tiou_reward": 0.9237571358680725,
      "rewards/forensics_sinkhorn_reward": 0.6816993951797485,
      "rewards/forensics_strict_edge_reward": 0.04085736349225044,
      "rewards/forensics_tiou_reward": 0.46811673045158386,
      "step": 635
    },
    {
      "completion_length": 21.421875,
      "epoch": 2.6610878661087867,
      "grad_norm": 5.53606653213501,
      "kl": 0.26171875,
      "learning_rate": 3.3577405857740584e-07,
      "loss": -0.0042,
      "reward": 0.7428376972675323,
      "reward_std": 0.038342325016856194,
      "reward_std_all": 0.05037446226924658,
      "rewards-max/forensics_sinkhorn_reward": 0.9823938608169556,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9831542670726776,
      "rewards/forensics_sinkhorn_reward": 0.7307012975215912,
      "rewards/forensics_strict_edge_reward": 0.05525635555386543,
      "rewards/forensics_tiou_reward": 0.5535272359848022,
      "step": 636
    },
    {
      "completion_length": 22.390625,
      "epoch": 2.6652719665271967,
      "grad_norm": 6.218846797943115,
      "kl": 0.337890625,
      "learning_rate": 3.347280334728033e-07,
      "loss": 0.0691,
      "reward": 0.7551800906658173,
      "reward_std": 0.060945503413677216,
      "reward_std_all": 0.05803024210035801,
      "rewards-max/forensics_sinkhorn_reward": 0.975915789604187,
      "rewards-max/forensics_strict_edge_reward": 0.5879503488540649,
      "rewards-max/forensics_tiou_reward": 0.968299150466919,
      "rewards/forensics_sinkhorn_reward": 0.7310789823532104,
      "rewards/forensics_strict_edge_reward": 0.04486913979053497,
      "rewards/forensics_tiou_reward": 0.5636971890926361,
      "step": 637
    },
    {
      "completion_length": 21.453125,
      "epoch": 2.6694560669456067,
      "grad_norm": 6.086441516876221,
      "kl": 0.3876953125,
      "learning_rate": 3.336820083682008e-07,
      "loss": 0.0133,
      "reward": 0.7151680588722229,
      "reward_std": 0.047151776030659676,
      "reward_std_all": 0.06633156538009644,
      "rewards-max/forensics_sinkhorn_reward": 0.9531884789466858,
      "rewards-max/forensics_strict_edge_reward": 0.2604241222143173,
      "rewards-max/forensics_tiou_reward": 0.90034019947052,
      "rewards/forensics_sinkhorn_reward": 0.6692900955677032,
      "rewards/forensics_strict_edge_reward": 0.0283057177439332,
      "rewards/forensics_tiou_reward": 0.433869868516922,
      "step": 638
    },
    {
      "completion_length": 20.96875,
      "epoch": 2.6736401673640167,
      "grad_norm": 6.041973114013672,
      "kl": 0.427734375,
      "learning_rate": 3.3263598326359833e-07,
      "loss": 0.0302,
      "reward": 0.7771761119365692,
      "reward_std": 0.03513783309608698,
      "reward_std_all": 0.03464546054601669,
      "rewards-max/forensics_sinkhorn_reward": 0.9625832736492157,
      "rewards-max/forensics_strict_edge_reward": 0.1602426841855049,
      "rewards-max/forensics_tiou_reward": 0.9287056028842926,
      "rewards/forensics_sinkhorn_reward": 0.7646926045417786,
      "rewards/forensics_strict_edge_reward": 0.010485471226274967,
      "rewards/forensics_tiou_reward": 0.5320554971694946,
      "step": 639
    },
    {
      "completion_length": 27.046875,
      "epoch": 2.6778242677824267,
      "grad_norm": 9.011261940002441,
      "kl": 0.4921875,
      "learning_rate": 3.3158995815899584e-07,
      "loss": -0.0025,
      "reward": 0.6440068483352661,
      "reward_std": 0.07906627282500267,
      "reward_std_all": 0.11727563291788101,
      "rewards-max/forensics_sinkhorn_reward": 0.9577662348747253,
      "rewards-max/forensics_strict_edge_reward": 0.19615432247519493,
      "rewards-max/forensics_tiou_reward": 0.9167105257511139,
      "rewards/forensics_sinkhorn_reward": 0.5685686767101288,
      "rewards/forensics_strict_edge_reward": 0.012094886973500252,
      "rewards/forensics_tiou_reward": 0.39059896767139435,
      "step": 640
    },
    {
      "completion_length": 22.171875,
      "epoch": 2.6820083682008367,
      "grad_norm": 7.613966464996338,
      "kl": 0.67578125,
      "learning_rate": 3.3054393305439325e-07,
      "loss": 0.0494,
      "reward": 0.7516288757324219,
      "reward_std": 0.050639137625694275,
      "reward_std_all": 0.05316955968737602,
      "rewards-max/forensics_sinkhorn_reward": 0.9521686732769012,
      "rewards-max/forensics_strict_edge_reward": 0.19514533132314682,
      "rewards-max/forensics_tiou_reward": 0.9065318405628204,
      "rewards/forensics_sinkhorn_reward": 0.6855869591236115,
      "rewards/forensics_strict_edge_reward": 0.019189128652215004,
      "rewards/forensics_tiou_reward": 0.42659372091293335,
      "step": 641
    },
    {
      "completion_length": 22.25,
      "epoch": 2.6861924686192467,
      "grad_norm": 5.362766265869141,
      "kl": 0.388671875,
      "learning_rate": 3.2949790794979076e-07,
      "loss": 0.0391,
      "reward": 0.7679418623447418,
      "reward_std": 0.05572282709181309,
      "reward_std_all": 0.05146223120391369,
      "rewards-max/forensics_sinkhorn_reward": 0.9690744876861572,
      "rewards-max/forensics_strict_edge_reward": 0.5050012171268463,
      "rewards-max/forensics_tiou_reward": 0.9527385234832764,
      "rewards/forensics_sinkhorn_reward": 0.7236821055412292,
      "rewards/forensics_strict_edge_reward": 0.048959964886307716,
      "rewards/forensics_tiou_reward": 0.523521214723587,
      "step": 642
    },
    {
      "completion_length": 20.328125,
      "epoch": 2.6903765690376567,
      "grad_norm": 5.993111610412598,
      "kl": 0.3095703125,
      "learning_rate": 3.2845188284518827e-07,
      "loss": 0.0495,
      "reward": 0.7364043295383453,
      "reward_std": 0.07561255618929863,
      "reward_std_all": 0.07419615611433983,
      "rewards-max/forensics_sinkhorn_reward": 0.9840478897094727,
      "rewards-max/forensics_strict_edge_reward": 0.7936499714851379,
      "rewards-max/forensics_tiou_reward": 0.9878466725349426,
      "rewards/forensics_sinkhorn_reward": 0.6859976053237915,
      "rewards/forensics_strict_edge_reward": 0.06246219016611576,
      "rewards/forensics_tiou_reward": 0.44142407178878784,
      "step": 643
    },
    {
      "completion_length": 24.234375,
      "epoch": 2.694560669456067,
      "grad_norm": 5.891589641571045,
      "kl": 0.5732421875,
      "learning_rate": 3.274058577405858e-07,
      "loss": 0.1855,
      "reward": 0.7606287598609924,
      "reward_std": 0.04738218989223242,
      "reward_std_all": 0.04703397490084171,
      "rewards-max/forensics_sinkhorn_reward": 0.9827788174152374,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9822231829166412,
      "rewards/forensics_sinkhorn_reward": 0.690250813961029,
      "rewards/forensics_strict_edge_reward": 0.037436915561556816,
      "rewards/forensics_tiou_reward": 0.486132487654686,
      "step": 644
    },
    {
      "completion_length": 20.578125,
      "epoch": 2.698744769874477,
      "grad_norm": 6.6217193603515625,
      "kl": 0.6015625,
      "learning_rate": 3.2635983263598325e-07,
      "loss": 0.0105,
      "reward": 0.7538917362689972,
      "reward_std": 0.02998372260481119,
      "reward_std_all": 0.041852083057165146,
      "rewards-max/forensics_sinkhorn_reward": 0.973576545715332,
      "rewards-max/forensics_strict_edge_reward": 0.18767917901277542,
      "rewards-max/forensics_tiou_reward": 0.9544002115726471,
      "rewards/forensics_sinkhorn_reward": 0.7072923481464386,
      "rewards/forensics_strict_edge_reward": 0.019484647549688816,
      "rewards/forensics_tiou_reward": 0.47809945046901703,
      "step": 645
    },
    {
      "completion_length": 21.6875,
      "epoch": 2.702928870292887,
      "grad_norm": 5.613775253295898,
      "kl": 0.396484375,
      "learning_rate": 3.253138075313807e-07,
      "loss": 0.1009,
      "reward": 0.7716404795646667,
      "reward_std": 0.04954829812049866,
      "reward_std_all": 0.046165239065885544,
      "rewards-max/forensics_sinkhorn_reward": 0.9754977822303772,
      "rewards-max/forensics_strict_edge_reward": 0.49658530950546265,
      "rewards-max/forensics_tiou_reward": 0.9632291793823242,
      "rewards/forensics_sinkhorn_reward": 0.735683411359787,
      "rewards/forensics_strict_edge_reward": 0.03942218795418739,
      "rewards/forensics_tiou_reward": 0.4968293607234955,
      "step": 646
    },
    {
      "completion_length": 21.203125,
      "epoch": 2.707112970711297,
      "grad_norm": 8.264657974243164,
      "kl": 0.3173828125,
      "learning_rate": 3.242677824267782e-07,
      "loss": -0.0141,
      "reward": 0.7514888048171997,
      "reward_std": 0.03309129364788532,
      "reward_std_all": 0.043795399367809296,
      "rewards-max/forensics_sinkhorn_reward": 0.956103503704071,
      "rewards-max/forensics_strict_edge_reward": 0.3916143849492073,
      "rewards-max/forensics_tiou_reward": 0.9208556711673737,
      "rewards/forensics_sinkhorn_reward": 0.7014825642108917,
      "rewards/forensics_strict_edge_reward": 0.025591567158699036,
      "rewards/forensics_tiou_reward": 0.5117515325546265,
      "step": 647
    },
    {
      "completion_length": 21.703125,
      "epoch": 2.711297071129707,
      "grad_norm": 4.911921501159668,
      "kl": 0.392578125,
      "learning_rate": 3.2322175732217573e-07,
      "loss": 0.0383,
      "reward": 0.7798506021499634,
      "reward_std": 0.03009773511439562,
      "reward_std_all": 0.030189432203769684,
      "rewards-max/forensics_sinkhorn_reward": 0.9875590801239014,
      "rewards-max/forensics_strict_edge_reward": 0.6505971103906631,
      "rewards-max/forensics_tiou_reward": 0.9957827627658844,
      "rewards/forensics_sinkhorn_reward": 0.7419974207878113,
      "rewards/forensics_strict_edge_reward": 0.05043949745595455,
      "rewards/forensics_tiou_reward": 0.5448229908943176,
      "step": 648
    },
    {
      "completion_length": 23.46875,
      "epoch": 2.715481171548117,
      "grad_norm": 6.062816619873047,
      "kl": 0.3779296875,
      "learning_rate": 3.221757322175732e-07,
      "loss": 0.0728,
      "reward": 0.7618711590766907,
      "reward_std": 0.06534496881067753,
      "reward_std_all": 0.05974392965435982,
      "rewards-max/forensics_sinkhorn_reward": 0.965182900428772,
      "rewards-max/forensics_strict_edge_reward": 0.3564630448818207,
      "rewards-max/forensics_tiou_reward": 0.9379447400569916,
      "rewards/forensics_sinkhorn_reward": 0.7287681996822357,
      "rewards/forensics_strict_edge_reward": 0.023645764216780663,
      "rewards/forensics_tiou_reward": 0.5245933532714844,
      "step": 649
    },
    {
      "completion_length": 21.0625,
      "epoch": 2.719665271966527,
      "grad_norm": 7.026313304901123,
      "kl": 0.3935546875,
      "learning_rate": 3.211297071129707e-07,
      "loss": 0.0835,
      "reward": 0.7552143931388855,
      "reward_std": 0.04738703928887844,
      "reward_std_all": 0.052239677868783474,
      "rewards-max/forensics_sinkhorn_reward": 0.9632093906402588,
      "rewards-max/forensics_strict_edge_reward": 0.297432079911232,
      "rewards-max/forensics_tiou_reward": 0.9207112193107605,
      "rewards/forensics_sinkhorn_reward": 0.7212975919246674,
      "rewards/forensics_strict_edge_reward": 0.027140326565131545,
      "rewards/forensics_tiou_reward": 0.508402481675148,
      "step": 650
    },
    {
      "completion_length": 21.09375,
      "epoch": 2.723849372384937,
      "grad_norm": 5.378275394439697,
      "kl": 0.4453125,
      "learning_rate": 3.2008368200836816e-07,
      "loss": 0.0401,
      "reward": 0.7806941270828247,
      "reward_std": 0.01999870454892516,
      "reward_std_all": 0.023430393543094397,
      "rewards-max/forensics_sinkhorn_reward": 0.9691861271858215,
      "rewards-max/forensics_strict_edge_reward": 0.5192448049783707,
      "rewards-max/forensics_tiou_reward": 0.9474660754203796,
      "rewards/forensics_sinkhorn_reward": 0.7037495970726013,
      "rewards/forensics_strict_edge_reward": 0.04591632168740034,
      "rewards/forensics_tiou_reward": 0.5256731510162354,
      "step": 651
    },
    {
      "completion_length": 21.765625,
      "epoch": 2.7280334728033475,
      "grad_norm": 7.399298667907715,
      "kl": 0.2900390625,
      "learning_rate": 3.190376569037657e-07,
      "loss": 0.0089,
      "reward": 0.7606344223022461,
      "reward_std": 0.059298474341630936,
      "reward_std_all": 0.05598537251353264,
      "rewards-max/forensics_sinkhorn_reward": 0.9714748859405518,
      "rewards-max/forensics_strict_edge_reward": 0.326669804751873,
      "rewards-max/forensics_tiou_reward": 0.9444057941436768,
      "rewards/forensics_sinkhorn_reward": 0.7420777678489685,
      "rewards/forensics_strict_edge_reward": 0.03311212779954076,
      "rewards/forensics_tiou_reward": 0.5249548256397247,
      "step": 652
    },
    {
      "completion_length": 21.03125,
      "epoch": 2.7322175732217575,
      "grad_norm": 6.308757781982422,
      "kl": 0.4736328125,
      "learning_rate": 3.1799163179916314e-07,
      "loss": 0.0214,
      "reward": 0.7292242050170898,
      "reward_std": 0.04727287404239178,
      "reward_std_all": 0.059823036193847656,
      "rewards-max/forensics_sinkhorn_reward": 0.9707145988941193,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.9555836617946625,
      "rewards/forensics_sinkhorn_reward": 0.6847737431526184,
      "rewards/forensics_strict_edge_reward": 0.046942333690822124,
      "rewards/forensics_tiou_reward": 0.4634881615638733,
      "step": 653
    },
    {
      "completion_length": 20.078125,
      "epoch": 2.7364016736401675,
      "grad_norm": 5.5517802238464355,
      "kl": 0.4189453125,
      "learning_rate": 3.1694560669456065e-07,
      "loss": 0.079,
      "reward": 0.7665370106697083,
      "reward_std": 0.04411851614713669,
      "reward_std_all": 0.04454753175377846,
      "rewards-max/forensics_sinkhorn_reward": 0.979423463344574,
      "rewards-max/forensics_strict_edge_reward": 0.5976387783885002,
      "rewards-max/forensics_tiou_reward": 0.968743771314621,
      "rewards/forensics_sinkhorn_reward": 0.7279086709022522,
      "rewards/forensics_strict_edge_reward": 0.07656606193631887,
      "rewards/forensics_tiou_reward": 0.49479545652866364,
      "step": 654
    },
    {
      "completion_length": 20.484375,
      "epoch": 2.7405857740585775,
      "grad_norm": 7.934657573699951,
      "kl": 0.361328125,
      "learning_rate": 3.1589958158995816e-07,
      "loss": -0.0595,
      "reward": 0.7530920803546906,
      "reward_std": 0.05200514104217291,
      "reward_std_all": 0.05326339229941368,
      "rewards-max/forensics_sinkhorn_reward": 0.9762302339076996,
      "rewards-max/forensics_strict_edge_reward": 0.3680838346481323,
      "rewards-max/forensics_tiou_reward": 0.9628848135471344,
      "rewards/forensics_sinkhorn_reward": 0.7085408866405487,
      "rewards/forensics_strict_edge_reward": 0.027962355874478817,
      "rewards/forensics_tiou_reward": 0.45177629590034485,
      "step": 655
    },
    {
      "completion_length": 21.0625,
      "epoch": 2.7447698744769875,
      "grad_norm": 4.674673080444336,
      "kl": 0.384765625,
      "learning_rate": 3.148535564853556e-07,
      "loss": 0.0911,
      "reward": 0.7666643261909485,
      "reward_std": 0.059171875938773155,
      "reward_std_all": 0.054027849808335304,
      "rewards-max/forensics_sinkhorn_reward": 0.9882762730121613,
      "rewards-max/forensics_strict_edge_reward": 0.7009071409702301,
      "rewards-max/forensics_tiou_reward": 0.9978695809841156,
      "rewards/forensics_sinkhorn_reward": 0.7763249278068542,
      "rewards/forensics_strict_edge_reward": 0.08858132362365723,
      "rewards/forensics_tiou_reward": 0.6165001690387726,
      "step": 656
    },
    {
      "completion_length": 21.859375,
      "epoch": 2.7489539748953975,
      "grad_norm": 6.2009077072143555,
      "kl": 0.4697265625,
      "learning_rate": 3.1380753138075313e-07,
      "loss": 0.0283,
      "reward": 0.7262082695960999,
      "reward_std": 0.046600005123764277,
      "reward_std_all": 0.06302658468484879,
      "rewards-max/forensics_sinkhorn_reward": 0.9862072467803955,
      "rewards-max/forensics_strict_edge_reward": 0.6862294971942902,
      "rewards-max/forensics_tiou_reward": 0.991062343120575,
      "rewards/forensics_sinkhorn_reward": 0.7216619849205017,
      "rewards/forensics_strict_edge_reward": 0.07967009022831917,
      "rewards/forensics_tiou_reward": 0.5544519275426865,
      "step": 657
    },
    {
      "completion_length": 23.375,
      "epoch": 2.7531380753138075,
      "grad_norm": 6.732489109039307,
      "kl": 0.322265625,
      "learning_rate": 3.127615062761506e-07,
      "loss": 0.134,
      "reward": 0.7556353807449341,
      "reward_std": 0.0586250014603138,
      "reward_std_all": 0.058384038507938385,
      "rewards-max/forensics_sinkhorn_reward": 0.9575965702533722,
      "rewards-max/forensics_strict_edge_reward": 0.3703685849905014,
      "rewards-max/forensics_tiou_reward": 0.9261582493782043,
      "rewards/forensics_sinkhorn_reward": 0.7011620700359344,
      "rewards/forensics_strict_edge_reward": 0.05227469466626644,
      "rewards/forensics_tiou_reward": 0.5041297227144241,
      "step": 658
    },
    {
      "completion_length": 23.484375,
      "epoch": 2.7573221757322175,
      "grad_norm": 8.499361038208008,
      "kl": 0.3876953125,
      "learning_rate": 3.117154811715481e-07,
      "loss": 0.1328,
      "reward": 0.7295399308204651,
      "reward_std": 0.04784116894006729,
      "reward_std_all": 0.06309475377202034,
      "rewards-max/forensics_sinkhorn_reward": 0.9716905951499939,
      "rewards-max/forensics_strict_edge_reward": 0.4472203999757767,
      "rewards-max/forensics_tiou_reward": 0.9587175846099854,
      "rewards/forensics_sinkhorn_reward": 0.6803861558437347,
      "rewards/forensics_strict_edge_reward": 0.02879028394818306,
      "rewards/forensics_tiou_reward": 0.46422915160655975,
      "step": 659
    },
    {
      "completion_length": 20.46875,
      "epoch": 2.7615062761506275,
      "grad_norm": 5.427284240722656,
      "kl": 0.4580078125,
      "learning_rate": 3.106694560669456e-07,
      "loss": -0.0022,
      "reward": 0.7564401030540466,
      "reward_std": 0.06093944422900677,
      "reward_std_all": 0.059594739228487015,
      "rewards-max/forensics_sinkhorn_reward": 0.9725891649723053,
      "rewards-max/forensics_strict_edge_reward": 0.47726282477378845,
      "rewards-max/forensics_tiou_reward": 0.9562130868434906,
      "rewards/forensics_sinkhorn_reward": 0.7467033863067627,
      "rewards/forensics_strict_edge_reward": 0.036176880821585655,
      "rewards/forensics_tiou_reward": 0.5291850566864014,
      "step": 660
    },
    {
      "completion_length": 22.46875,
      "epoch": 2.7656903765690375,
      "grad_norm": 5.4391398429870605,
      "kl": 0.4697265625,
      "learning_rate": 3.096234309623431e-07,
      "loss": 0.0634,
      "reward": 0.7530242204666138,
      "reward_std": 0.0501578152179718,
      "reward_std_all": 0.051857421174645424,
      "rewards-max/forensics_sinkhorn_reward": 0.9763275682926178,
      "rewards-max/forensics_strict_edge_reward": 0.5129558444023132,
      "rewards-max/forensics_tiou_reward": 0.9654819965362549,
      "rewards/forensics_sinkhorn_reward": 0.7075425386428833,
      "rewards/forensics_strict_edge_reward": 0.03379256231710315,
      "rewards/forensics_tiou_reward": 0.4363008588552475,
      "step": 661
    },
    {
      "completion_length": 21.265625,
      "epoch": 2.7698744769874475,
      "grad_norm": 6.851397514343262,
      "kl": 0.65625,
      "learning_rate": 3.0857740585774054e-07,
      "loss": 0.124,
      "reward": 0.7291472852230072,
      "reward_std": 0.07296661660075188,
      "reward_std_all": 0.07337609305977821,
      "rewards-max/forensics_sinkhorn_reward": 0.9629503786563873,
      "rewards-max/forensics_strict_edge_reward": 0.4374866336584091,
      "rewards-max/forensics_tiou_reward": 0.9346369802951813,
      "rewards/forensics_sinkhorn_reward": 0.6804036796092987,
      "rewards/forensics_strict_edge_reward": 0.03731884993612766,
      "rewards/forensics_tiou_reward": 0.42513425648212433,
      "step": 662
    },
    {
      "completion_length": 20.484375,
      "epoch": 2.7740585774058575,
      "grad_norm": 4.950932502746582,
      "kl": 0.3427734375,
      "learning_rate": 3.0753138075313805e-07,
      "loss": 0.0092,
      "reward": 0.7602799832820892,
      "reward_std": 0.03402034938335419,
      "reward_std_all": 0.04216019902378321,
      "rewards-max/forensics_sinkhorn_reward": 0.9853448271751404,
      "rewards-max/forensics_strict_edge_reward": 0.4760446548461914,
      "rewards-max/forensics_tiou_reward": 0.9884500205516815,
      "rewards/forensics_sinkhorn_reward": 0.7503634393215179,
      "rewards/forensics_strict_edge_reward": 0.03360439417883754,
      "rewards/forensics_tiou_reward": 0.5731718838214874,
      "step": 663
    },
    {
      "completion_length": 23.40625,
      "epoch": 2.778242677824268,
      "grad_norm": 5.240385055541992,
      "kl": 0.3115234375,
      "learning_rate": 3.0648535564853557e-07,
      "loss": -0.0328,
      "reward": 0.7119823098182678,
      "reward_std": 0.08094033971428871,
      "reward_std_all": 0.08974403515458107,
      "rewards-max/forensics_sinkhorn_reward": 0.9839526116847992,
      "rewards-max/forensics_strict_edge_reward": 0.7331632375717163,
      "rewards-max/forensics_tiou_reward": 0.9869481027126312,
      "rewards/forensics_sinkhorn_reward": 0.6607449352741241,
      "rewards/forensics_strict_edge_reward": 0.05501333624124527,
      "rewards/forensics_tiou_reward": 0.5269457697868347,
      "step": 664
    },
    {
      "completion_length": 19.703125,
      "epoch": 2.782426778242678,
      "grad_norm": 6.212289810180664,
      "kl": 0.4052734375,
      "learning_rate": 3.054393305439331e-07,
      "loss": 0.0249,
      "reward": 0.6929146647453308,
      "reward_std": 0.10046860203146935,
      "reward_std_all": 0.11078276857733727,
      "rewards-max/forensics_sinkhorn_reward": 0.9755388796329498,
      "rewards-max/forensics_strict_edge_reward": 0.39520856738090515,
      "rewards-max/forensics_tiou_reward": 0.9661073684692383,
      "rewards/forensics_sinkhorn_reward": 0.6507003605365753,
      "rewards/forensics_strict_edge_reward": 0.03177755232900381,
      "rewards/forensics_tiou_reward": 0.4292939007282257,
      "step": 665
    },
    {
      "completion_length": 22.90625,
      "epoch": 2.786610878661088,
      "grad_norm": 6.327404499053955,
      "kl": 0.3330078125,
      "learning_rate": 3.043933054393305e-07,
      "loss": 0.1123,
      "reward": 0.7888858616352081,
      "reward_std": 0.02666629059240222,
      "reward_std_all": 0.02550814114511013,
      "rewards-max/forensics_sinkhorn_reward": 0.9733073711395264,
      "rewards-max/forensics_strict_edge_reward": 0.44024358689785004,
      "rewards-max/forensics_tiou_reward": 0.9583349525928497,
      "rewards/forensics_sinkhorn_reward": 0.7711983621120453,
      "rewards/forensics_strict_edge_reward": 0.04684313386678696,
      "rewards/forensics_tiou_reward": 0.5666609406471252,
      "step": 666
    },
    {
      "completion_length": 21.375,
      "epoch": 2.790794979079498,
      "grad_norm": 7.641754150390625,
      "kl": 0.3154296875,
      "learning_rate": 3.03347280334728e-07,
      "loss": 0.0757,
      "reward": 0.7481009364128113,
      "reward_std": 0.07346336543560028,
      "reward_std_all": 0.06878338009119034,
      "rewards-max/forensics_sinkhorn_reward": 0.9778668582439423,
      "rewards-max/forensics_strict_edge_reward": 0.3871150016784668,
      "rewards-max/forensics_tiou_reward": 0.9682987034320831,
      "rewards/forensics_sinkhorn_reward": 0.7334122061729431,
      "rewards/forensics_strict_edge_reward": 0.03434673044830561,
      "rewards/forensics_tiou_reward": 0.5418274998664856,
      "step": 667
    },
    {
      "completion_length": 22.78125,
      "epoch": 2.794979079497908,
      "grad_norm": 4.997477054595947,
      "kl": 0.3271484375,
      "learning_rate": 3.023012552301255e-07,
      "loss": 0.0049,
      "reward": 0.7106581330299377,
      "reward_std": 0.0300195156596601,
      "reward_std_all": 0.0640448909252882,
      "rewards-max/forensics_sinkhorn_reward": 0.9831049144268036,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9843913614749908,
      "rewards/forensics_sinkhorn_reward": 0.6858643293380737,
      "rewards/forensics_strict_edge_reward": 0.06909487582743168,
      "rewards/forensics_tiou_reward": 0.5103691518306732,
      "step": 668
    },
    {
      "completion_length": 20.765625,
      "epoch": 2.799163179916318,
      "grad_norm": 8.277026176452637,
      "kl": 0.318359375,
      "learning_rate": 3.01255230125523e-07,
      "loss": 0.022,
      "reward": 0.7327607274055481,
      "reward_std": 0.04602820239961147,
      "reward_std_all": 0.0622134692966938,
      "rewards-max/forensics_sinkhorn_reward": 0.966234564781189,
      "rewards-max/forensics_strict_edge_reward": 0.11302661895751953,
      "rewards-max/forensics_tiou_reward": 0.9384731650352478,
      "rewards/forensics_sinkhorn_reward": 0.6787143051624298,
      "rewards/forensics_strict_edge_reward": 0.006078742910176516,
      "rewards/forensics_tiou_reward": 0.4363012760877609,
      "step": 669
    },
    {
      "completion_length": 21.171875,
      "epoch": 2.803347280334728,
      "grad_norm": 5.616754531860352,
      "kl": 0.3515625,
      "learning_rate": 3.002092050209205e-07,
      "loss": 0.0644,
      "reward": 0.7474966645240784,
      "reward_std": 0.05016354285180569,
      "reward_std_all": 0.05422371532768011,
      "rewards-max/forensics_sinkhorn_reward": 0.9735101461410522,
      "rewards-max/forensics_strict_edge_reward": 0.5548860281705856,
      "rewards-max/forensics_tiou_reward": 0.9611872136592865,
      "rewards/forensics_sinkhorn_reward": 0.7168025076389313,
      "rewards/forensics_strict_edge_reward": 0.06036294810473919,
      "rewards/forensics_tiou_reward": 0.5418623089790344,
      "step": 670
    },
    {
      "completion_length": 20.296875,
      "epoch": 2.8075313807531384,
      "grad_norm": 6.018398761749268,
      "kl": 0.5068359375,
      "learning_rate": 2.9916317991631794e-07,
      "loss": 0.0342,
      "reward": 0.7856022715568542,
      "reward_std": 0.02980705164372921,
      "reward_std_all": 0.028520075604319572,
      "rewards-max/forensics_sinkhorn_reward": 0.9852894246578217,
      "rewards-max/forensics_strict_edge_reward": 0.765928328037262,
      "rewards-max/forensics_tiou_reward": 0.9895849525928497,
      "rewards/forensics_sinkhorn_reward": 0.7554551959037781,
      "rewards/forensics_strict_edge_reward": 0.05519452318549156,
      "rewards/forensics_tiou_reward": 0.575054407119751,
      "step": 671
    },
    {
      "completion_length": 20.40625,
      "epoch": 2.8117154811715483,
      "grad_norm": 6.527590751647949,
      "kl": 0.404296875,
      "learning_rate": 2.9811715481171546e-07,
      "loss": -0.0071,
      "reward": 0.7857187986373901,
      "reward_std": 0.026835248805582523,
      "reward_std_all": 0.02656039036810398,
      "rewards-max/forensics_sinkhorn_reward": 0.9741091728210449,
      "rewards-max/forensics_strict_edge_reward": 0.4713352173566818,
      "rewards-max/forensics_tiou_reward": 0.9634622633457184,
      "rewards/forensics_sinkhorn_reward": 0.7829404473304749,
      "rewards/forensics_strict_edge_reward": 0.050123135559260845,
      "rewards/forensics_tiou_reward": 0.5742649137973785,
      "step": 672
    },
    {
      "completion_length": 21.203125,
      "epoch": 2.8158995815899583,
      "grad_norm": 5.264060020446777,
      "kl": 0.37890625,
      "learning_rate": 2.9707112970711297e-07,
      "loss": 0.0786,
      "reward": 0.761126697063446,
      "reward_std": 0.07160656526684761,
      "reward_std_all": 0.06502383202314377,
      "rewards-max/forensics_sinkhorn_reward": 0.9413877427577972,
      "rewards-max/forensics_strict_edge_reward": 0.5094682248309255,
      "rewards-max/forensics_tiou_reward": 0.8913130164146423,
      "rewards/forensics_sinkhorn_reward": 0.7296528816223145,
      "rewards/forensics_strict_edge_reward": 0.044768884137738496,
      "rewards/forensics_tiou_reward": 0.5004457831382751,
      "step": 673
    },
    {
      "completion_length": 21.15625,
      "epoch": 2.8200836820083683,
      "grad_norm": 6.139864444732666,
      "kl": 0.3115234375,
      "learning_rate": 2.960251046025105e-07,
      "loss": 0.0294,
      "reward": 0.7347671687602997,
      "reward_std": 0.03353785630315542,
      "reward_std_all": 0.05406430549919605,
      "rewards-max/forensics_sinkhorn_reward": 0.9788782894611359,
      "rewards-max/forensics_strict_edge_reward": 0.6076286435127258,
      "rewards-max/forensics_tiou_reward": 0.9742078483104706,
      "rewards/forensics_sinkhorn_reward": 0.6738282442092896,
      "rewards/forensics_strict_edge_reward": 0.06326437275856733,
      "rewards/forensics_tiou_reward": 0.48269444704055786,
      "step": 674
    },
    {
      "completion_length": 19.65625,
      "epoch": 2.8242677824267783,
      "grad_norm": 6.234304428100586,
      "kl": 0.3369140625,
      "learning_rate": 2.9497907949790794e-07,
      "loss": 0.0052,
      "reward": 0.7807992398738861,
      "reward_std": 0.01631585438735783,
      "reward_std_all": 0.024091563187539577,
      "rewards-max/forensics_sinkhorn_reward": 0.9591147005558014,
      "rewards-max/forensics_strict_edge_reward": 0.4231441020965576,
      "rewards-max/forensics_tiou_reward": 0.9288534224033356,
      "rewards/forensics_sinkhorn_reward": 0.7692461609840393,
      "rewards/forensics_strict_edge_reward": 0.02431343588978052,
      "rewards/forensics_tiou_reward": 0.5292256772518158,
      "step": 675
    },
    {
      "completion_length": 20.546875,
      "epoch": 2.8284518828451883,
      "grad_norm": 6.957098007202148,
      "kl": 0.431640625,
      "learning_rate": 2.939330543933054e-07,
      "loss": 0.0228,
      "reward": 0.7711698114871979,
      "reward_std": 0.04066258296370506,
      "reward_std_all": 0.0397810023277998,
      "rewards-max/forensics_sinkhorn_reward": 0.9698246419429779,
      "rewards-max/forensics_strict_edge_reward": 0.5074994415044785,
      "rewards-max/forensics_tiou_reward": 0.9517010450363159,
      "rewards/forensics_sinkhorn_reward": 0.7782354950904846,
      "rewards/forensics_strict_edge_reward": 0.06278194114565849,
      "rewards/forensics_tiou_reward": 0.6070257425308228,
      "step": 676
    },
    {
      "completion_length": 21.90625,
      "epoch": 2.8326359832635983,
      "grad_norm": 7.564986705780029,
      "kl": 0.361328125,
      "learning_rate": 2.928870292887029e-07,
      "loss": 0.0407,
      "reward": 0.7709662020206451,
      "reward_std": 0.05537877604365349,
      "reward_std_all": 0.05043211951851845,
      "rewards-max/forensics_sinkhorn_reward": 0.97858065366745,
      "rewards-max/forensics_strict_edge_reward": 0.707136332988739,
      "rewards-max/forensics_tiou_reward": 0.9754639565944672,
      "rewards/forensics_sinkhorn_reward": 0.7422964572906494,
      "rewards/forensics_strict_edge_reward": 0.05800171382725239,
      "rewards/forensics_tiou_reward": 0.5241059064865112,
      "step": 677
    },
    {
      "completion_length": 22.875,
      "epoch": 2.8368200836820083,
      "grad_norm": 11.520102500915527,
      "kl": 0.3056640625,
      "learning_rate": 2.9184100418410043e-07,
      "loss": 0.1109,
      "reward": 0.6791470348834991,
      "reward_std": 0.06490935944020748,
      "reward_std_all": 0.09738622233271599,
      "rewards-max/forensics_sinkhorn_reward": 0.9649932384490967,
      "rewards-max/forensics_strict_edge_reward": 0.5253343880176544,
      "rewards-max/forensics_tiou_reward": 0.9469847083091736,
      "rewards/forensics_sinkhorn_reward": 0.6412725746631622,
      "rewards/forensics_strict_edge_reward": 0.040826842188835144,
      "rewards/forensics_tiou_reward": 0.45341622829437256,
      "step": 678
    },
    {
      "completion_length": 21.625,
      "epoch": 2.8410041841004183,
      "grad_norm": 8.165231704711914,
      "kl": 0.3154296875,
      "learning_rate": 2.907949790794979e-07,
      "loss": 0.0556,
      "reward": 0.7669598162174225,
      "reward_std": 0.04357006214559078,
      "reward_std_all": 0.04483684245496988,
      "rewards-max/forensics_sinkhorn_reward": 0.9578536450862885,
      "rewards-max/forensics_strict_edge_reward": 0.4839124232530594,
      "rewards-max/forensics_tiou_reward": 0.9263882040977478,
      "rewards/forensics_sinkhorn_reward": 0.7174812853336334,
      "rewards/forensics_strict_edge_reward": 0.036218960769474506,
      "rewards/forensics_tiou_reward": 0.5046760141849518,
      "step": 679
    },
    {
      "completion_length": 21.765625,
      "epoch": 2.8451882845188283,
      "grad_norm": 4.819461345672607,
      "kl": 0.40625,
      "learning_rate": 2.897489539748954e-07,
      "loss": -0.1013,
      "reward": 0.7792471945285797,
      "reward_std": 0.025009071978274733,
      "reward_std_all": 0.027006050571799278,
      "rewards-max/forensics_sinkhorn_reward": 0.9845481216907501,
      "rewards-max/forensics_strict_edge_reward": 0.7482926547527313,
      "rewards-max/forensics_tiou_reward": 0.9860994815826416,
      "rewards/forensics_sinkhorn_reward": 0.7413030564785004,
      "rewards/forensics_strict_edge_reward": 0.04205530695617199,
      "rewards/forensics_tiou_reward": 0.5122984349727631,
      "step": 680
    },
    {
      "completion_length": 24.421875,
      "epoch": 2.8493723849372383,
      "grad_norm": 5.79154109954834,
      "kl": 0.3857421875,
      "learning_rate": 2.8870292887029286e-07,
      "loss": 0.0994,
      "reward": 0.7146403789520264,
      "reward_std": 0.07778982073068619,
      "reward_std_all": 0.08152865618467331,
      "rewards-max/forensics_sinkhorn_reward": 0.9424578547477722,
      "rewards-max/forensics_strict_edge_reward": 0.4374866336584091,
      "rewards-max/forensics_tiou_reward": 0.8962584435939789,
      "rewards/forensics_sinkhorn_reward": 0.6156485378742218,
      "rewards/forensics_strict_edge_reward": 0.026479098480194807,
      "rewards/forensics_tiou_reward": 0.3602249026298523,
      "step": 681
    },
    {
      "completion_length": 20.671875,
      "epoch": 2.8535564853556483,
      "grad_norm": 5.5949811935424805,
      "kl": 0.375,
      "learning_rate": 2.8765690376569037e-07,
      "loss": -0.053,
      "reward": 0.7974998950958252,
      "reward_std": 0.013771103229373693,
      "reward_std_all": 0.01426029298454523,
      "rewards-max/forensics_sinkhorn_reward": 0.9505260288715363,
      "rewards-max/forensics_strict_edge_reward": 0.559220913797617,
      "rewards-max/forensics_tiou_reward": 0.91033536195755,
      "rewards/forensics_sinkhorn_reward": 0.7780574560165405,
      "rewards/forensics_strict_edge_reward": 0.047883258666843176,
      "rewards/forensics_tiou_reward": 0.5450438559055328,
      "step": 682
    },
    {
      "completion_length": 20.890625,
      "epoch": 2.8577405857740583,
      "grad_norm": 9.2949857711792,
      "kl": 0.3125,
      "learning_rate": 2.8661087866108783e-07,
      "loss": -0.0063,
      "reward": 0.7252087295055389,
      "reward_std": 0.08128818869590759,
      "reward_std_all": 0.08236763626337051,
      "rewards-max/forensics_sinkhorn_reward": 0.9727694392204285,
      "rewards-max/forensics_strict_edge_reward": 0.48575712740421295,
      "rewards-max/forensics_tiou_reward": 0.9550026655197144,
      "rewards/forensics_sinkhorn_reward": 0.6942809522151947,
      "rewards/forensics_strict_edge_reward": 0.04422299563884735,
      "rewards/forensics_tiou_reward": 0.49807700514793396,
      "step": 683
    },
    {
      "completion_length": 21.953125,
      "epoch": 2.8619246861924688,
      "grad_norm": 6.588461875915527,
      "kl": 0.3056640625,
      "learning_rate": 2.8556485355648535e-07,
      "loss": 0.0894,
      "reward": 0.7548307776451111,
      "reward_std": 0.07408250868320465,
      "reward_std_all": 0.06827386282384396,
      "rewards-max/forensics_sinkhorn_reward": 0.9567137658596039,
      "rewards-max/forensics_strict_edge_reward": 0.14518444705754519,
      "rewards-max/forensics_tiou_reward": 0.9208010733127594,
      "rewards/forensics_sinkhorn_reward": 0.686851441860199,
      "rewards/forensics_strict_edge_reward": 0.007870594039559364,
      "rewards/forensics_tiou_reward": 0.4326549619436264,
      "step": 684
    },
    {
      "completion_length": 20.71875,
      "epoch": 2.8661087866108788,
      "grad_norm": 4.943036079406738,
      "kl": 0.29345703125,
      "learning_rate": 2.8451882845188286e-07,
      "loss": 0.0481,
      "reward": 0.7614331245422363,
      "reward_std": 0.04522025864571333,
      "reward_std_all": 0.046336100436747074,
      "rewards-max/forensics_sinkhorn_reward": 0.9564098417758942,
      "rewards-max/forensics_strict_edge_reward": 0.3434525206685066,
      "rewards-max/forensics_tiou_reward": 0.9184644520282745,
      "rewards/forensics_sinkhorn_reward": 0.714019387960434,
      "rewards/forensics_strict_edge_reward": 0.03507204167544842,
      "rewards/forensics_tiou_reward": 0.5329421758651733,
      "step": 685
    },
    {
      "completion_length": 21.578125,
      "epoch": 2.8702928870292888,
      "grad_norm": 5.057350158691406,
      "kl": 0.3525390625,
      "learning_rate": 2.834728033472803e-07,
      "loss": 0.0078,
      "reward": 0.7823624312877655,
      "reward_std": 0.030119483009912074,
      "reward_std_all": 0.029201569384895265,
      "rewards-max/forensics_sinkhorn_reward": 0.98700150847435,
      "rewards-max/forensics_strict_edge_reward": 0.8187680542469025,
      "rewards-max/forensics_tiou_reward": 0.9939373433589935,
      "rewards/forensics_sinkhorn_reward": 0.7875535786151886,
      "rewards/forensics_strict_edge_reward": 0.11204876750707626,
      "rewards/forensics_tiou_reward": 0.6168302893638611,
      "step": 686
    },
    {
      "completion_length": 19.828125,
      "epoch": 2.8744769874476988,
      "grad_norm": 6.157914638519287,
      "kl": 0.3359375,
      "learning_rate": 2.824267782426778e-07,
      "loss": -0.0072,
      "reward": 0.7504193484783173,
      "reward_std": 0.03278946503996849,
      "reward_std_all": 0.04450366646051407,
      "rewards-max/forensics_sinkhorn_reward": 0.9764873087406158,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9677926599979401,
      "rewards/forensics_sinkhorn_reward": 0.7299673855304718,
      "rewards/forensics_strict_edge_reward": 0.041667722165584564,
      "rewards/forensics_tiou_reward": 0.5392099618911743,
      "step": 687
    },
    {
      "completion_length": 21.546875,
      "epoch": 2.8786610878661087,
      "grad_norm": 7.287499904632568,
      "kl": 0.328125,
      "learning_rate": 2.813807531380753e-07,
      "loss": 0.0376,
      "reward": 0.769800454378128,
      "reward_std": 0.026919135823845863,
      "reward_std_all": 0.03404415398836136,
      "rewards-max/forensics_sinkhorn_reward": 0.9758972525596619,
      "rewards-max/forensics_strict_edge_reward": 0.42500291764736176,
      "rewards-max/forensics_tiou_reward": 0.9630067348480225,
      "rewards/forensics_sinkhorn_reward": 0.7350730299949646,
      "rewards/forensics_strict_edge_reward": 0.05167915113270283,
      "rewards/forensics_tiou_reward": 0.5291432738304138,
      "step": 688
    },
    {
      "completion_length": 22.265625,
      "epoch": 2.8828451882845187,
      "grad_norm": 5.942516803741455,
      "kl": 0.3310546875,
      "learning_rate": 2.803347280334728e-07,
      "loss": 0.0088,
      "reward": 0.77603679895401,
      "reward_std": 0.026862522587180138,
      "reward_std_all": 0.0292995423078537,
      "rewards-max/forensics_sinkhorn_reward": 0.9564151167869568,
      "rewards-max/forensics_strict_edge_reward": 0.09770023263990879,
      "rewards-max/forensics_tiou_reward": 0.9146660566329956,
      "rewards/forensics_sinkhorn_reward": 0.7211261987686157,
      "rewards/forensics_strict_edge_reward": 0.0054516581585630774,
      "rewards/forensics_tiou_reward": 0.5066761672496796,
      "step": 689
    },
    {
      "completion_length": 22.640625,
      "epoch": 2.8870292887029287,
      "grad_norm": 9.057245254516602,
      "kl": 0.302734375,
      "learning_rate": 2.792887029288703e-07,
      "loss": 0.0396,
      "reward": 0.7197985649108887,
      "reward_std": 0.06629389896988869,
      "reward_std_all": 0.07636742107570171,
      "rewards-max/forensics_sinkhorn_reward": 0.9636827707290649,
      "rewards-max/forensics_strict_edge_reward": 0.3503752648830414,
      "rewards-max/forensics_tiou_reward": 0.9326073229312897,
      "rewards/forensics_sinkhorn_reward": 0.653252363204956,
      "rewards/forensics_strict_edge_reward": 0.0189457843080163,
      "rewards/forensics_tiou_reward": 0.3748626857995987,
      "step": 690
    },
    {
      "completion_length": 22.328125,
      "epoch": 2.891213389121339,
      "grad_norm": 6.241459846496582,
      "kl": 0.3837890625,
      "learning_rate": 2.782426778242677e-07,
      "loss": 0.0523,
      "reward": 0.74165078997612,
      "reward_std": 0.06662430614233017,
      "reward_std_all": 0.06636291556060314,
      "rewards-max/forensics_sinkhorn_reward": 0.9507991075515747,
      "rewards-max/forensics_strict_edge_reward": 0.22191549092531204,
      "rewards-max/forensics_tiou_reward": 0.8890549540519714,
      "rewards/forensics_sinkhorn_reward": 0.7034066021442413,
      "rewards/forensics_strict_edge_reward": 0.012406219728291035,
      "rewards/forensics_tiou_reward": 0.4662623405456543,
      "step": 691
    },
    {
      "completion_length": 19.65625,
      "epoch": 2.895397489539749,
      "grad_norm": 7.966770648956299,
      "kl": 0.39453125,
      "learning_rate": 2.7719665271966524e-07,
      "loss": -0.0459,
      "reward": 0.7731303870677948,
      "reward_std": 0.035682376474142075,
      "reward_std_all": 0.03554805554449558,
      "rewards-max/forensics_sinkhorn_reward": 0.9806125462055206,
      "rewards-max/forensics_strict_edge_reward": 0.6312568187713623,
      "rewards-max/forensics_tiou_reward": 0.978866845369339,
      "rewards/forensics_sinkhorn_reward": 0.7783738970756531,
      "rewards/forensics_strict_edge_reward": 0.06365383043885231,
      "rewards/forensics_tiou_reward": 0.584458976984024,
      "step": 692
    },
    {
      "completion_length": 20.765625,
      "epoch": 2.899581589958159,
      "grad_norm": 8.088898658752441,
      "kl": 0.57421875,
      "learning_rate": 2.7615062761506275e-07,
      "loss": 0.0773,
      "reward": 0.7118070423603058,
      "reward_std": 0.06984739005565643,
      "reward_std_all": 0.07852444052696228,
      "rewards-max/forensics_sinkhorn_reward": 0.9316661357879639,
      "rewards-max/forensics_strict_edge_reward": 0.1288958564400673,
      "rewards-max/forensics_tiou_reward": 0.8650543987751007,
      "rewards/forensics_sinkhorn_reward": 0.6291514933109283,
      "rewards/forensics_strict_edge_reward": 0.00942315743304789,
      "rewards/forensics_tiou_reward": 0.34888820350170135,
      "step": 693
    },
    {
      "completion_length": 21.109375,
      "epoch": 2.903765690376569,
      "grad_norm": 5.747859954833984,
      "kl": 0.28369140625,
      "learning_rate": 2.7510460251046026e-07,
      "loss": -0.0675,
      "reward": 0.7864765524864197,
      "reward_std": 0.028688893653452396,
      "reward_std_all": 0.027709096670150757,
      "rewards-max/forensics_sinkhorn_reward": 0.9644889831542969,
      "rewards-max/forensics_strict_edge_reward": 0.44934946298599243,
      "rewards-max/forensics_tiou_reward": 0.9421136975288391,
      "rewards/forensics_sinkhorn_reward": 0.7558699548244476,
      "rewards/forensics_strict_edge_reward": 0.04050474427640438,
      "rewards/forensics_tiou_reward": 0.5460643172264099,
      "step": 694
    },
    {
      "completion_length": 20.75,
      "epoch": 2.907949790794979,
      "grad_norm": 5.288061618804932,
      "kl": 0.3701171875,
      "learning_rate": 2.740585774058578e-07,
      "loss": 0.0745,
      "reward": 0.7540296018123627,
      "reward_std": 0.06749747879803181,
      "reward_std_all": 0.06402761861681938,
      "rewards-max/forensics_sinkhorn_reward": 0.9689400494098663,
      "rewards-max/forensics_strict_edge_reward": 0.4685862995684147,
      "rewards-max/forensics_tiou_reward": 0.9455219805240631,
      "rewards/forensics_sinkhorn_reward": 0.7426197826862335,
      "rewards/forensics_strict_edge_reward": 0.028938230127096176,
      "rewards/forensics_tiou_reward": 0.5311965048313141,
      "step": 695
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.912133891213389,
      "grad_norm": 5.330924034118652,
      "kl": 0.359375,
      "learning_rate": 2.730125523012552e-07,
      "loss": 0.0429,
      "reward": 0.7869099974632263,
      "reward_std": 0.02856281027197838,
      "reward_std_all": 0.026836754754185677,
      "rewards-max/forensics_sinkhorn_reward": 0.9802334606647491,
      "rewards-max/forensics_strict_edge_reward": 0.7558856904506683,
      "rewards-max/forensics_tiou_reward": 0.97695192694664,
      "rewards/forensics_sinkhorn_reward": 0.7858806550502777,
      "rewards/forensics_strict_edge_reward": 0.08540020510554314,
      "rewards/forensics_tiou_reward": 0.6196934282779694,
      "step": 696
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.916317991631799,
      "grad_norm": 7.714076519012451,
      "kl": 0.306640625,
      "learning_rate": 2.719665271966527e-07,
      "loss": 0.0642,
      "reward": 0.7824734449386597,
      "reward_std": 0.03598117083311081,
      "reward_std_all": 0.034049270674586296,
      "rewards-max/forensics_sinkhorn_reward": 0.9597460925579071,
      "rewards-max/forensics_strict_edge_reward": 0.3564630448818207,
      "rewards-max/forensics_tiou_reward": 0.9292954802513123,
      "rewards/forensics_sinkhorn_reward": 0.7062547206878662,
      "rewards/forensics_strict_edge_reward": 0.029893336817622185,
      "rewards/forensics_tiou_reward": 0.48401685059070587,
      "step": 697
    },
    {
      "completion_length": 20.375,
      "epoch": 2.920502092050209,
      "grad_norm": 8.445547103881836,
      "kl": 0.4169921875,
      "learning_rate": 2.709205020920502e-07,
      "loss": -0.0159,
      "reward": 0.7773952782154083,
      "reward_std": 0.024865017272531986,
      "reward_std_all": 0.027787726372480392,
      "rewards-max/forensics_sinkhorn_reward": 0.9746714234352112,
      "rewards-max/forensics_strict_edge_reward": 0.608120322227478,
      "rewards-max/forensics_tiou_reward": 0.9617607295513153,
      "rewards/forensics_sinkhorn_reward": 0.7398686110973358,
      "rewards/forensics_strict_edge_reward": 0.08344816230237484,
      "rewards/forensics_tiou_reward": 0.5466679036617279,
      "step": 698
    },
    {
      "completion_length": 21.625,
      "epoch": 2.924686192468619,
      "grad_norm": 5.980527400970459,
      "kl": 0.4697265625,
      "learning_rate": 2.698744769874477e-07,
      "loss": 0.0994,
      "reward": 0.738659143447876,
      "reward_std": 0.042324747424572706,
      "reward_std_all": 0.05525086261332035,
      "rewards-max/forensics_sinkhorn_reward": 0.9782897531986237,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9713726043701172,
      "rewards/forensics_sinkhorn_reward": 0.7355929315090179,
      "rewards/forensics_strict_edge_reward": 0.038694280199706554,
      "rewards/forensics_tiou_reward": 0.5474412441253662,
      "step": 699
    },
    {
      "completion_length": 22.1875,
      "epoch": 2.928870292887029,
      "grad_norm": 6.546034812927246,
      "kl": 0.3369140625,
      "learning_rate": 2.688284518828452e-07,
      "loss": 0.0709,
      "reward": 0.7665168642997742,
      "reward_std": 0.04905044939368963,
      "reward_std_all": 0.046903425827622414,
      "rewards-max/forensics_sinkhorn_reward": 0.9866777658462524,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.993552029132843,
      "rewards/forensics_sinkhorn_reward": 0.7333124876022339,
      "rewards/forensics_strict_edge_reward": 0.07788598351180553,
      "rewards/forensics_tiou_reward": 0.5390593856573105,
      "step": 700
    },
    {
      "completion_length": 23.90625,
      "epoch": 2.933054393305439,
      "grad_norm": 6.495804786682129,
      "kl": 0.4267578125,
      "learning_rate": 2.6778242677824264e-07,
      "loss": -0.082,
      "reward": 0.7499671578407288,
      "reward_std": 0.03605207987129688,
      "reward_std_all": 0.043808743357658386,
      "rewards-max/forensics_sinkhorn_reward": 0.9789443016052246,
      "rewards-max/forensics_strict_edge_reward": 0.4346180260181427,
      "rewards-max/forensics_tiou_reward": 0.9713272154331207,
      "rewards/forensics_sinkhorn_reward": 0.6790817677974701,
      "rewards/forensics_strict_edge_reward": 0.03340817987918854,
      "rewards/forensics_tiou_reward": 0.5050303936004639,
      "step": 701
    },
    {
      "completion_length": 24.78125,
      "epoch": 2.937238493723849,
      "grad_norm": 8.804327011108398,
      "kl": 0.2939453125,
      "learning_rate": 2.6673640167364015e-07,
      "loss": 0.0592,
      "reward": 0.7624693512916565,
      "reward_std": 0.03173314034938812,
      "reward_std_all": 0.03780250437557697,
      "rewards-max/forensics_sinkhorn_reward": 0.982060432434082,
      "rewards-max/forensics_strict_edge_reward": 0.42371489852666855,
      "rewards-max/forensics_tiou_reward": 0.9803202450275421,
      "rewards/forensics_sinkhorn_reward": 0.7143575549125671,
      "rewards/forensics_strict_edge_reward": 0.025966698303818703,
      "rewards/forensics_tiou_reward": 0.5518068373203278,
      "step": 702
    },
    {
      "completion_length": 21.046875,
      "epoch": 2.941422594142259,
      "grad_norm": 5.305054187774658,
      "kl": 0.3486328125,
      "learning_rate": 2.6569037656903767e-07,
      "loss": 0.0311,
      "reward": 0.7433605790138245,
      "reward_std": 0.08561187982559204,
      "reward_std_all": 0.07877514138817787,
      "rewards-max/forensics_sinkhorn_reward": 0.9817048013210297,
      "rewards-max/forensics_strict_edge_reward": 0.5533463060855865,
      "rewards-max/forensics_tiou_reward": 0.9769164025783539,
      "rewards/forensics_sinkhorn_reward": 0.7370882034301758,
      "rewards/forensics_strict_edge_reward": 0.06014817953109741,
      "rewards/forensics_tiou_reward": 0.573894590139389,
      "step": 703
    },
    {
      "completion_length": 21.046875,
      "epoch": 2.9456066945606696,
      "grad_norm": 5.138610363006592,
      "kl": 0.37109375,
      "learning_rate": 2.6464435146443513e-07,
      "loss": 0.0613,
      "reward": 0.7567468881607056,
      "reward_std": 0.02627447433769703,
      "reward_std_all": 0.041198753751814365,
      "rewards-max/forensics_sinkhorn_reward": 0.9233106672763824,
      "rewards-max/forensics_strict_edge_reward": 0.10632913187146187,
      "rewards-max/forensics_tiou_reward": 0.8264243006706238,
      "rewards/forensics_sinkhorn_reward": 0.7005705833435059,
      "rewards/forensics_strict_edge_reward": 0.00760434684343636,
      "rewards/forensics_tiou_reward": 0.494235634803772,
      "step": 704
    },
    {
      "completion_length": 21.125,
      "epoch": 2.9497907949790796,
      "grad_norm": 17.598955154418945,
      "kl": 0.357421875,
      "learning_rate": 2.6359832635983264e-07,
      "loss": 0.1704,
      "reward": 0.7100631296634674,
      "reward_std": 0.06995988450944424,
      "reward_std_all": 0.08153491094708443,
      "rewards-max/forensics_sinkhorn_reward": 0.9646576344966888,
      "rewards-max/forensics_strict_edge_reward": 0.3385289905127138,
      "rewards-max/forensics_tiou_reward": 0.9337009489536285,
      "rewards/forensics_sinkhorn_reward": 0.6626569330692291,
      "rewards/forensics_strict_edge_reward": 0.018388744967523962,
      "rewards/forensics_tiou_reward": 0.4448316991329193,
      "step": 705
    },
    {
      "completion_length": 22.453125,
      "epoch": 2.9539748953974896,
      "grad_norm": 5.786745071411133,
      "kl": 0.40625,
      "learning_rate": 2.625523012552301e-07,
      "loss": 0.0404,
      "reward": 0.7114052772521973,
      "reward_std": 0.06886380817741156,
      "reward_std_all": 0.08037366718053818,
      "rewards-max/forensics_sinkhorn_reward": 0.9795095026493073,
      "rewards-max/forensics_strict_edge_reward": 0.375261589884758,
      "rewards-max/forensics_tiou_reward": 0.9705302715301514,
      "rewards/forensics_sinkhorn_reward": 0.6816008985042572,
      "rewards/forensics_strict_edge_reward": 0.023972198367118835,
      "rewards/forensics_tiou_reward": 0.49136602878570557,
      "step": 706
    },
    {
      "completion_length": 19.765625,
      "epoch": 2.9581589958158996,
      "grad_norm": 6.9801106452941895,
      "kl": 0.345703125,
      "learning_rate": 2.615062761506276e-07,
      "loss": -0.0363,
      "reward": 0.792458564043045,
      "reward_std": 0.020024420227855444,
      "reward_std_all": 0.02002362860366702,
      "rewards-max/forensics_sinkhorn_reward": 0.9648207426071167,
      "rewards-max/forensics_strict_edge_reward": 0.47726282477378845,
      "rewards-max/forensics_tiou_reward": 0.9415770769119263,
      "rewards/forensics_sinkhorn_reward": 0.7646647989749908,
      "rewards/forensics_strict_edge_reward": 0.04867195338010788,
      "rewards/forensics_tiou_reward": 0.5512132048606873,
      "step": 707
    },
    {
      "completion_length": 21.765625,
      "epoch": 2.9623430962343096,
      "grad_norm": 5.592495918273926,
      "kl": 0.302734375,
      "learning_rate": 2.6046025104602507e-07,
      "loss": 0.1103,
      "reward": 0.7399880290031433,
      "reward_std": 0.07069243490695953,
      "reward_std_all": 0.06984934955835342,
      "rewards-max/forensics_sinkhorn_reward": 0.9739220440387726,
      "rewards-max/forensics_strict_edge_reward": 0.47008779272437096,
      "rewards-max/forensics_tiou_reward": 0.9636575281620026,
      "rewards/forensics_sinkhorn_reward": 0.7175254225730896,
      "rewards/forensics_strict_edge_reward": 0.04733597254380584,
      "rewards/forensics_tiou_reward": 0.5519620776176453,
      "step": 708
    },
    {
      "completion_length": 22.78125,
      "epoch": 2.9665271966527196,
      "grad_norm": 6.485538005828857,
      "kl": 0.3818359375,
      "learning_rate": 2.594142259414226e-07,
      "loss": 0.0301,
      "reward": 0.7793497145175934,
      "reward_std": 0.027868518605828285,
      "reward_std_all": 0.028650786727666855,
      "rewards-max/forensics_sinkhorn_reward": 0.9823931157588959,
      "rewards-max/forensics_strict_edge_reward": 0.6573699712753296,
      "rewards-max/forensics_tiou_reward": 0.9836337864398956,
      "rewards/forensics_sinkhorn_reward": 0.7634562849998474,
      "rewards/forensics_strict_edge_reward": 0.048316389322280884,
      "rewards/forensics_tiou_reward": 0.5816117525100708,
      "step": 709
    },
    {
      "completion_length": 20.15625,
      "epoch": 2.9707112970711296,
      "grad_norm": 7.2236528396606445,
      "kl": 0.4248046875,
      "learning_rate": 2.583682008368201e-07,
      "loss": -0.0622,
      "reward": 0.7882266938686371,
      "reward_std": 0.030413884902372956,
      "reward_std_all": 0.028232034062966704,
      "rewards-max/forensics_sinkhorn_reward": 0.9782046675682068,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.975255012512207,
      "rewards/forensics_sinkhorn_reward": 0.7663844525814056,
      "rewards/forensics_strict_edge_reward": 0.07824758440256119,
      "rewards/forensics_tiou_reward": 0.5820452868938446,
      "step": 710
    },
    {
      "completion_length": 21.15625,
      "epoch": 2.97489539748954,
      "grad_norm": 6.868934631347656,
      "kl": 0.337890625,
      "learning_rate": 2.5732217573221756e-07,
      "loss": 0.067,
      "reward": 0.7914076745510101,
      "reward_std": 0.018116148188710213,
      "reward_std_all": 0.01874202862381935,
      "rewards-max/forensics_sinkhorn_reward": 0.9813697040081024,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9817725718021393,
      "rewards/forensics_sinkhorn_reward": 0.7636487782001495,
      "rewards/forensics_strict_edge_reward": 0.05418355576694012,
      "rewards/forensics_tiou_reward": 0.5556806027889252,
      "step": 711
    },
    {
      "completion_length": 20.953125,
      "epoch": 2.97907949790795,
      "grad_norm": 6.458691596984863,
      "kl": 0.3896484375,
      "learning_rate": 2.56276150627615e-07,
      "loss": -0.0018,
      "reward": 0.7622056007385254,
      "reward_std": 0.017109688371419907,
      "reward_std_all": 0.03288707276806235,
      "rewards-max/forensics_sinkhorn_reward": 0.9768152236938477,
      "rewards-max/forensics_strict_edge_reward": 0.34872467815876007,
      "rewards-max/forensics_tiou_reward": 0.9594531059265137,
      "rewards/forensics_sinkhorn_reward": 0.7195181846618652,
      "rewards/forensics_strict_edge_reward": 0.03125750366598368,
      "rewards/forensics_tiou_reward": 0.5190520584583282,
      "step": 712
    },
    {
      "completion_length": 20.828125,
      "epoch": 2.98326359832636,
      "grad_norm": 7.133660316467285,
      "kl": 0.4033203125,
      "learning_rate": 2.5523012552301253e-07,
      "loss": 0.0127,
      "reward": 0.7792862951755524,
      "reward_std": 0.029221860226243734,
      "reward_std_all": 0.029481634497642517,
      "rewards-max/forensics_sinkhorn_reward": 0.9833921194076538,
      "rewards-max/forensics_strict_edge_reward": 0.6304196715354919,
      "rewards-max/forensics_tiou_reward": 0.9842309057712555,
      "rewards/forensics_sinkhorn_reward": 0.785777360200882,
      "rewards/forensics_strict_edge_reward": 0.08030183240771294,
      "rewards/forensics_tiou_reward": 0.6340508162975311,
      "step": 713
    },
    {
      "completion_length": 23.1875,
      "epoch": 2.98744769874477,
      "grad_norm": 5.29656457901001,
      "kl": 0.408203125,
      "learning_rate": 2.5418410041841004e-07,
      "loss": 0.0213,
      "reward": 0.7923600077629089,
      "reward_std": 0.023575241677463055,
      "reward_std_all": 0.022253483533859253,
      "rewards-max/forensics_sinkhorn_reward": 0.9839348196983337,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9845693409442902,
      "rewards/forensics_sinkhorn_reward": 0.7853022515773773,
      "rewards/forensics_strict_edge_reward": 0.0585787296295166,
      "rewards/forensics_tiou_reward": 0.596257746219635,
      "step": 714
    },
    {
      "completion_length": 23.4375,
      "epoch": 2.99163179916318,
      "grad_norm": 5.68564510345459,
      "kl": 0.3701171875,
      "learning_rate": 2.5313807531380756e-07,
      "loss": -0.0574,
      "reward": 0.7658307552337646,
      "reward_std": 0.04238072596490383,
      "reward_std_all": 0.0430446844547987,
      "rewards-max/forensics_sinkhorn_reward": 0.9806645810604095,
      "rewards-max/forensics_strict_edge_reward": 0.4628827217966318,
      "rewards-max/forensics_tiou_reward": 0.9727858006954193,
      "rewards/forensics_sinkhorn_reward": 0.7248095273971558,
      "rewards/forensics_strict_edge_reward": 0.02313197811599821,
      "rewards/forensics_tiou_reward": 0.5165197551250458,
      "step": 715
    },
    {
      "completion_length": 20.71875,
      "epoch": 2.99581589958159,
      "grad_norm": 5.603687286376953,
      "kl": 0.353515625,
      "learning_rate": 2.5209205020920496e-07,
      "loss": 0.1181,
      "reward": 0.7412345111370087,
      "reward_std": 0.049320803955197334,
      "reward_std_all": 0.06160077266395092,
      "rewards-max/forensics_sinkhorn_reward": 0.9834585785865784,
      "rewards-max/forensics_strict_edge_reward": 0.5361751317977905,
      "rewards-max/forensics_tiou_reward": 0.9860873818397522,
      "rewards/forensics_sinkhorn_reward": 0.6997491419315338,
      "rewards/forensics_strict_edge_reward": 0.02882628981024027,
      "rewards/forensics_tiou_reward": 0.4900110960006714,
      "step": 716
    },
    {
      "completion_length": 20.000000953674316,
      "epoch": 3.0,
      "grad_norm": 6.29195499420166,
      "kl": 0.34765625,
      "learning_rate": 2.510460251046025e-07,
      "loss": 0.0441,
      "reward": 0.7224670052528381,
      "reward_std": 0.11247023940086365,
      "reward_std_all": 0.11214932799339294,
      "rewards-max/forensics_sinkhorn_reward": 0.9128738939762115,
      "rewards-max/forensics_strict_edge_reward": 0.13721626518054109,
      "rewards-max/forensics_tiou_reward": 0.7968700230121613,
      "rewards/forensics_sinkhorn_reward": 0.6546131670475006,
      "rewards/forensics_strict_edge_reward": 0.031289934291862664,
      "rewards/forensics_tiou_reward": 0.4239110201597214,
      "step": 717
    },
    {
      "completion_length": 20.40625,
      "epoch": 3.00418410041841,
      "grad_norm": 7.367331027984619,
      "kl": 0.39453125,
      "learning_rate": 2.5e-07,
      "loss": 0.1192,
      "reward": 0.7605308890342712,
      "reward_std": 0.05378176108933985,
      "reward_std_all": 0.051916792057454586,
      "rewards-max/forensics_sinkhorn_reward": 0.937309741973877,
      "rewards-max/forensics_strict_edge_reward": 0.1787973027676344,
      "rewards-max/forensics_tiou_reward": 0.8671968281269073,
      "rewards/forensics_sinkhorn_reward": 0.6923734843730927,
      "rewards/forensics_strict_edge_reward": 0.021847677184268832,
      "rewards/forensics_tiou_reward": 0.4400465786457062,
      "step": 718
    },
    {
      "completion_length": 20.671875,
      "epoch": 3.00836820083682,
      "grad_norm": 5.905706882476807,
      "kl": 0.421875,
      "learning_rate": 2.489539748953975e-07,
      "loss": 0.015,
      "reward": 0.7257315814495087,
      "reward_std": 0.027719469740986824,
      "reward_std_all": 0.056930061429739,
      "rewards-max/forensics_sinkhorn_reward": 0.9742434322834015,
      "rewards-max/forensics_strict_edge_reward": 0.42794930934906006,
      "rewards-max/forensics_tiou_reward": 0.9582347571849823,
      "rewards/forensics_sinkhorn_reward": 0.7142667174339294,
      "rewards/forensics_strict_edge_reward": 0.05042589642107487,
      "rewards/forensics_tiou_reward": 0.543819010257721,
      "step": 719
    },
    {
      "completion_length": 20.109375,
      "epoch": 3.01255230125523,
      "grad_norm": 8.714695930480957,
      "kl": 0.3330078125,
      "learning_rate": 2.4790794979079496e-07,
      "loss": -0.174,
      "reward": 0.8055227100849152,
      "reward_std": 0.0017649949295446277,
      "reward_std_all": 0.00349819241091609,
      "rewards-max/forensics_sinkhorn_reward": 0.9812089800834656,
      "rewards-max/forensics_strict_edge_reward": 0.6816803216934204,
      "rewards-max/forensics_tiou_reward": 0.9801709651947021,
      "rewards/forensics_sinkhorn_reward": 0.8450483977794647,
      "rewards/forensics_strict_edge_reward": 0.0746698584407568,
      "rewards/forensics_tiou_reward": 0.6589147448539734,
      "step": 720
    },
    {
      "completion_length": 21.140625,
      "epoch": 3.01673640167364,
      "grad_norm": 5.840479850769043,
      "kl": 0.404296875,
      "learning_rate": 2.468619246861925e-07,
      "loss": -0.0548,
      "reward": 0.7448194921016693,
      "reward_std": 0.027067395858466625,
      "reward_std_all": 0.04286180064082146,
      "rewards-max/forensics_sinkhorn_reward": 0.9684319496154785,
      "rewards-max/forensics_strict_edge_reward": 0.259224871173501,
      "rewards-max/forensics_tiou_reward": 0.9264281690120697,
      "rewards/forensics_sinkhorn_reward": 0.6980473101139069,
      "rewards/forensics_strict_edge_reward": 0.012394146004226059,
      "rewards/forensics_tiou_reward": 0.4661649018526077,
      "step": 721
    },
    {
      "completion_length": 20.546875,
      "epoch": 3.02092050209205,
      "grad_norm": 6.404719352722168,
      "kl": 0.4140625,
      "learning_rate": 2.4581589958158993e-07,
      "loss": 0.0431,
      "reward": 0.7802045047283173,
      "reward_std": 0.040515962056815624,
      "reward_std_all": 0.03732144646346569,
      "rewards-max/forensics_sinkhorn_reward": 0.9795141518115997,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9740491211414337,
      "rewards/forensics_sinkhorn_reward": 0.7841263711452484,
      "rewards/forensics_strict_edge_reward": 0.057413212954998016,
      "rewards/forensics_tiou_reward": 0.6338179111480713,
      "step": 722
    },
    {
      "completion_length": 21.28125,
      "epoch": 3.0251046025104604,
      "grad_norm": 5.051059722900391,
      "kl": 0.392578125,
      "learning_rate": 2.4476987447698745e-07,
      "loss": 0.028,
      "reward": 0.7710379958152771,
      "reward_std": 0.06081956624984741,
      "reward_std_all": 0.054405149072408676,
      "rewards-max/forensics_sinkhorn_reward": 0.9834331572055817,
      "rewards-max/forensics_strict_edge_reward": 0.6548857688903809,
      "rewards-max/forensics_tiou_reward": 0.9818161427974701,
      "rewards/forensics_sinkhorn_reward": 0.7842771410942078,
      "rewards/forensics_strict_edge_reward": 0.06309554725885391,
      "rewards/forensics_tiou_reward": 0.6145294606685638,
      "step": 723
    },
    {
      "completion_length": 20.34375,
      "epoch": 3.0292887029288704,
      "grad_norm": 6.971574306488037,
      "kl": 0.4501953125,
      "learning_rate": 2.4372384937238496e-07,
      "loss": -0.0239,
      "reward": 0.775736927986145,
      "reward_std": 0.031196744181215763,
      "reward_std_all": 0.032026070170104504,
      "rewards-max/forensics_sinkhorn_reward": 0.9775200486183167,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9663048088550568,
      "rewards/forensics_sinkhorn_reward": 0.7555276155471802,
      "rewards/forensics_strict_edge_reward": 0.07770400308072567,
      "rewards/forensics_tiou_reward": 0.5292686522006989,
      "step": 724
    },
    {
      "completion_length": 22.203125,
      "epoch": 3.0334728033472804,
      "grad_norm": 6.879291534423828,
      "kl": 0.412109375,
      "learning_rate": 2.426778242677824e-07,
      "loss": -0.0852,
      "reward": 0.7930493950843811,
      "reward_std": 0.02163117495365441,
      "reward_std_all": 0.020836890675127506,
      "rewards-max/forensics_sinkhorn_reward": 0.9755329787731171,
      "rewards-max/forensics_strict_edge_reward": 0.4645801782608032,
      "rewards-max/forensics_tiou_reward": 0.9594547152519226,
      "rewards/forensics_sinkhorn_reward": 0.8100157678127289,
      "rewards/forensics_strict_edge_reward": 0.055075500160455704,
      "rewards/forensics_tiou_reward": 0.6384115815162659,
      "step": 725
    },
    {
      "completion_length": 22.359375,
      "epoch": 3.0376569037656904,
      "grad_norm": 7.568723201751709,
      "kl": 0.349609375,
      "learning_rate": 2.4163179916317993e-07,
      "loss": 0.0531,
      "reward": 0.7201474905014038,
      "reward_std": 0.07689451798796654,
      "reward_std_all": 0.0793684758245945,
      "rewards-max/forensics_sinkhorn_reward": 0.9698738753795624,
      "rewards-max/forensics_strict_edge_reward": 0.48844531178474426,
      "rewards-max/forensics_tiou_reward": 0.9541395306587219,
      "rewards/forensics_sinkhorn_reward": 0.6781178116798401,
      "rewards/forensics_strict_edge_reward": 0.03878462314605713,
      "rewards/forensics_tiou_reward": 0.4760560244321823,
      "step": 726
    },
    {
      "completion_length": 20.96875,
      "epoch": 3.0418410041841004,
      "grad_norm": 5.997412204742432,
      "kl": 0.345703125,
      "learning_rate": 2.405857740585774e-07,
      "loss": -0.0171,
      "reward": 0.7925541400909424,
      "reward_std": 0.014953725971281528,
      "reward_std_all": 0.016450892202556133,
      "rewards-max/forensics_sinkhorn_reward": 0.9742212295532227,
      "rewards-max/forensics_strict_edge_reward": 0.6365986764431,
      "rewards-max/forensics_tiou_reward": 0.9625762104988098,
      "rewards/forensics_sinkhorn_reward": 0.7788552045822144,
      "rewards/forensics_strict_edge_reward": 0.044347938150167465,
      "rewards/forensics_tiou_reward": 0.5571343302726746,
      "step": 727
    },
    {
      "completion_length": 20.453125,
      "epoch": 3.0460251046025104,
      "grad_norm": 13.123812675476074,
      "kl": 0.8935546875,
      "learning_rate": 2.395397489539749e-07,
      "loss": 0.1108,
      "reward": 0.7336853444576263,
      "reward_std": 0.06112259562360123,
      "reward_std_all": 0.06702487356960773,
      "rewards-max/forensics_sinkhorn_reward": 0.9858705401420593,
      "rewards-max/forensics_strict_edge_reward": 0.2862449362874031,
      "rewards-max/forensics_tiou_reward": 0.9880664050579071,
      "rewards/forensics_sinkhorn_reward": 0.7216561436653137,
      "rewards/forensics_strict_edge_reward": 0.02932615391910076,
      "rewards/forensics_tiou_reward": 0.5083530843257904,
      "step": 728
    },
    {
      "completion_length": 19.734375,
      "epoch": 3.0502092050209204,
      "grad_norm": 8.895817756652832,
      "kl": 0.3642578125,
      "learning_rate": 2.3849372384937237e-07,
      "loss": -0.0797,
      "reward": 0.7762610018253326,
      "reward_std": 0.02587074227631092,
      "reward_std_all": 0.02838142216205597,
      "rewards-max/forensics_sinkhorn_reward": 0.9781273007392883,
      "rewards-max/forensics_strict_edge_reward": 0.5566954910755157,
      "rewards-max/forensics_tiou_reward": 0.969586193561554,
      "rewards/forensics_sinkhorn_reward": 0.7652328610420227,
      "rewards/forensics_strict_edge_reward": 0.02708104532212019,
      "rewards/forensics_tiou_reward": 0.5190754234790802,
      "step": 729
    },
    {
      "completion_length": 22.5625,
      "epoch": 3.0543933054393304,
      "grad_norm": 7.117015838623047,
      "kl": 0.416015625,
      "learning_rate": 2.3744769874476985e-07,
      "loss": 0.0043,
      "reward": 0.7663634419441223,
      "reward_std": 0.042546749114990234,
      "reward_std_all": 0.043433768674731255,
      "rewards-max/forensics_sinkhorn_reward": 0.9758436977863312,
      "rewards-max/forensics_strict_edge_reward": 0.36915770173072815,
      "rewards-max/forensics_tiou_reward": 0.9655137360095978,
      "rewards/forensics_sinkhorn_reward": 0.7164759635925293,
      "rewards/forensics_strict_edge_reward": 0.026394523680210114,
      "rewards/forensics_tiou_reward": 0.45975834131240845,
      "step": 730
    },
    {
      "completion_length": 23.078125,
      "epoch": 3.0585774058577404,
      "grad_norm": 8.104706764221191,
      "kl": 0.5654296875,
      "learning_rate": 2.3640167364016736e-07,
      "loss": -0.0644,
      "reward": 0.7966207563877106,
      "reward_std": 0.016221519617829472,
      "reward_std_all": 0.016084517585113645,
      "rewards-max/forensics_sinkhorn_reward": 0.9807088375091553,
      "rewards-max/forensics_strict_edge_reward": 0.8187680542469025,
      "rewards-max/forensics_tiou_reward": 0.9794958531856537,
      "rewards/forensics_sinkhorn_reward": 0.7919489741325378,
      "rewards/forensics_strict_edge_reward": 0.07918673940002918,
      "rewards/forensics_tiou_reward": 0.618013471364975,
      "step": 731
    },
    {
      "completion_length": 19.6875,
      "epoch": 3.062761506276151,
      "grad_norm": 6.820475101470947,
      "kl": 0.49609375,
      "learning_rate": 2.3535564853556482e-07,
      "loss": -0.0506,
      "reward": 0.7972164750099182,
      "reward_std": 0.015692186541855335,
      "reward_std_all": 0.015470186714082956,
      "rewards-max/forensics_sinkhorn_reward": 0.9777187407016754,
      "rewards-max/forensics_strict_edge_reward": 0.5995397567749023,
      "rewards-max/forensics_tiou_reward": 0.9665556252002716,
      "rewards/forensics_sinkhorn_reward": 0.8114650249481201,
      "rewards/forensics_strict_edge_reward": 0.06834918446838856,
      "rewards/forensics_tiou_reward": 0.6375554203987122,
      "step": 732
    },
    {
      "completion_length": 21.421875,
      "epoch": 3.066945606694561,
      "grad_norm": 6.1349778175354,
      "kl": 0.4052734375,
      "learning_rate": 2.3430962343096234e-07,
      "loss": -0.0045,
      "reward": 0.782696396112442,
      "reward_std": 0.005378989037126303,
      "reward_std_all": 0.016532053239643574,
      "rewards-max/forensics_sinkhorn_reward": 0.9726689755916595,
      "rewards-max/forensics_strict_edge_reward": 0.5294047147035599,
      "rewards-max/forensics_tiou_reward": 0.9567462801933289,
      "rewards/forensics_sinkhorn_reward": 0.7193971574306488,
      "rewards/forensics_strict_edge_reward": 0.051233191043138504,
      "rewards/forensics_tiou_reward": 0.4977319538593292,
      "step": 733
    },
    {
      "completion_length": 20.734375,
      "epoch": 3.071129707112971,
      "grad_norm": 5.438304901123047,
      "kl": 0.3720703125,
      "learning_rate": 2.3326359832635982e-07,
      "loss": 0.074,
      "reward": 0.772230714559555,
      "reward_std": 0.0500246062874794,
      "reward_std_all": 0.04651273414492607,
      "rewards-max/forensics_sinkhorn_reward": 0.9834835827350616,
      "rewards-max/forensics_strict_edge_reward": 0.37596484273672104,
      "rewards-max/forensics_tiou_reward": 0.9815758168697357,
      "rewards/forensics_sinkhorn_reward": 0.7645100057125092,
      "rewards/forensics_strict_edge_reward": 0.017662787344306707,
      "rewards/forensics_tiou_reward": 0.5573413670063019,
      "step": 734
    },
    {
      "completion_length": 20.90625,
      "epoch": 3.075313807531381,
      "grad_norm": 6.26978874206543,
      "kl": 0.412109375,
      "learning_rate": 2.322175732217573e-07,
      "loss": -0.067,
      "reward": 0.7903612852096558,
      "reward_std": 0.024749090895056725,
      "reward_std_all": 0.023773313965648413,
      "rewards-max/forensics_sinkhorn_reward": 0.9844125509262085,
      "rewards-max/forensics_strict_edge_reward": 0.5072369128465652,
      "rewards-max/forensics_tiou_reward": 0.9834838807582855,
      "rewards/forensics_sinkhorn_reward": 0.7888282835483551,
      "rewards/forensics_strict_edge_reward": 0.03199051879346371,
      "rewards/forensics_tiou_reward": 0.5817357003688812,
      "step": 735
    },
    {
      "completion_length": 19.734375,
      "epoch": 3.079497907949791,
      "grad_norm": 7.335554599761963,
      "kl": 0.654296875,
      "learning_rate": 2.311715481171548e-07,
      "loss": -0.0293,
      "reward": 0.7884289026260376,
      "reward_std": 0.024549636989831924,
      "reward_std_all": 0.02410469390451908,
      "rewards-max/forensics_sinkhorn_reward": 0.974475234746933,
      "rewards-max/forensics_strict_edge_reward": 0.46758662164211273,
      "rewards-max/forensics_tiou_reward": 0.9656545221805573,
      "rewards/forensics_sinkhorn_reward": 0.7844330370426178,
      "rewards/forensics_strict_edge_reward": 0.05962786078453064,
      "rewards/forensics_tiou_reward": 0.5935205072164536,
      "step": 736
    },
    {
      "completion_length": 21.875,
      "epoch": 3.083682008368201,
      "grad_norm": 5.695874214172363,
      "kl": 0.416015625,
      "learning_rate": 2.3012552301255228e-07,
      "loss": 0.0824,
      "reward": 0.7854725420475006,
      "reward_std": 0.03291842993348837,
      "reward_std_all": 0.03067940939217806,
      "rewards-max/forensics_sinkhorn_reward": 0.9791520237922668,
      "rewards-max/forensics_strict_edge_reward": 0.6188668608665466,
      "rewards-max/forensics_tiou_reward": 0.9752054512500763,
      "rewards/forensics_sinkhorn_reward": 0.7875953614711761,
      "rewards/forensics_strict_edge_reward": 0.08228093013167381,
      "rewards/forensics_tiou_reward": 0.625724583864212,
      "step": 737
    },
    {
      "completion_length": 20.96875,
      "epoch": 3.087866108786611,
      "grad_norm": 7.512513160705566,
      "kl": 0.390625,
      "learning_rate": 2.2907949790794977e-07,
      "loss": -0.0173,
      "reward": 0.77351313829422,
      "reward_std": 0.04987902380526066,
      "reward_std_all": 0.04606210999190807,
      "rewards-max/forensics_sinkhorn_reward": 0.9788105487823486,
      "rewards-max/forensics_strict_edge_reward": 0.5984183251857758,
      "rewards-max/forensics_tiou_reward": 0.9757663011550903,
      "rewards/forensics_sinkhorn_reward": 0.7454746961593628,
      "rewards/forensics_strict_edge_reward": 0.0644181240350008,
      "rewards/forensics_tiou_reward": 0.5035071969032288,
      "step": 738
    },
    {
      "completion_length": 21.015625,
      "epoch": 3.092050209205021,
      "grad_norm": 6.618000030517578,
      "kl": 0.4091796875,
      "learning_rate": 2.2803347280334728e-07,
      "loss": 0.0207,
      "reward": 0.75190070271492,
      "reward_std": 0.05309166945517063,
      "reward_std_all": 0.053472019731998444,
      "rewards-max/forensics_sinkhorn_reward": 0.9705925583839417,
      "rewards-max/forensics_strict_edge_reward": 0.4067516513168812,
      "rewards-max/forensics_tiou_reward": 0.9519627392292023,
      "rewards/forensics_sinkhorn_reward": 0.705877959728241,
      "rewards/forensics_strict_edge_reward": 0.0338589265011251,
      "rewards/forensics_tiou_reward": 0.530381590127945,
      "step": 739
    },
    {
      "completion_length": 22.359375,
      "epoch": 3.096234309623431,
      "grad_norm": 5.345078945159912,
      "kl": 0.2978515625,
      "learning_rate": 2.2698744769874474e-07,
      "loss": 0.0562,
      "reward": 0.7873150110244751,
      "reward_std": 0.028584999963641167,
      "reward_std_all": 0.027288970537483692,
      "rewards-max/forensics_sinkhorn_reward": 0.9886473417282104,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9988400042057037,
      "rewards/forensics_sinkhorn_reward": 0.7747530341148376,
      "rewards/forensics_strict_edge_reward": 0.07764005288481712,
      "rewards/forensics_tiou_reward": 0.6377155482769012,
      "step": 740
    },
    {
      "completion_length": 21.9375,
      "epoch": 3.100418410041841,
      "grad_norm": 5.215949058532715,
      "kl": 0.349609375,
      "learning_rate": 2.2594142259414225e-07,
      "loss": 0.0921,
      "reward": 0.7745077311992645,
      "reward_std": 0.04405122064054012,
      "reward_std_all": 0.041417188942432404,
      "rewards-max/forensics_sinkhorn_reward": 0.9749042987823486,
      "rewards-max/forensics_strict_edge_reward": 0.4697008728981018,
      "rewards-max/forensics_tiou_reward": 0.9621991515159607,
      "rewards/forensics_sinkhorn_reward": 0.7501508891582489,
      "rewards/forensics_strict_edge_reward": 0.03850653022527695,
      "rewards/forensics_tiou_reward": 0.5631610453128815,
      "step": 741
    },
    {
      "completion_length": 21.5,
      "epoch": 3.104602510460251,
      "grad_norm": 6.863692283630371,
      "kl": 0.3798828125,
      "learning_rate": 2.2489539748953974e-07,
      "loss": 0.0927,
      "reward": 0.7292253375053406,
      "reward_std": 0.08492155186831951,
      "reward_std_all": 0.08622204884886742,
      "rewards-max/forensics_sinkhorn_reward": 0.9575952887535095,
      "rewards-max/forensics_strict_edge_reward": 0.5424859970808029,
      "rewards-max/forensics_tiou_reward": 0.9295748174190521,
      "rewards/forensics_sinkhorn_reward": 0.6907134056091309,
      "rewards/forensics_strict_edge_reward": 0.037962350994348526,
      "rewards/forensics_tiou_reward": 0.5229480564594269,
      "step": 742
    },
    {
      "completion_length": 23.21875,
      "epoch": 3.1087866108786613,
      "grad_norm": 6.7414093017578125,
      "kl": 0.345703125,
      "learning_rate": 2.2384937238493723e-07,
      "loss": -0.0159,
      "reward": 0.7272591292858124,
      "reward_std": 0.04416786739602685,
      "reward_std_all": 0.059539093635976315,
      "rewards-max/forensics_sinkhorn_reward": 0.9536026120185852,
      "rewards-max/forensics_strict_edge_reward": 0.305109478533268,
      "rewards-max/forensics_tiou_reward": 0.9072185754776001,
      "rewards/forensics_sinkhorn_reward": 0.6807260811328888,
      "rewards/forensics_strict_edge_reward": 0.022625823505222797,
      "rewards/forensics_tiou_reward": 0.4289417862892151,
      "step": 743
    },
    {
      "completion_length": 20.4375,
      "epoch": 3.1129707112970713,
      "grad_norm": 6.553038597106934,
      "kl": 0.2763671875,
      "learning_rate": 2.2280334728033474e-07,
      "loss": -0.0209,
      "reward": 0.7616872191429138,
      "reward_std": 0.017998868599534035,
      "reward_std_all": 0.032658107578754425,
      "rewards-max/forensics_sinkhorn_reward": 0.9812500476837158,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9803272485733032,
      "rewards/forensics_sinkhorn_reward": 0.7221789956092834,
      "rewards/forensics_strict_edge_reward": 0.06472263857722282,
      "rewards/forensics_tiou_reward": 0.5188865065574646,
      "step": 744
    },
    {
      "completion_length": 20.234375,
      "epoch": 3.1171548117154813,
      "grad_norm": 7.620046138763428,
      "kl": 0.34375,
      "learning_rate": 2.217573221757322e-07,
      "loss": -0.0248,
      "reward": 0.7812020182609558,
      "reward_std": 0.032850589603185654,
      "reward_std_all": 0.03171803615987301,
      "rewards-max/forensics_sinkhorn_reward": 0.9693970382213593,
      "rewards-max/forensics_strict_edge_reward": 0.3456989713013172,
      "rewards-max/forensics_tiou_reward": 0.9464252889156342,
      "rewards/forensics_sinkhorn_reward": 0.7943995296955109,
      "rewards/forensics_strict_edge_reward": 0.04669771855697036,
      "rewards/forensics_tiou_reward": 0.6137807667255402,
      "step": 745
    },
    {
      "completion_length": 21.515625,
      "epoch": 3.1213389121338913,
      "grad_norm": 5.882922649383545,
      "kl": 0.3974609375,
      "learning_rate": 2.207112970711297e-07,
      "loss": -0.0865,
      "reward": 0.7628433704376221,
      "reward_std": 0.045935457572340965,
      "reward_std_all": 0.045575063675642014,
      "rewards-max/forensics_sinkhorn_reward": 0.9713696837425232,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9547494053840637,
      "rewards/forensics_sinkhorn_reward": 0.7215452790260315,
      "rewards/forensics_strict_edge_reward": 0.061192531138658524,
      "rewards/forensics_tiou_reward": 0.4936218857765198,
      "step": 746
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.1255230125523012,
      "grad_norm": 6.520961761474609,
      "kl": 0.5380859375,
      "learning_rate": 2.1966527196652717e-07,
      "loss": 0.1459,
      "reward": 0.7319231927394867,
      "reward_std": 0.0849575325846672,
      "reward_std_all": 0.08157774806022644,
      "rewards-max/forensics_sinkhorn_reward": 0.955874502658844,
      "rewards-max/forensics_strict_edge_reward": 0.17881613411009312,
      "rewards-max/forensics_tiou_reward": 0.9059584140777588,
      "rewards/forensics_sinkhorn_reward": 0.621989518404007,
      "rewards/forensics_strict_edge_reward": 0.012913767714053392,
      "rewards/forensics_tiou_reward": 0.44714514911174774,
      "step": 747
    },
    {
      "completion_length": 21.984375,
      "epoch": 3.1297071129707112,
      "grad_norm": 6.275663375854492,
      "kl": 0.3642578125,
      "learning_rate": 2.1861924686192469e-07,
      "loss": 0.0413,
      "reward": 0.7518697381019592,
      "reward_std": 0.020984958857297897,
      "reward_std_all": 0.03757043834775686,
      "rewards-max/forensics_sinkhorn_reward": 0.979464590549469,
      "rewards-max/forensics_strict_edge_reward": 0.3095986396074295,
      "rewards-max/forensics_tiou_reward": 0.9725522398948669,
      "rewards/forensics_sinkhorn_reward": 0.6914756298065186,
      "rewards/forensics_strict_edge_reward": 0.030614529736340046,
      "rewards/forensics_tiou_reward": 0.5378432869911194,
      "step": 748
    },
    {
      "completion_length": 21.71875,
      "epoch": 3.1338912133891212,
      "grad_norm": 6.788873195648193,
      "kl": 0.4150390625,
      "learning_rate": 2.1757322175732217e-07,
      "loss": 0.0207,
      "reward": 0.7764644622802734,
      "reward_std": 0.031111798947677016,
      "reward_std_all": 0.031622710172086954,
      "rewards-max/forensics_sinkhorn_reward": 0.9442318677902222,
      "rewards-max/forensics_strict_edge_reward": 0.36915770173072815,
      "rewards-max/forensics_tiou_reward": 0.8953251540660858,
      "rewards/forensics_sinkhorn_reward": 0.7468056082725525,
      "rewards/forensics_strict_edge_reward": 0.0268399971537292,
      "rewards/forensics_tiou_reward": 0.5205926299095154,
      "step": 749
    },
    {
      "completion_length": 23.09375,
      "epoch": 3.1380753138075312,
      "grad_norm": 7.1027512550354,
      "kl": 0.4296875,
      "learning_rate": 2.1652719665271966e-07,
      "loss": 0.0782,
      "reward": 0.7490617334842682,
      "reward_std": 0.04408632777631283,
      "reward_std_all": 0.05028895474970341,
      "rewards-max/forensics_sinkhorn_reward": 0.9805770516395569,
      "rewards-max/forensics_strict_edge_reward": 0.5548860281705856,
      "rewards-max/forensics_tiou_reward": 0.9723809361457825,
      "rewards/forensics_sinkhorn_reward": 0.7103621959686279,
      "rewards/forensics_strict_edge_reward": 0.04800752829760313,
      "rewards/forensics_tiou_reward": 0.5443159341812134,
      "step": 750
    },
    {
      "completion_length": 19.15625,
      "epoch": 3.1422594142259412,
      "grad_norm": 6.294376850128174,
      "kl": 0.4462890625,
      "learning_rate": 2.1548117154811714e-07,
      "loss": -0.0341,
      "reward": 0.7723839581012726,
      "reward_std": 0.0013328043860383332,
      "reward_std_all": 0.019863904220983386,
      "rewards-max/forensics_sinkhorn_reward": 0.9669330716133118,
      "rewards-max/forensics_strict_edge_reward": 0.5020816326141357,
      "rewards-max/forensics_tiou_reward": 0.9461461901664734,
      "rewards/forensics_sinkhorn_reward": 0.8011967837810516,
      "rewards/forensics_strict_edge_reward": 0.044883265160024166,
      "rewards/forensics_tiou_reward": 0.6214067041873932,
      "step": 751
    },
    {
      "completion_length": 19.96875,
      "epoch": 3.1464435146443517,
      "grad_norm": 7.761989116668701,
      "kl": 0.3037109375,
      "learning_rate": 2.1443514644351463e-07,
      "loss": -0.0809,
      "reward": 0.7653100788593292,
      "reward_std": 0.023814517073333263,
      "reward_std_all": 0.030937161296606064,
      "rewards-max/forensics_sinkhorn_reward": 0.9814087450504303,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.978266566991806,
      "rewards/forensics_sinkhorn_reward": 0.7728846669197083,
      "rewards/forensics_strict_edge_reward": 0.04397632647305727,
      "rewards/forensics_tiou_reward": 0.5639492869377136,
      "step": 752
    },
    {
      "completion_length": 20.96875,
      "epoch": 3.1506276150627617,
      "grad_norm": 6.514647960662842,
      "kl": 0.439453125,
      "learning_rate": 2.1338912133891212e-07,
      "loss": -0.0375,
      "reward": 0.7876312136650085,
      "reward_std": 0.024426008574664593,
      "reward_std_all": 0.024256561882793903,
      "rewards-max/forensics_sinkhorn_reward": 0.9806969165802002,
      "rewards-max/forensics_strict_edge_reward": 0.57895427942276,
      "rewards-max/forensics_tiou_reward": 0.977313369512558,
      "rewards/forensics_sinkhorn_reward": 0.8208247423171997,
      "rewards/forensics_strict_edge_reward": 0.08685570955276489,
      "rewards/forensics_tiou_reward": 0.6203001737594604,
      "step": 753
    },
    {
      "completion_length": 22.484375,
      "epoch": 3.1548117154811717,
      "grad_norm": 6.288075923919678,
      "kl": 0.4208984375,
      "learning_rate": 2.1234309623430963e-07,
      "loss": 0.1431,
      "reward": 0.7690431773662567,
      "reward_std": 0.04839985631406307,
      "reward_std_all": 0.04591694287955761,
      "rewards-max/forensics_sinkhorn_reward": 0.9804314374923706,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.977865993976593,
      "rewards/forensics_sinkhorn_reward": 0.7095660865306854,
      "rewards/forensics_strict_edge_reward": 0.05249273218214512,
      "rewards/forensics_tiou_reward": 0.5328131914138794,
      "step": 754
    },
    {
      "completion_length": 20.578125,
      "epoch": 3.1589958158995817,
      "grad_norm": 5.895881175994873,
      "kl": 0.4130859375,
      "learning_rate": 2.112970711297071e-07,
      "loss": 0.0074,
      "reward": 0.7926619350910187,
      "reward_std": 0.022735384292900562,
      "reward_std_all": 0.02201634831726551,
      "rewards-max/forensics_sinkhorn_reward": 0.97684046626091,
      "rewards-max/forensics_strict_edge_reward": 0.24938895553350449,
      "rewards-max/forensics_tiou_reward": 0.9657919108867645,
      "rewards/forensics_sinkhorn_reward": 0.805549293756485,
      "rewards/forensics_strict_edge_reward": 0.03810679726302624,
      "rewards/forensics_tiou_reward": 0.6299921572208405,
      "step": 755
    },
    {
      "completion_length": 22.265625,
      "epoch": 3.1631799163179917,
      "grad_norm": 7.204300403594971,
      "kl": 0.380859375,
      "learning_rate": 2.102510460251046e-07,
      "loss": 0.0429,
      "reward": 0.7567355334758759,
      "reward_std": 0.035855150781571865,
      "reward_std_all": 0.0451029185205698,
      "rewards-max/forensics_sinkhorn_reward": 0.8965078890323639,
      "rewards-max/forensics_strict_edge_reward": 0.01552046649158001,
      "rewards-max/forensics_tiou_reward": 0.7881697714328766,
      "rewards/forensics_sinkhorn_reward": 0.6817869544029236,
      "rewards/forensics_strict_edge_reward": 0.0017068099114112556,
      "rewards/forensics_tiou_reward": 0.4400555044412613,
      "step": 756
    },
    {
      "completion_length": 21.609375,
      "epoch": 3.1673640167364017,
      "grad_norm": 7.993045806884766,
      "kl": 0.3330078125,
      "learning_rate": 2.0920502092050206e-07,
      "loss": 0.0004,
      "reward": 0.7809162139892578,
      "reward_std": 0.03240752685815096,
      "reward_std_all": 0.03154588770121336,
      "rewards-max/forensics_sinkhorn_reward": 0.9827595949172974,
      "rewards-max/forensics_strict_edge_reward": 0.6119300425052643,
      "rewards-max/forensics_tiou_reward": 0.981579452753067,
      "rewards/forensics_sinkhorn_reward": 0.7630977630615234,
      "rewards/forensics_strict_edge_reward": 0.077524583786726,
      "rewards/forensics_tiou_reward": 0.5942295491695404,
      "step": 757
    },
    {
      "completion_length": 21.8125,
      "epoch": 3.1715481171548117,
      "grad_norm": 5.952986240386963,
      "kl": 0.4306640625,
      "learning_rate": 2.0815899581589958e-07,
      "loss": -0.0057,
      "reward": 0.746954470872879,
      "reward_std": 0.06082524172961712,
      "reward_std_all": 0.06131654232740402,
      "rewards-max/forensics_sinkhorn_reward": 0.9804389178752899,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9790741801261902,
      "rewards/forensics_sinkhorn_reward": 0.7342014610767365,
      "rewards/forensics_strict_edge_reward": 0.10678253322839737,
      "rewards/forensics_tiou_reward": 0.6083865463733673,
      "step": 758
    },
    {
      "completion_length": 20.78125,
      "epoch": 3.1757322175732217,
      "grad_norm": 6.947588920593262,
      "kl": 0.4970703125,
      "learning_rate": 2.0711297071129706e-07,
      "loss": 0.0754,
      "reward": 0.776480495929718,
      "reward_std": 0.02034006267786026,
      "reward_std_all": 0.0271413279697299,
      "rewards-max/forensics_sinkhorn_reward": 0.9691048860549927,
      "rewards-max/forensics_strict_edge_reward": 0.47726282477378845,
      "rewards-max/forensics_tiou_reward": 0.949832558631897,
      "rewards/forensics_sinkhorn_reward": 0.71661576628685,
      "rewards/forensics_strict_edge_reward": 0.033167785964906216,
      "rewards/forensics_tiou_reward": 0.46970607340335846,
      "step": 759
    },
    {
      "completion_length": 22.84375,
      "epoch": 3.1799163179916317,
      "grad_norm": 6.048208236694336,
      "kl": 0.361328125,
      "learning_rate": 2.0606694560669455e-07,
      "loss": -0.0263,
      "reward": 0.7551026046276093,
      "reward_std": 0.04808294586837292,
      "reward_std_all": 0.05144146457314491,
      "rewards-max/forensics_sinkhorn_reward": 0.9721102714538574,
      "rewards-max/forensics_strict_edge_reward": 0.6135751008987427,
      "rewards-max/forensics_tiou_reward": 0.9592106938362122,
      "rewards/forensics_sinkhorn_reward": 0.7071130871772766,
      "rewards/forensics_strict_edge_reward": 0.05374146066606045,
      "rewards/forensics_tiou_reward": 0.49650660157203674,
      "step": 760
    },
    {
      "completion_length": 20.703125,
      "epoch": 3.1841004184100417,
      "grad_norm": 6.915902137756348,
      "kl": 0.3291015625,
      "learning_rate": 2.0502092050209206e-07,
      "loss": 0.0878,
      "reward": 0.7619299292564392,
      "reward_std": 0.03485005907714367,
      "reward_std_all": 0.04195393808186054,
      "rewards-max/forensics_sinkhorn_reward": 0.9774017930030823,
      "rewards-max/forensics_strict_edge_reward": 0.5878151059150696,
      "rewards-max/forensics_tiou_reward": 0.9692191481590271,
      "rewards/forensics_sinkhorn_reward": 0.7083101570606232,
      "rewards/forensics_strict_edge_reward": 0.0647018700838089,
      "rewards/forensics_tiou_reward": 0.45030903816223145,
      "step": 761
    },
    {
      "completion_length": 22.296875,
      "epoch": 3.1882845188284517,
      "grad_norm": 5.90652322769165,
      "kl": 0.5419921875,
      "learning_rate": 2.0397489539748952e-07,
      "loss": 0.031,
      "reward": 0.7669480443000793,
      "reward_std": 0.043392036110162735,
      "reward_std_all": 0.04577906057238579,
      "rewards-max/forensics_sinkhorn_reward": 0.9758478105068207,
      "rewards-max/forensics_strict_edge_reward": 0.5905451774597168,
      "rewards-max/forensics_tiou_reward": 0.9691028296947479,
      "rewards/forensics_sinkhorn_reward": 0.7727335095405579,
      "rewards/forensics_strict_edge_reward": 0.06279039569199085,
      "rewards/forensics_tiou_reward": 0.5927856266498566,
      "step": 762
    },
    {
      "completion_length": 19.34375,
      "epoch": 3.192468619246862,
      "grad_norm": 7.666386604309082,
      "kl": 0.384765625,
      "learning_rate": 2.0292887029288703e-07,
      "loss": -0.0314,
      "reward": 0.7941925823688507,
      "reward_std": 0.007090492406859994,
      "reward_std_all": 0.01173366466537118,
      "rewards-max/forensics_sinkhorn_reward": 0.9790831506252289,
      "rewards-max/forensics_strict_edge_reward": 0.7055691182613373,
      "rewards-max/forensics_tiou_reward": 0.9753408133983612,
      "rewards/forensics_sinkhorn_reward": 0.7670564651489258,
      "rewards/forensics_strict_edge_reward": 0.07740404736250639,
      "rewards/forensics_tiou_reward": 0.502173662185669,
      "step": 763
    },
    {
      "completion_length": 21.125,
      "epoch": 3.196652719665272,
      "grad_norm": 8.764870643615723,
      "kl": 0.3203125,
      "learning_rate": 2.0188284518828452e-07,
      "loss": 0.0008,
      "reward": 0.766832560300827,
      "reward_std": 0.03492269944399595,
      "reward_std_all": 0.036732289008796215,
      "rewards-max/forensics_sinkhorn_reward": 0.9791339635848999,
      "rewards-max/forensics_strict_edge_reward": 0.3422802984714508,
      "rewards-max/forensics_tiou_reward": 0.9711081981658936,
      "rewards/forensics_sinkhorn_reward": 0.7347878813743591,
      "rewards/forensics_strict_edge_reward": 0.03176738880574703,
      "rewards/forensics_tiou_reward": 0.5470689833164215,
      "step": 764
    },
    {
      "completion_length": 21.390625,
      "epoch": 3.200836820083682,
      "grad_norm": 6.897782802581787,
      "kl": 0.3603515625,
      "learning_rate": 2.00836820083682e-07,
      "loss": 0.0241,
      "reward": 0.7826993763446808,
      "reward_std": 0.038072701543569565,
      "reward_std_all": 0.035327122546732426,
      "rewards-max/forensics_sinkhorn_reward": 0.9741043746471405,
      "rewards-max/forensics_strict_edge_reward": 0.6231168210506439,
      "rewards-max/forensics_tiou_reward": 0.959399938583374,
      "rewards/forensics_sinkhorn_reward": 0.7740559279918671,
      "rewards/forensics_strict_edge_reward": 0.055031128227710724,
      "rewards/forensics_tiou_reward": 0.5811223685741425,
      "step": 765
    },
    {
      "completion_length": 21.21875,
      "epoch": 3.205020920502092,
      "grad_norm": 6.938661098480225,
      "kl": 0.4248046875,
      "learning_rate": 1.997907949790795e-07,
      "loss": -0.0411,
      "reward": 0.7698054313659668,
      "reward_std": 0.037581710144877434,
      "reward_std_all": 0.04062322061508894,
      "rewards-max/forensics_sinkhorn_reward": 0.9328449368476868,
      "rewards-max/forensics_strict_edge_reward": 0.29517203150317073,
      "rewards-max/forensics_tiou_reward": 0.8394329249858856,
      "rewards/forensics_sinkhorn_reward": 0.7180553078651428,
      "rewards/forensics_strict_edge_reward": 0.018633425468578935,
      "rewards/forensics_tiou_reward": 0.4910683333873749,
      "step": 766
    },
    {
      "completion_length": 22.09375,
      "epoch": 3.209205020920502,
      "grad_norm": 7.422050476074219,
      "kl": 0.361328125,
      "learning_rate": 1.9874476987447698e-07,
      "loss": -0.0502,
      "reward": 0.7811101377010345,
      "reward_std": 0.025203879456967115,
      "reward_std_all": 0.02639410551637411,
      "rewards-max/forensics_sinkhorn_reward": 0.981786698102951,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.981061726808548,
      "rewards/forensics_sinkhorn_reward": 0.759021520614624,
      "rewards/forensics_strict_edge_reward": 0.0832257941365242,
      "rewards/forensics_tiou_reward": 0.5721327364444733,
      "step": 767
    },
    {
      "completion_length": 19.953125,
      "epoch": 3.213389121338912,
      "grad_norm": 7.502931118011475,
      "kl": 0.3681640625,
      "learning_rate": 1.9769874476987447e-07,
      "loss": 0.0984,
      "reward": 0.7619127333164215,
      "reward_std": 0.0544346384704113,
      "reward_std_all": 0.052109092473983765,
      "rewards-max/forensics_sinkhorn_reward": 0.9771654903888702,
      "rewards-max/forensics_strict_edge_reward": 0.4496074318885803,
      "rewards-max/forensics_tiou_reward": 0.956596314907074,
      "rewards/forensics_sinkhorn_reward": 0.748761922121048,
      "rewards/forensics_strict_edge_reward": 0.028165563941001892,
      "rewards/forensics_tiou_reward": 0.5512509644031525,
      "step": 768
    },
    {
      "completion_length": 20.40625,
      "epoch": 3.217573221757322,
      "grad_norm": 6.843645095825195,
      "kl": 0.4951171875,
      "learning_rate": 1.9665271966527198e-07,
      "loss": -0.0193,
      "reward": 0.7906454801559448,
      "reward_std": 0.025877241045236588,
      "reward_std_all": 0.024509551003575325,
      "rewards-max/forensics_sinkhorn_reward": 0.981868326663971,
      "rewards-max/forensics_strict_edge_reward": 0.7676310539245605,
      "rewards-max/forensics_tiou_reward": 0.9828798770904541,
      "rewards/forensics_sinkhorn_reward": 0.7649953961372375,
      "rewards/forensics_strict_edge_reward": 0.07090546749532223,
      "rewards/forensics_tiou_reward": 0.5576704442501068,
      "step": 769
    },
    {
      "completion_length": 23.78125,
      "epoch": 3.221757322175732,
      "grad_norm": 6.533547878265381,
      "kl": 0.37890625,
      "learning_rate": 1.9560669456066944e-07,
      "loss": 0.0135,
      "reward": 0.7425503432750702,
      "reward_std": 0.06114981323480606,
      "reward_std_all": 0.06252466514706612,
      "rewards-max/forensics_sinkhorn_reward": 0.978721410036087,
      "rewards-max/forensics_strict_edge_reward": 0.3816564530134201,
      "rewards-max/forensics_tiou_reward": 0.9721808135509491,
      "rewards/forensics_sinkhorn_reward": 0.7048723697662354,
      "rewards/forensics_strict_edge_reward": 0.026216271333396435,
      "rewards/forensics_tiou_reward": 0.5409983396530151,
      "step": 770
    },
    {
      "completion_length": 21.421875,
      "epoch": 3.225941422594142,
      "grad_norm": 7.061727523803711,
      "kl": 0.27880859375,
      "learning_rate": 1.9456066945606695e-07,
      "loss": -0.061,
      "reward": 0.7376858592033386,
      "reward_std": 0.042661565355956554,
      "reward_std_all": 0.05392925441265106,
      "rewards-max/forensics_sinkhorn_reward": 0.9854894578456879,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9910281300544739,
      "rewards/forensics_sinkhorn_reward": 0.7487214207649231,
      "rewards/forensics_strict_edge_reward": 0.10655540600419044,
      "rewards/forensics_tiou_reward": 0.5960623025894165,
      "step": 771
    },
    {
      "completion_length": 22.859375,
      "epoch": 3.2301255230125525,
      "grad_norm": 8.382287979125977,
      "kl": 0.37890625,
      "learning_rate": 1.935146443514644e-07,
      "loss": 0.0657,
      "reward": 0.7697811722755432,
      "reward_std": 0.05692438967525959,
      "reward_std_all": 0.051870270632207394,
      "rewards-max/forensics_sinkhorn_reward": 0.9817278385162354,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.9810631573200226,
      "rewards/forensics_sinkhorn_reward": 0.7427296936511993,
      "rewards/forensics_strict_edge_reward": 0.07005777768790722,
      "rewards/forensics_tiou_reward": 0.6105220019817352,
      "step": 772
    },
    {
      "completion_length": 21.546875,
      "epoch": 3.2343096234309625,
      "grad_norm": 7.461892127990723,
      "kl": 0.529296875,
      "learning_rate": 1.9246861924686192e-07,
      "loss": 0.0213,
      "reward": 0.7560569643974304,
      "reward_std": 0.06750261038541794,
      "reward_std_all": 0.06346071884036064,
      "rewards-max/forensics_sinkhorn_reward": 0.9705875813961029,
      "rewards-max/forensics_strict_edge_reward": 0.5599738955497742,
      "rewards-max/forensics_tiou_reward": 0.9585360586643219,
      "rewards/forensics_sinkhorn_reward": 0.741060197353363,
      "rewards/forensics_strict_edge_reward": 0.030518850311636925,
      "rewards/forensics_tiou_reward": 0.5262659192085266,
      "step": 773
    },
    {
      "completion_length": 20.71875,
      "epoch": 3.2384937238493725,
      "grad_norm": 5.97835636138916,
      "kl": 0.443359375,
      "learning_rate": 1.914225941422594e-07,
      "loss": -0.0135,
      "reward": 0.7879038751125336,
      "reward_std": 0.02907137107104063,
      "reward_std_all": 0.027449412271380424,
      "rewards-max/forensics_sinkhorn_reward": 0.9833619594573975,
      "rewards-max/forensics_strict_edge_reward": 0.5599624961614609,
      "rewards-max/forensics_tiou_reward": 0.9855401813983917,
      "rewards/forensics_sinkhorn_reward": 0.7855965793132782,
      "rewards/forensics_strict_edge_reward": 0.05143258906900883,
      "rewards/forensics_tiou_reward": 0.6365357637405396,
      "step": 774
    },
    {
      "completion_length": 20.703125,
      "epoch": 3.2426778242677825,
      "grad_norm": 7.278228759765625,
      "kl": 0.376953125,
      "learning_rate": 1.903765690376569e-07,
      "loss": -0.0166,
      "reward": 0.7820200324058533,
      "reward_std": 0.011906114756129682,
      "reward_std_all": 0.01799425238277763,
      "rewards-max/forensics_sinkhorn_reward": 0.9466989636421204,
      "rewards-max/forensics_strict_edge_reward": 0.47754265088588,
      "rewards-max/forensics_tiou_reward": 0.8928514122962952,
      "rewards/forensics_sinkhorn_reward": 0.7773694694042206,
      "rewards/forensics_strict_edge_reward": 0.06414238642901182,
      "rewards/forensics_tiou_reward": 0.5977602005004883,
      "step": 775
    },
    {
      "completion_length": 20.375,
      "epoch": 3.2468619246861925,
      "grad_norm": 7.450084209442139,
      "kl": 0.447265625,
      "learning_rate": 1.8933054393305438e-07,
      "loss": -0.0036,
      "reward": 0.7713479995727539,
      "reward_std": 0.059804767370224,
      "reward_std_all": 0.05336640030145645,
      "rewards-max/forensics_sinkhorn_reward": 0.9850050508975983,
      "rewards-max/forensics_strict_edge_reward": 0.7789089381694794,
      "rewards-max/forensics_tiou_reward": 0.9897593855857849,
      "rewards/forensics_sinkhorn_reward": 0.7991541922092438,
      "rewards/forensics_strict_edge_reward": 0.12704402208328247,
      "rewards/forensics_tiou_reward": 0.6527398228645325,
      "step": 776
    },
    {
      "completion_length": 20.546875,
      "epoch": 3.2510460251046025,
      "grad_norm": 8.654191017150879,
      "kl": 0.3642578125,
      "learning_rate": 1.8828451882845187e-07,
      "loss": 0.0276,
      "reward": 0.7795502245426178,
      "reward_std": 0.04604863375425339,
      "reward_std_all": 0.04185120388865471,
      "rewards-max/forensics_sinkhorn_reward": 0.9659708142280579,
      "rewards-max/forensics_strict_edge_reward": 0.5842172801494598,
      "rewards-max/forensics_tiou_reward": 0.9402792751789093,
      "rewards/forensics_sinkhorn_reward": 0.7921424508094788,
      "rewards/forensics_strict_edge_reward": 0.06680505909025669,
      "rewards/forensics_tiou_reward": 0.6070936024188995,
      "step": 777
    },
    {
      "completion_length": 21.625,
      "epoch": 3.2552301255230125,
      "grad_norm": 7.915487289428711,
      "kl": 0.30078125,
      "learning_rate": 1.8723849372384936e-07,
      "loss": 0.1671,
      "reward": 0.7445138394832611,
      "reward_std": 0.02934013493359089,
      "reward_std_all": 0.04657900892198086,
      "rewards-max/forensics_sinkhorn_reward": 0.9799278378486633,
      "rewards-max/forensics_strict_edge_reward": 0.5674136877059937,
      "rewards-max/forensics_tiou_reward": 0.9756142795085907,
      "rewards/forensics_sinkhorn_reward": 0.6898480653762817,
      "rewards/forensics_strict_edge_reward": 0.04750129207968712,
      "rewards/forensics_tiou_reward": 0.5009050965309143,
      "step": 778
    },
    {
      "completion_length": 22.375,
      "epoch": 3.2594142259414225,
      "grad_norm": 5.978592872619629,
      "kl": 0.3720703125,
      "learning_rate": 1.8619246861924684e-07,
      "loss": 0.0649,
      "reward": 0.7504746913909912,
      "reward_std": 0.023512943647801876,
      "reward_std_all": 0.04057210125029087,
      "rewards-max/forensics_sinkhorn_reward": 0.970952183008194,
      "rewards-max/forensics_strict_edge_reward": 0.42345638386905193,
      "rewards-max/forensics_tiou_reward": 0.9518176317214966,
      "rewards/forensics_sinkhorn_reward": 0.6971301436424255,
      "rewards/forensics_strict_edge_reward": 0.031398562248796225,
      "rewards/forensics_tiou_reward": 0.5034883171319962,
      "step": 779
    },
    {
      "completion_length": 21.78125,
      "epoch": 3.2635983263598325,
      "grad_norm": 5.523356914520264,
      "kl": 0.48046875,
      "learning_rate": 1.8514644351464433e-07,
      "loss": 0.0292,
      "reward": 0.7900877296924591,
      "reward_std": 0.017546418181154877,
      "reward_std_all": 0.01873498223721981,
      "rewards-max/forensics_sinkhorn_reward": 0.9807886779308319,
      "rewards-max/forensics_strict_edge_reward": 0.6799494326114655,
      "rewards-max/forensics_tiou_reward": 0.9777131974697113,
      "rewards/forensics_sinkhorn_reward": 0.8365085422992706,
      "rewards/forensics_strict_edge_reward": 0.0782395601272583,
      "rewards/forensics_tiou_reward": 0.6593524515628815,
      "step": 780
    },
    {
      "completion_length": 20.984375,
      "epoch": 3.2677824267782425,
      "grad_norm": 5.986364841461182,
      "kl": 0.4970703125,
      "learning_rate": 1.8410041841004184e-07,
      "loss": -0.0023,
      "reward": 0.8027259409427643,
      "reward_std": 0.005173437122721225,
      "reward_std_all": 0.0068765649339184165,
      "rewards-max/forensics_sinkhorn_reward": 0.9752268195152283,
      "rewards-max/forensics_strict_edge_reward": 0.3073139265179634,
      "rewards-max/forensics_tiou_reward": 0.9564521312713623,
      "rewards/forensics_sinkhorn_reward": 0.8213570713996887,
      "rewards/forensics_strict_edge_reward": 0.03545781411230564,
      "rewards/forensics_tiou_reward": 0.6322677731513977,
      "step": 781
    },
    {
      "completion_length": 20.703125,
      "epoch": 3.2719665271966525,
      "grad_norm": 7.781723499298096,
      "kl": 0.333984375,
      "learning_rate": 1.830543933054393e-07,
      "loss": 0.1019,
      "reward": 0.7577628791332245,
      "reward_std": 0.019016888923943043,
      "reward_std_all": 0.03442146722227335,
      "rewards-max/forensics_sinkhorn_reward": 0.9580096900463104,
      "rewards-max/forensics_strict_edge_reward": 0.29955458641052246,
      "rewards-max/forensics_tiou_reward": 0.9188042283058167,
      "rewards/forensics_sinkhorn_reward": 0.7468124330043793,
      "rewards/forensics_strict_edge_reward": 0.031151749193668365,
      "rewards/forensics_tiou_reward": 0.5455209612846375,
      "step": 782
    },
    {
      "completion_length": 19.640625,
      "epoch": 3.276150627615063,
      "grad_norm": 9.059541702270508,
      "kl": 0.5595703125,
      "learning_rate": 1.8200836820083681e-07,
      "loss": 0.0379,
      "reward": 0.709956169128418,
      "reward_std": 0.06306671537458897,
      "reward_std_all": 0.07842978090047836,
      "rewards-max/forensics_sinkhorn_reward": 0.9680561125278473,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9507627189159393,
      "rewards/forensics_sinkhorn_reward": 0.6907787322998047,
      "rewards/forensics_strict_edge_reward": 0.05226367712020874,
      "rewards/forensics_tiou_reward": 0.4519684612751007,
      "step": 783
    },
    {
      "completion_length": 23.25,
      "epoch": 3.280334728033473,
      "grad_norm": 6.567273139953613,
      "kl": 0.3876953125,
      "learning_rate": 1.8096234309623427e-07,
      "loss": 0.1593,
      "reward": 0.7466086447238922,
      "reward_std": 0.08500656671822071,
      "reward_std_all": 0.07761558890342712,
      "rewards-max/forensics_sinkhorn_reward": 0.9580501616001129,
      "rewards-max/forensics_strict_edge_reward": 0.4264601767063141,
      "rewards-max/forensics_tiou_reward": 0.9254061281681061,
      "rewards/forensics_sinkhorn_reward": 0.6773332357406616,
      "rewards/forensics_strict_edge_reward": 0.023058888968080282,
      "rewards/forensics_tiou_reward": 0.4797671437263489,
      "step": 784
    },
    {
      "completion_length": 20.0625,
      "epoch": 3.284518828451883,
      "grad_norm": 7.927498817443848,
      "kl": 0.4326171875,
      "learning_rate": 1.799163179916318e-07,
      "loss": 0.0602,
      "reward": 0.7631827890872955,
      "reward_std": 0.03120154794305563,
      "reward_std_all": 0.04078665282577276,
      "rewards-max/forensics_sinkhorn_reward": 0.953567624092102,
      "rewards-max/forensics_strict_edge_reward": 0.2714046686887741,
      "rewards-max/forensics_tiou_reward": 0.9176006615161896,
      "rewards/forensics_sinkhorn_reward": 0.7122520506381989,
      "rewards/forensics_strict_edge_reward": 0.018091400619596243,
      "rewards/forensics_tiou_reward": 0.450994536280632,
      "step": 785
    },
    {
      "completion_length": 22.9375,
      "epoch": 3.288702928870293,
      "grad_norm": 6.239333152770996,
      "kl": 0.3408203125,
      "learning_rate": 1.788702928870293e-07,
      "loss": -0.0752,
      "reward": 0.7431821227073669,
      "reward_std": 0.04764893278479576,
      "reward_std_all": 0.058638665825128555,
      "rewards-max/forensics_sinkhorn_reward": 0.9456849992275238,
      "rewards-max/forensics_strict_edge_reward": 0.3307340331375599,
      "rewards-max/forensics_tiou_reward": 0.8949432671070099,
      "rewards/forensics_sinkhorn_reward": 0.691461592912674,
      "rewards/forensics_strict_edge_reward": 0.031109060801099986,
      "rewards/forensics_tiou_reward": 0.5028467178344727,
      "step": 786
    },
    {
      "completion_length": 21.96875,
      "epoch": 3.292887029288703,
      "grad_norm": 5.508175849914551,
      "kl": 0.35546875,
      "learning_rate": 1.7782426778242676e-07,
      "loss": -0.0454,
      "reward": 0.7701730728149414,
      "reward_std": 0.04411143250763416,
      "reward_std_all": 0.04246266931295395,
      "rewards-max/forensics_sinkhorn_reward": 0.9755933582782745,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.9646316170692444,
      "rewards/forensics_sinkhorn_reward": 0.7763047516345978,
      "rewards/forensics_strict_edge_reward": 0.07116478309035301,
      "rewards/forensics_tiou_reward": 0.6588554978370667,
      "step": 787
    },
    {
      "completion_length": 21.390625,
      "epoch": 3.297071129707113,
      "grad_norm": 6.57735538482666,
      "kl": 0.306640625,
      "learning_rate": 1.7677824267782427e-07,
      "loss": -0.1286,
      "reward": 0.7671180665493011,
      "reward_std": 0.002162480785045773,
      "reward_std_all": 0.022558929747901857,
      "rewards-max/forensics_sinkhorn_reward": 0.9785647690296173,
      "rewards-max/forensics_strict_edge_reward": 0.447468563914299,
      "rewards-max/forensics_tiou_reward": 0.9720197021961212,
      "rewards/forensics_sinkhorn_reward": 0.7683510780334473,
      "rewards/forensics_strict_edge_reward": 0.06304881907999516,
      "rewards/forensics_tiou_reward": 0.5928429365158081,
      "step": 788
    },
    {
      "completion_length": 20.75,
      "epoch": 3.301255230125523,
      "grad_norm": 6.660752296447754,
      "kl": 0.3984375,
      "learning_rate": 1.7573221757322173e-07,
      "loss": -0.0416,
      "reward": 0.7916804850101471,
      "reward_std": 0.014118732477072626,
      "reward_std_all": 0.01604854059405625,
      "rewards-max/forensics_sinkhorn_reward": 0.9719114601612091,
      "rewards-max/forensics_strict_edge_reward": 0.6329656541347504,
      "rewards-max/forensics_tiou_reward": 0.9551575481891632,
      "rewards/forensics_sinkhorn_reward": 0.7746077179908752,
      "rewards/forensics_strict_edge_reward": 0.048336658626794815,
      "rewards/forensics_tiou_reward": 0.5546551048755646,
      "step": 789
    },
    {
      "completion_length": 21.1875,
      "epoch": 3.305439330543933,
      "grad_norm": 9.132847785949707,
      "kl": 0.4072265625,
      "learning_rate": 1.7468619246861925e-07,
      "loss": 0.101,
      "reward": 0.7677805423736572,
      "reward_std": 0.018792220391333103,
      "reward_std_all": 0.030649456195533276,
      "rewards-max/forensics_sinkhorn_reward": 0.9837975800037384,
      "rewards-max/forensics_strict_edge_reward": 0.42810201877728105,
      "rewards-max/forensics_tiou_reward": 0.980342447757721,
      "rewards/forensics_sinkhorn_reward": 0.7043859660625458,
      "rewards/forensics_strict_edge_reward": 0.02163740061223507,
      "rewards/forensics_tiou_reward": 0.5282243192195892,
      "step": 790
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.309623430962343,
      "grad_norm": 6.220984935760498,
      "kl": 0.3056640625,
      "learning_rate": 1.7364016736401673e-07,
      "loss": 0.0491,
      "reward": 0.7313710451126099,
      "reward_std": 0.07114763371646404,
      "reward_std_all": 0.0759793072938919,
      "rewards-max/forensics_sinkhorn_reward": 0.9778036177158356,
      "rewards-max/forensics_strict_edge_reward": 0.6071998775005341,
      "rewards-max/forensics_tiou_reward": 0.9678852260112762,
      "rewards/forensics_sinkhorn_reward": 0.7211908996105194,
      "rewards/forensics_strict_edge_reward": 0.0566883310675621,
      "rewards/forensics_tiou_reward": 0.5950093269348145,
      "step": 791
    },
    {
      "completion_length": 21.546875,
      "epoch": 3.3138075313807533,
      "grad_norm": 8.06882095336914,
      "kl": 0.3125,
      "learning_rate": 1.7259414225941422e-07,
      "loss": -0.0384,
      "reward": 0.7885322570800781,
      "reward_std": 0.00749534135684371,
      "reward_std_all": 0.014772506430745125,
      "rewards-max/forensics_sinkhorn_reward": 0.984860897064209,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9898083806037903,
      "rewards/forensics_sinkhorn_reward": 0.761884868144989,
      "rewards/forensics_strict_edge_reward": 0.11030701547861099,
      "rewards/forensics_tiou_reward": 0.6051622033119202,
      "step": 792
    },
    {
      "completion_length": 21.578125,
      "epoch": 3.3179916317991633,
      "grad_norm": 6.893744468688965,
      "kl": 0.4169921875,
      "learning_rate": 1.715481171548117e-07,
      "loss": -0.0086,
      "reward": 0.7698929309844971,
      "reward_std": 0.03930641710758209,
      "reward_std_all": 0.03983998764306307,
      "rewards-max/forensics_sinkhorn_reward": 0.9809042513370514,
      "rewards-max/forensics_strict_edge_reward": 0.5948129594326019,
      "rewards-max/forensics_tiou_reward": 0.9762109518051147,
      "rewards/forensics_sinkhorn_reward": 0.7627754807472229,
      "rewards/forensics_strict_edge_reward": 0.05914683640003204,
      "rewards/forensics_tiou_reward": 0.553452342748642,
      "step": 793
    },
    {
      "completion_length": 23.265625,
      "epoch": 3.3221757322175733,
      "grad_norm": 8.741007804870605,
      "kl": 0.3564453125,
      "learning_rate": 1.705020920502092e-07,
      "loss": -0.0444,
      "reward": 0.7811527252197266,
      "reward_std": 0.023960530757904053,
      "reward_std_all": 0.025540808215737343,
      "rewards-max/forensics_sinkhorn_reward": 0.9721740484237671,
      "rewards-max/forensics_strict_edge_reward": 0.7126307189464569,
      "rewards-max/forensics_tiou_reward": 0.9618959724903107,
      "rewards/forensics_sinkhorn_reward": 0.7476806640625,
      "rewards/forensics_strict_edge_reward": 0.07948921620845795,
      "rewards/forensics_tiou_reward": 0.5630234479904175,
      "step": 794
    },
    {
      "completion_length": 22.359375,
      "epoch": 3.3263598326359833,
      "grad_norm": 7.070977210998535,
      "kl": 0.41162109375,
      "learning_rate": 1.6945606694560668e-07,
      "loss": -0.0014,
      "reward": 0.7871197462081909,
      "reward_std": 0.017282592598348856,
      "reward_std_all": 0.019155648071318865,
      "rewards-max/forensics_sinkhorn_reward": 0.9710702300071716,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9464830756187439,
      "rewards/forensics_sinkhorn_reward": 0.771152526140213,
      "rewards/forensics_strict_edge_reward": 0.05131570249795914,
      "rewards/forensics_tiou_reward": 0.5955201089382172,
      "step": 795
    },
    {
      "completion_length": 20.234375,
      "epoch": 3.3305439330543933,
      "grad_norm": 8.060636520385742,
      "kl": 0.37109375,
      "learning_rate": 1.684100418410042e-07,
      "loss": -0.015,
      "reward": 0.7489282786846161,
      "reward_std": 0.04690774530172348,
      "reward_std_all": 0.05148688517510891,
      "rewards-max/forensics_sinkhorn_reward": 0.9557225406169891,
      "rewards-max/forensics_strict_edge_reward": 0.1352136954665184,
      "rewards-max/forensics_tiou_reward": 0.9035377502441406,
      "rewards/forensics_sinkhorn_reward": 0.7360593974590302,
      "rewards/forensics_strict_edge_reward": 0.013903260231018066,
      "rewards/forensics_tiou_reward": 0.5124699026346207,
      "step": 796
    },
    {
      "completion_length": 22.84375,
      "epoch": 3.3347280334728033,
      "grad_norm": 8.759705543518066,
      "kl": 0.30859375,
      "learning_rate": 1.6736401673640165e-07,
      "loss": -0.0361,
      "reward": 0.7723678946495056,
      "reward_std": 0.03361000493168831,
      "reward_std_all": 0.034711774438619614,
      "rewards-max/forensics_sinkhorn_reward": 0.9778217077255249,
      "rewards-max/forensics_strict_edge_reward": 0.40107065439224243,
      "rewards-max/forensics_tiou_reward": 0.9684641659259796,
      "rewards/forensics_sinkhorn_reward": 0.7372790575027466,
      "rewards/forensics_strict_edge_reward": 0.03909878572449088,
      "rewards/forensics_tiou_reward": 0.5659291446208954,
      "step": 797
    },
    {
      "completion_length": 20.4375,
      "epoch": 3.3389121338912133,
      "grad_norm": 5.877114295959473,
      "kl": 0.3408203125,
      "learning_rate": 1.6631799163179916e-07,
      "loss": -0.0135,
      "reward": 0.7736591100692749,
      "reward_std": 0.0381347022484988,
      "reward_std_all": 0.037539954762905836,
      "rewards-max/forensics_sinkhorn_reward": 0.9784127473831177,
      "rewards-max/forensics_strict_edge_reward": 0.4199727475643158,
      "rewards-max/forensics_tiou_reward": 0.9694953262805939,
      "rewards/forensics_sinkhorn_reward": 0.7890813648700714,
      "rewards/forensics_strict_edge_reward": 0.03383406437933445,
      "rewards/forensics_tiou_reward": 0.6253439784049988,
      "step": 798
    },
    {
      "completion_length": 20.109375,
      "epoch": 3.3430962343096233,
      "grad_norm": 6.579657077789307,
      "kl": 0.3828125,
      "learning_rate": 1.6527196652719662e-07,
      "loss": 0.1293,
      "reward": 0.7763651311397552,
      "reward_std": 0.042637135833501816,
      "reward_std_all": 0.04011820629239082,
      "rewards-max/forensics_sinkhorn_reward": 0.95868781208992,
      "rewards-max/forensics_strict_edge_reward": 0.3243762031197548,
      "rewards-max/forensics_tiou_reward": 0.926929384469986,
      "rewards/forensics_sinkhorn_reward": 0.7838765382766724,
      "rewards/forensics_strict_edge_reward": 0.03568363655358553,
      "rewards/forensics_tiou_reward": 0.5958737134933472,
      "step": 799
    },
    {
      "completion_length": 20.109375,
      "epoch": 3.3472803347280333,
      "grad_norm": 7.66799259185791,
      "kl": 0.3642578125,
      "learning_rate": 1.6422594142259414e-07,
      "loss": -0.0049,
      "reward": 0.7994895279407501,
      "reward_std": 0.007613949186634272,
      "reward_std_all": 0.009388919919729233,
      "rewards-max/forensics_sinkhorn_reward": 0.9670813679695129,
      "rewards-max/forensics_strict_edge_reward": 0.38679473102092743,
      "rewards-max/forensics_tiou_reward": 0.9392412006855011,
      "rewards/forensics_sinkhorn_reward": 0.7885564863681793,
      "rewards/forensics_strict_edge_reward": 0.03812688775360584,
      "rewards/forensics_tiou_reward": 0.574322521686554,
      "step": 800
    },
    {
      "completion_length": 21.640625,
      "epoch": 3.3514644351464433,
      "grad_norm": 5.523834228515625,
      "kl": 0.3671875,
      "learning_rate": 1.6317991631799162e-07,
      "loss": -0.081,
      "reward": 0.748080313205719,
      "reward_std": 0.0422613222617656,
      "reward_std_all": 0.05429101409390569,
      "rewards-max/forensics_sinkhorn_reward": 0.9808062016963959,
      "rewards-max/forensics_strict_edge_reward": 0.7032848298549652,
      "rewards-max/forensics_tiou_reward": 0.9799481928348541,
      "rewards/forensics_sinkhorn_reward": 0.7191286087036133,
      "rewards/forensics_strict_edge_reward": 0.04700538143515587,
      "rewards/forensics_tiou_reward": 0.5357489287853241,
      "step": 801
    },
    {
      "completion_length": 21.1875,
      "epoch": 3.3556485355648533,
      "grad_norm": 7.32793664932251,
      "kl": 0.494140625,
      "learning_rate": 1.621338912133891e-07,
      "loss": 0.0118,
      "reward": 0.7509387135505676,
      "reward_std": 0.066734179854393,
      "reward_std_all": 0.0647725909948349,
      "rewards-max/forensics_sinkhorn_reward": 0.9729966521263123,
      "rewards-max/forensics_strict_edge_reward": 0.39585426449775696,
      "rewards-max/forensics_tiou_reward": 0.9578446745872498,
      "rewards/forensics_sinkhorn_reward": 0.7172185182571411,
      "rewards/forensics_strict_edge_reward": 0.03212697897106409,
      "rewards/forensics_tiou_reward": 0.5009692311286926,
      "step": 802
    },
    {
      "completion_length": 19.875,
      "epoch": 3.3598326359832638,
      "grad_norm": 17.784446716308594,
      "kl": 1.021484375,
      "learning_rate": 1.610878661087866e-07,
      "loss": 0.1538,
      "reward": 0.7702782154083252,
      "reward_std": 0.03907695319503546,
      "reward_std_all": 0.03914143145084381,
      "rewards-max/forensics_sinkhorn_reward": 0.979125052690506,
      "rewards-max/forensics_strict_edge_reward": 0.7697611749172211,
      "rewards-max/forensics_tiou_reward": 0.9766809046268463,
      "rewards/forensics_sinkhorn_reward": 0.760374128818512,
      "rewards/forensics_strict_edge_reward": 0.05924663692712784,
      "rewards/forensics_tiou_reward": 0.5841096639633179,
      "step": 803
    },
    {
      "completion_length": 22.03125,
      "epoch": 3.3640167364016738,
      "grad_norm": 6.621249675750732,
      "kl": 0.408203125,
      "learning_rate": 1.6004184100418408e-07,
      "loss": 0.0043,
      "reward": 0.7529688477516174,
      "reward_std": 0.032969060353934765,
      "reward_std_all": 0.04483179817907512,
      "rewards-max/forensics_sinkhorn_reward": 0.9550533294677734,
      "rewards-max/forensics_strict_edge_reward": 0.3617284297943115,
      "rewards-max/forensics_tiou_reward": 0.9216579794883728,
      "rewards/forensics_sinkhorn_reward": 0.7149167954921722,
      "rewards/forensics_strict_edge_reward": 0.029716182500123978,
      "rewards/forensics_tiou_reward": 0.509994626045227,
      "step": 804
    },
    {
      "completion_length": 21.84375,
      "epoch": 3.3682008368200838,
      "grad_norm": 6.258186340332031,
      "kl": 0.46484375,
      "learning_rate": 1.5899581589958157e-07,
      "loss": 0.0474,
      "reward": 0.7798207402229309,
      "reward_std": 0.03263443359173834,
      "reward_std_all": 0.03197407815605402,
      "rewards-max/forensics_sinkhorn_reward": 0.9653770327568054,
      "rewards-max/forensics_strict_edge_reward": 0.42036955058574677,
      "rewards-max/forensics_tiou_reward": 0.9372091293334961,
      "rewards/forensics_sinkhorn_reward": 0.746066689491272,
      "rewards/forensics_strict_edge_reward": 0.05496158450841904,
      "rewards/forensics_tiou_reward": 0.540263295173645,
      "step": 805
    },
    {
      "completion_length": 19.859375,
      "epoch": 3.3723849372384938,
      "grad_norm": 8.029386520385742,
      "kl": 0.3349609375,
      "learning_rate": 1.5794979079497908e-07,
      "loss": -0.0309,
      "reward": 0.7929491102695465,
      "reward_std": 0.018696858664043248,
      "reward_std_all": 0.018849698826670647,
      "rewards-max/forensics_sinkhorn_reward": 0.967987984418869,
      "rewards-max/forensics_strict_edge_reward": 0.5303425639867783,
      "rewards-max/forensics_tiou_reward": 0.9470493495464325,
      "rewards/forensics_sinkhorn_reward": 0.8220845758914948,
      "rewards/forensics_strict_edge_reward": 0.02750046644359827,
      "rewards/forensics_tiou_reward": 0.6431486308574677,
      "step": 806
    },
    {
      "completion_length": 21.890625,
      "epoch": 3.3765690376569037,
      "grad_norm": 6.937732219696045,
      "kl": 0.345703125,
      "learning_rate": 1.5690376569037657e-07,
      "loss": 0.0523,
      "reward": 0.7720559537410736,
      "reward_std": 0.04223136976361275,
      "reward_std_all": 0.040846776217222214,
      "rewards-max/forensics_sinkhorn_reward": 0.9763859510421753,
      "rewards-max/forensics_strict_edge_reward": 0.538444846868515,
      "rewards-max/forensics_tiou_reward": 0.9657335877418518,
      "rewards/forensics_sinkhorn_reward": 0.7509004771709442,
      "rewards/forensics_strict_edge_reward": 0.05203245393931866,
      "rewards/forensics_tiou_reward": 0.5587925016880035,
      "step": 807
    },
    {
      "completion_length": 21.109375,
      "epoch": 3.3807531380753137,
      "grad_norm": 6.966831684112549,
      "kl": 0.458984375,
      "learning_rate": 1.5585774058577405e-07,
      "loss": -0.1391,
      "reward": 0.8025777041912079,
      "reward_std": 0.0028570304857566953,
      "reward_std_all": 0.00523855141364038,
      "rewards-max/forensics_sinkhorn_reward": 0.9747357070446014,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9617063999176025,
      "rewards/forensics_sinkhorn_reward": 0.7895414233207703,
      "rewards/forensics_strict_edge_reward": 0.0477710273116827,
      "rewards/forensics_tiou_reward": 0.5500428080558777,
      "step": 808
    },
    {
      "completion_length": 20.875,
      "epoch": 3.3849372384937237,
      "grad_norm": 7.574100971221924,
      "kl": 0.3974609375,
      "learning_rate": 1.5481171548117154e-07,
      "loss": 0.0478,
      "reward": 0.7770828604698181,
      "reward_std": 0.029553562868386507,
      "reward_std_all": 0.030217858962714672,
      "rewards-max/forensics_sinkhorn_reward": 0.9874862432479858,
      "rewards-max/forensics_strict_edge_reward": 0.7082725167274475,
      "rewards-max/forensics_tiou_reward": 0.9948723614215851,
      "rewards/forensics_sinkhorn_reward": 0.7931621074676514,
      "rewards/forensics_strict_edge_reward": 0.054892534390091896,
      "rewards/forensics_tiou_reward": 0.6157583594322205,
      "step": 809
    },
    {
      "completion_length": 21.21875,
      "epoch": 3.3891213389121337,
      "grad_norm": 9.359411239624023,
      "kl": 0.353515625,
      "learning_rate": 1.5376569037656903e-07,
      "loss": 0.1044,
      "reward": 0.7708273828029633,
      "reward_std": 0.033632973209023476,
      "reward_std_all": 0.036914754658937454,
      "rewards-max/forensics_sinkhorn_reward": 0.9616182148456573,
      "rewards-max/forensics_strict_edge_reward": 0.4211113452911377,
      "rewards-max/forensics_tiou_reward": 0.9335988759994507,
      "rewards/forensics_sinkhorn_reward": 0.725519061088562,
      "rewards/forensics_strict_edge_reward": 0.021261338144540787,
      "rewards/forensics_tiou_reward": 0.44075845181941986,
      "step": 810
    },
    {
      "completion_length": 21.640625,
      "epoch": 3.393305439330544,
      "grad_norm": 6.827967643737793,
      "kl": 0.3466796875,
      "learning_rate": 1.5271966527196654e-07,
      "loss": -0.0199,
      "reward": 0.7744188904762268,
      "reward_std": 0.04806117620319128,
      "reward_std_all": 0.0446312977001071,
      "rewards-max/forensics_sinkhorn_reward": 0.9790072441101074,
      "rewards-max/forensics_strict_edge_reward": 0.6516932249069214,
      "rewards-max/forensics_tiou_reward": 0.9756821691989899,
      "rewards/forensics_sinkhorn_reward": 0.7255043983459473,
      "rewards/forensics_strict_edge_reward": 0.055999502539634705,
      "rewards/forensics_tiou_reward": 0.5145313590764999,
      "step": 811
    },
    {
      "completion_length": 24.90625,
      "epoch": 3.397489539748954,
      "grad_norm": 5.950974941253662,
      "kl": 0.44921875,
      "learning_rate": 1.51673640167364e-07,
      "loss": 0.0749,
      "reward": 0.7354532480239868,
      "reward_std": 0.03393243998289108,
      "reward_std_all": 0.05336579121649265,
      "rewards-max/forensics_sinkhorn_reward": 0.9689257740974426,
      "rewards-max/forensics_strict_edge_reward": 0.13495207950472832,
      "rewards-max/forensics_tiou_reward": 0.9412494897842407,
      "rewards/forensics_sinkhorn_reward": 0.6845520436763763,
      "rewards/forensics_strict_edge_reward": 0.010460261604748666,
      "rewards/forensics_tiou_reward": 0.4691345989704132,
      "step": 812
    },
    {
      "completion_length": 26.90625,
      "epoch": 3.401673640167364,
      "grad_norm": 6.3884992599487305,
      "kl": 0.404296875,
      "learning_rate": 1.506276150627615e-07,
      "loss": 0.0109,
      "reward": 0.680547833442688,
      "reward_std": 0.054462955333292484,
      "reward_std_all": 0.08581947907805443,
      "rewards-max/forensics_sinkhorn_reward": 0.9635565876960754,
      "rewards-max/forensics_strict_edge_reward": 0.29384535178542137,
      "rewards-max/forensics_tiou_reward": 0.9352599382400513,
      "rewards/forensics_sinkhorn_reward": 0.6262286007404327,
      "rewards/forensics_strict_edge_reward": 0.020351424114778638,
      "rewards/forensics_tiou_reward": 0.4585167169570923,
      "step": 813
    },
    {
      "completion_length": 23.828125,
      "epoch": 3.405857740585774,
      "grad_norm": 5.8320488929748535,
      "kl": 0.3779296875,
      "learning_rate": 1.4958158995815897e-07,
      "loss": 0.0281,
      "reward": 0.7599406242370605,
      "reward_std": 0.03326094523072243,
      "reward_std_all": 0.03922251798212528,
      "rewards-max/forensics_sinkhorn_reward": 0.9688684046268463,
      "rewards-max/forensics_strict_edge_reward": 0.4565335884690285,
      "rewards-max/forensics_tiou_reward": 0.9502643942832947,
      "rewards/forensics_sinkhorn_reward": 0.7638768553733826,
      "rewards/forensics_strict_edge_reward": 0.05054555821698159,
      "rewards/forensics_tiou_reward": 0.5772241950035095,
      "step": 814
    },
    {
      "completion_length": 23.8125,
      "epoch": 3.410041841004184,
      "grad_norm": 6.044538497924805,
      "kl": 0.380859375,
      "learning_rate": 1.4853556485355648e-07,
      "loss": 0.0662,
      "reward": 0.7307977378368378,
      "reward_std": 0.07132145017385483,
      "reward_std_all": 0.07706005126237869,
      "rewards-max/forensics_sinkhorn_reward": 0.9318303763866425,
      "rewards-max/forensics_strict_edge_reward": 0.13078086450695992,
      "rewards-max/forensics_tiou_reward": 0.8437730073928833,
      "rewards/forensics_sinkhorn_reward": 0.6807572543621063,
      "rewards/forensics_strict_edge_reward": 0.013854974415153265,
      "rewards/forensics_tiou_reward": 0.48382073640823364,
      "step": 815
    },
    {
      "completion_length": 22.578125,
      "epoch": 3.414225941422594,
      "grad_norm": 5.92069673538208,
      "kl": 0.314453125,
      "learning_rate": 1.4748953974895397e-07,
      "loss": 0.1261,
      "reward": 0.77509805560112,
      "reward_std": 0.04915682412683964,
      "reward_std_all": 0.04463639482855797,
      "rewards-max/forensics_sinkhorn_reward": 0.985777884721756,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9921179711818695,
      "rewards/forensics_sinkhorn_reward": 0.7750486433506012,
      "rewards/forensics_strict_edge_reward": 0.07321873307228088,
      "rewards/forensics_tiou_reward": 0.6574290990829468,
      "step": 816
    },
    {
      "completion_length": 23.15625,
      "epoch": 3.418410041841004,
      "grad_norm": 7.278539180755615,
      "kl": 0.4189453125,
      "learning_rate": 1.4644351464435146e-07,
      "loss": 0.0956,
      "reward": 0.7752028703689575,
      "reward_std": 0.03125095972791314,
      "reward_std_all": 0.03228925075381994,
      "rewards-max/forensics_sinkhorn_reward": 0.9423194825649261,
      "rewards-max/forensics_strict_edge_reward": 0.1794937588274479,
      "rewards-max/forensics_tiou_reward": 0.88166344165802,
      "rewards/forensics_sinkhorn_reward": 0.7068197727203369,
      "rewards/forensics_strict_edge_reward": 0.0206728707998991,
      "rewards/forensics_tiou_reward": 0.5314223170280457,
      "step": 817
    },
    {
      "completion_length": 21.09375,
      "epoch": 3.422594142259414,
      "grad_norm": 6.099677085876465,
      "kl": 0.4091796875,
      "learning_rate": 1.4539748953974894e-07,
      "loss": 0.0436,
      "reward": 0.7603373229503632,
      "reward_std": 0.054163675755262375,
      "reward_std_all": 0.05306771211326122,
      "rewards-max/forensics_sinkhorn_reward": 0.9707043170928955,
      "rewards-max/forensics_strict_edge_reward": 0.44934946298599243,
      "rewards-max/forensics_tiou_reward": 0.9558939933776855,
      "rewards/forensics_sinkhorn_reward": 0.7406930327415466,
      "rewards/forensics_strict_edge_reward": 0.05829968675971031,
      "rewards/forensics_tiou_reward": 0.5832608938217163,
      "step": 818
    },
    {
      "completion_length": 22.34375,
      "epoch": 3.426778242677824,
      "grad_norm": 7.35740852355957,
      "kl": 0.3583984375,
      "learning_rate": 1.4435146443514643e-07,
      "loss": 0.1718,
      "reward": 0.7657943964004517,
      "reward_std": 0.04362721275538206,
      "reward_std_all": 0.04353518784046173,
      "rewards-max/forensics_sinkhorn_reward": 0.9412575662136078,
      "rewards-max/forensics_strict_edge_reward": 0.557279422879219,
      "rewards-max/forensics_tiou_reward": 0.8824307322502136,
      "rewards/forensics_sinkhorn_reward": 0.6568180918693542,
      "rewards/forensics_strict_edge_reward": 0.033956184051930904,
      "rewards/forensics_tiou_reward": 0.48120109736919403,
      "step": 819
    },
    {
      "completion_length": 21.125,
      "epoch": 3.430962343096234,
      "grad_norm": 6.637228488922119,
      "kl": 0.3740234375,
      "learning_rate": 1.4330543933054392e-07,
      "loss": 0.0112,
      "reward": 0.7654933929443359,
      "reward_std": 0.061839308589696884,
      "reward_std_all": 0.0566537044942379,
      "rewards-max/forensics_sinkhorn_reward": 0.9705776572227478,
      "rewards-max/forensics_strict_edge_reward": 0.5985658019781113,
      "rewards-max/forensics_tiou_reward": 0.9518717229366302,
      "rewards/forensics_sinkhorn_reward": 0.722503274679184,
      "rewards/forensics_strict_edge_reward": 0.04006964713335037,
      "rewards/forensics_tiou_reward": 0.5059535205364227,
      "step": 820
    },
    {
      "completion_length": 22.140625,
      "epoch": 3.435146443514644,
      "grad_norm": 6.4422926902771,
      "kl": 0.47265625,
      "learning_rate": 1.4225941422594143e-07,
      "loss": 0.008,
      "reward": 0.7195838689804077,
      "reward_std": 0.023094854317605495,
      "reward_std_all": 0.05648980662226677,
      "rewards-max/forensics_sinkhorn_reward": 0.976131409406662,
      "rewards-max/forensics_strict_edge_reward": 0.34924091398715973,
      "rewards-max/forensics_tiou_reward": 0.9645052850246429,
      "rewards/forensics_sinkhorn_reward": 0.6932184994220734,
      "rewards/forensics_strict_edge_reward": 0.028675531968474388,
      "rewards/forensics_tiou_reward": 0.5137563347816467,
      "step": 821
    },
    {
      "completion_length": 19.765625,
      "epoch": 3.439330543933054,
      "grad_norm": 5.524932384490967,
      "kl": 0.3974609375,
      "learning_rate": 1.412133891213389e-07,
      "loss": -0.0136,
      "reward": 0.7762265801429749,
      "reward_std": 0.03653988288715482,
      "reward_std_all": 0.03574567288160324,
      "rewards-max/forensics_sinkhorn_reward": 0.9770691990852356,
      "rewards-max/forensics_strict_edge_reward": 0.7558856904506683,
      "rewards-max/forensics_tiou_reward": 0.9723942577838898,
      "rewards/forensics_sinkhorn_reward": 0.7885571420192719,
      "rewards/forensics_strict_edge_reward": 0.05715354159474373,
      "rewards/forensics_tiou_reward": 0.5971341729164124,
      "step": 822
    },
    {
      "completion_length": 21.71875,
      "epoch": 3.4435146443514646,
      "grad_norm": 6.363757133483887,
      "kl": 0.3759765625,
      "learning_rate": 1.401673640167364e-07,
      "loss": -0.041,
      "reward": 0.7891710102558136,
      "reward_std": 0.028054147376678884,
      "reward_std_all": 0.02667602291330695,
      "rewards-max/forensics_sinkhorn_reward": 0.9627071917057037,
      "rewards-max/forensics_strict_edge_reward": 0.5448362529277802,
      "rewards-max/forensics_tiou_reward": 0.9341365694999695,
      "rewards/forensics_sinkhorn_reward": 0.7544427216053009,
      "rewards/forensics_strict_edge_reward": 0.060894581489264965,
      "rewards/forensics_tiou_reward": 0.5759961903095245,
      "step": 823
    },
    {
      "completion_length": 21.078125,
      "epoch": 3.4476987447698746,
      "grad_norm": 8.651427268981934,
      "kl": 0.662109375,
      "learning_rate": 1.3912133891213386e-07,
      "loss": 0.0042,
      "reward": 0.7346669733524323,
      "reward_std": 0.09149198234081268,
      "reward_std_all": 0.08516361936926842,
      "rewards-max/forensics_sinkhorn_reward": 0.9728347957134247,
      "rewards-max/forensics_strict_edge_reward": 0.6660739481449127,
      "rewards-max/forensics_tiou_reward": 0.962032675743103,
      "rewards/forensics_sinkhorn_reward": 0.6937438547611237,
      "rewards/forensics_strict_edge_reward": 0.05671770125627518,
      "rewards/forensics_tiou_reward": 0.44462548196315765,
      "step": 824
    },
    {
      "completion_length": 22.078125,
      "epoch": 3.4518828451882846,
      "grad_norm": 5.703679084777832,
      "kl": 0.4658203125,
      "learning_rate": 1.3807531380753137e-07,
      "loss": 0.0373,
      "reward": 0.738155871629715,
      "reward_std": 0.04554485157132149,
      "reward_std_all": 0.05927329510450363,
      "rewards-max/forensics_sinkhorn_reward": 0.9734642803668976,
      "rewards-max/forensics_strict_edge_reward": 0.6388925462961197,
      "rewards-max/forensics_tiou_reward": 0.9608678221702576,
      "rewards/forensics_sinkhorn_reward": 0.658788412809372,
      "rewards/forensics_strict_edge_reward": 0.06756412610411644,
      "rewards/forensics_tiou_reward": 0.48869381844997406,
      "step": 825
    },
    {
      "completion_length": 24.0625,
      "epoch": 3.4560669456066946,
      "grad_norm": 5.7595133781433105,
      "kl": 0.404296875,
      "learning_rate": 1.370292887029289e-07,
      "loss": 0.1443,
      "reward": 0.7370258271694183,
      "reward_std": 0.08330840617418289,
      "reward_std_all": 0.07996333763003349,
      "rewards-max/forensics_sinkhorn_reward": 0.9868581295013428,
      "rewards-max/forensics_strict_edge_reward": 0.4712325632572174,
      "rewards-max/forensics_tiou_reward": 0.993254691362381,
      "rewards/forensics_sinkhorn_reward": 0.7121036648750305,
      "rewards/forensics_strict_edge_reward": 0.054006464779376984,
      "rewards/forensics_tiou_reward": 0.5500648617744446,
      "step": 826
    },
    {
      "completion_length": 22.828125,
      "epoch": 3.4602510460251046,
      "grad_norm": 7.052113056182861,
      "kl": 0.310546875,
      "learning_rate": 1.3598326359832635e-07,
      "loss": 0.1647,
      "reward": 0.7546384930610657,
      "reward_std": 0.07303765788674355,
      "reward_std_all": 0.06832148507237434,
      "rewards-max/forensics_sinkhorn_reward": 0.9816852807998657,
      "rewards-max/forensics_strict_edge_reward": 0.24325479194521904,
      "rewards-max/forensics_tiou_reward": 0.9768209159374237,
      "rewards/forensics_sinkhorn_reward": 0.7325581610202789,
      "rewards/forensics_strict_edge_reward": 0.02340748906135559,
      "rewards/forensics_tiou_reward": 0.5531102120876312,
      "step": 827
    },
    {
      "completion_length": 21.78125,
      "epoch": 3.4644351464435146,
      "grad_norm": 6.492366790771484,
      "kl": 0.3056640625,
      "learning_rate": 1.3493723849372386e-07,
      "loss": 0.0842,
      "reward": 0.7673837244510651,
      "reward_std": 0.051435165107250214,
      "reward_std_all": 0.04826752468943596,
      "rewards-max/forensics_sinkhorn_reward": 0.9859803915023804,
      "rewards-max/forensics_strict_edge_reward": 0.7047859728336334,
      "rewards-max/forensics_tiou_reward": 0.9917881190776825,
      "rewards/forensics_sinkhorn_reward": 0.7543653547763824,
      "rewards/forensics_strict_edge_reward": 0.07314925640821457,
      "rewards/forensics_tiou_reward": 0.5452267229557037,
      "step": 828
    },
    {
      "completion_length": 21.578125,
      "epoch": 3.4686192468619246,
      "grad_norm": 7.428469181060791,
      "kl": 0.4326171875,
      "learning_rate": 1.3389121338912132e-07,
      "loss": 0.1028,
      "reward": 0.7593212127685547,
      "reward_std": 0.052031999453902245,
      "reward_std_all": 0.050630033016204834,
      "rewards-max/forensics_sinkhorn_reward": 0.9641145467758179,
      "rewards-max/forensics_strict_edge_reward": 0.32197293639183044,
      "rewards-max/forensics_tiou_reward": 0.9387044608592987,
      "rewards/forensics_sinkhorn_reward": 0.7334127724170685,
      "rewards/forensics_strict_edge_reward": 0.03384522441774607,
      "rewards/forensics_tiou_reward": 0.5490262806415558,
      "step": 829
    },
    {
      "completion_length": 20.078125,
      "epoch": 3.4728033472803346,
      "grad_norm": 6.428531169891357,
      "kl": 0.4169921875,
      "learning_rate": 1.3284518828451883e-07,
      "loss": 0.043,
      "reward": 0.7652644515037537,
      "reward_std": 0.05881619080901146,
      "reward_std_all": 0.054396942257881165,
      "rewards-max/forensics_sinkhorn_reward": 0.9827115535736084,
      "rewards-max/forensics_strict_edge_reward": 0.5933051109313965,
      "rewards-max/forensics_tiou_reward": 0.9791945815086365,
      "rewards/forensics_sinkhorn_reward": 0.7954525351524353,
      "rewards/forensics_strict_edge_reward": 0.03902386873960495,
      "rewards/forensics_tiou_reward": 0.6219479143619537,
      "step": 830
    },
    {
      "completion_length": 21.109375,
      "epoch": 3.476987447698745,
      "grad_norm": 8.500510215759277,
      "kl": 0.439453125,
      "learning_rate": 1.3179916317991632e-07,
      "loss": -0.0486,
      "reward": 0.7930036783218384,
      "reward_std": 0.023157700896263123,
      "reward_std_all": 0.021853262092918158,
      "rewards-max/forensics_sinkhorn_reward": 0.971463531255722,
      "rewards-max/forensics_strict_edge_reward": 0.46695858240127563,
      "rewards-max/forensics_tiou_reward": 0.9604404270648956,
      "rewards/forensics_sinkhorn_reward": 0.8046203255653381,
      "rewards/forensics_strict_edge_reward": 0.05813055485486984,
      "rewards/forensics_tiou_reward": 0.6309422552585602,
      "step": 831
    },
    {
      "completion_length": 20.515625,
      "epoch": 3.481171548117155,
      "grad_norm": 8.570734024047852,
      "kl": 0.3154296875,
      "learning_rate": 1.307531380753138e-07,
      "loss": -0.0515,
      "reward": 0.7502815127372742,
      "reward_std": 0.03713281825184822,
      "reward_std_all": 0.047417568042874336,
      "rewards-max/forensics_sinkhorn_reward": 0.9378682971000671,
      "rewards-max/forensics_strict_edge_reward": 0.2202961416915059,
      "rewards-max/forensics_tiou_reward": 0.8392433226108551,
      "rewards/forensics_sinkhorn_reward": 0.7076318860054016,
      "rewards/forensics_strict_edge_reward": 0.02391781349433586,
      "rewards/forensics_tiou_reward": 0.4808010458946228,
      "step": 832
    },
    {
      "completion_length": 21.859375,
      "epoch": 3.485355648535565,
      "grad_norm": 5.717413902282715,
      "kl": 0.42578125,
      "learning_rate": 1.297071129707113e-07,
      "loss": 0.0028,
      "reward": 0.7744128108024597,
      "reward_std": 0.03347266651690006,
      "reward_std_all": 0.033892713487148285,
      "rewards-max/forensics_sinkhorn_reward": 0.9780336320400238,
      "rewards-max/forensics_strict_edge_reward": 0.8617841005325317,
      "rewards-max/forensics_tiou_reward": 0.9747595489025116,
      "rewards/forensics_sinkhorn_reward": 0.7583439946174622,
      "rewards/forensics_strict_edge_reward": 0.055273519828915596,
      "rewards/forensics_tiou_reward": 0.5650992393493652,
      "step": 833
    },
    {
      "completion_length": 20.9375,
      "epoch": 3.489539748953975,
      "grad_norm": 7.590490341186523,
      "kl": 0.38671875,
      "learning_rate": 1.2866108786610878e-07,
      "loss": -0.0881,
      "reward": 0.7669942677021027,
      "reward_std": 0.0470981951802969,
      "reward_std_all": 0.04522604122757912,
      "rewards-max/forensics_sinkhorn_reward": 0.9812867939472198,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9807295501232147,
      "rewards/forensics_sinkhorn_reward": 0.760725200176239,
      "rewards/forensics_strict_edge_reward": 0.07336471788585186,
      "rewards/forensics_tiou_reward": 0.6009339392185211,
      "step": 834
    },
    {
      "completion_length": 22.21875,
      "epoch": 3.493723849372385,
      "grad_norm": 6.2654571533203125,
      "kl": 0.3798828125,
      "learning_rate": 1.2761506276150627e-07,
      "loss": -0.0119,
      "reward": 0.722489058971405,
      "reward_std": 0.060434890910983086,
      "reward_std_all": 0.07287270575761795,
      "rewards-max/forensics_sinkhorn_reward": 0.9861713945865631,
      "rewards-max/forensics_strict_edge_reward": 0.47726282477378845,
      "rewards-max/forensics_tiou_reward": 0.9902045428752899,
      "rewards/forensics_sinkhorn_reward": 0.730731338262558,
      "rewards/forensics_strict_edge_reward": 0.07891185767948627,
      "rewards/forensics_tiou_reward": 0.5727764368057251,
      "step": 835
    },
    {
      "completion_length": 20.703125,
      "epoch": 3.497907949790795,
      "grad_norm": 7.592506408691406,
      "kl": 0.4306640625,
      "learning_rate": 1.2656903765690378e-07,
      "loss": 0.1002,
      "reward": 0.7798541188240051,
      "reward_std": 0.032431396190077066,
      "reward_std_all": 0.03296216391026974,
      "rewards-max/forensics_sinkhorn_reward": 0.9749158024787903,
      "rewards-max/forensics_strict_edge_reward": 0.1869880985468626,
      "rewards-max/forensics_tiou_reward": 0.9546585381031036,
      "rewards/forensics_sinkhorn_reward": 0.7506495118141174,
      "rewards/forensics_strict_edge_reward": 0.025665449968073517,
      "rewards/forensics_tiou_reward": 0.543654203414917,
      "step": 836
    },
    {
      "completion_length": 20.765625,
      "epoch": 3.502092050209205,
      "grad_norm": 6.746850490570068,
      "kl": 0.396484375,
      "learning_rate": 1.2552301255230124e-07,
      "loss": 0.064,
      "reward": 0.7626737654209137,
      "reward_std": 0.045451585203409195,
      "reward_std_all": 0.045465302653610706,
      "rewards-max/forensics_sinkhorn_reward": 0.9698953628540039,
      "rewards-max/forensics_strict_edge_reward": 0.6302271485328674,
      "rewards-max/forensics_tiou_reward": 0.9509344100952148,
      "rewards/forensics_sinkhorn_reward": 0.7715773582458496,
      "rewards/forensics_strict_edge_reward": 0.05181986838579178,
      "rewards/forensics_tiou_reward": 0.5756885707378387,
      "step": 837
    },
    {
      "completion_length": 21.75,
      "epoch": 3.506276150627615,
      "grad_norm": 6.6057844161987305,
      "kl": 0.298828125,
      "learning_rate": 1.2447698744769875e-07,
      "loss": 0.0146,
      "reward": 0.798139899969101,
      "reward_std": 0.011664937541354448,
      "reward_std_all": 0.01259975298307836,
      "rewards-max/forensics_sinkhorn_reward": 0.9706458151340485,
      "rewards-max/forensics_strict_edge_reward": 0.3977042883634567,
      "rewards-max/forensics_tiou_reward": 0.9503660500049591,
      "rewards/forensics_sinkhorn_reward": 0.7620263993740082,
      "rewards/forensics_strict_edge_reward": 0.024498244281858206,
      "rewards/forensics_tiou_reward": 0.5572780072689056,
      "step": 838
    },
    {
      "completion_length": 20.203125,
      "epoch": 3.510460251046025,
      "grad_norm": 8.400137901306152,
      "kl": 0.3740234375,
      "learning_rate": 1.2343096234309624e-07,
      "loss": 0.0896,
      "reward": 0.763768345117569,
      "reward_std": 0.05611922964453697,
      "reward_std_all": 0.052741196006536484,
      "rewards-max/forensics_sinkhorn_reward": 0.9831834137439728,
      "rewards-max/forensics_strict_edge_reward": 0.7047859728336334,
      "rewards-max/forensics_tiou_reward": 0.9866345524787903,
      "rewards/forensics_sinkhorn_reward": 0.7743991315364838,
      "rewards/forensics_strict_edge_reward": 0.04843908175826073,
      "rewards/forensics_tiou_reward": 0.6093569695949554,
      "step": 839
    },
    {
      "completion_length": 20.34375,
      "epoch": 3.514644351464435,
      "grad_norm": 6.569969654083252,
      "kl": 0.3134765625,
      "learning_rate": 1.2238493723849372e-07,
      "loss": -0.016,
      "reward": 0.7806909382343292,
      "reward_std": 0.038910748437047005,
      "reward_std_all": 0.03622715175151825,
      "rewards-max/forensics_sinkhorn_reward": 0.9659965932369232,
      "rewards-max/forensics_strict_edge_reward": 0.7709200084209442,
      "rewards-max/forensics_tiou_reward": 0.9463485479354858,
      "rewards/forensics_sinkhorn_reward": 0.7722267508506775,
      "rewards/forensics_strict_edge_reward": 0.09167592786252499,
      "rewards/forensics_tiou_reward": 0.5679045021533966,
      "step": 840
    },
    {
      "completion_length": 21.390625,
      "epoch": 3.518828451882845,
      "grad_norm": 7.0353474617004395,
      "kl": 0.580078125,
      "learning_rate": 1.213389121338912e-07,
      "loss": -0.0011,
      "reward": 0.7703302800655365,
      "reward_std": 0.050030190497636795,
      "reward_std_all": 0.046651531010866165,
      "rewards-max/forensics_sinkhorn_reward": 0.9770655632019043,
      "rewards-max/forensics_strict_edge_reward": 0.6076286435127258,
      "rewards-max/forensics_tiou_reward": 0.9708060324192047,
      "rewards/forensics_sinkhorn_reward": 0.775580495595932,
      "rewards/forensics_strict_edge_reward": 0.07371993735432625,
      "rewards/forensics_tiou_reward": 0.6129489541053772,
      "step": 841
    },
    {
      "completion_length": 20.96875,
      "epoch": 3.523012552301255,
      "grad_norm": 7.824171543121338,
      "kl": 0.5478515625,
      "learning_rate": 1.202928870292887e-07,
      "loss": 0.1835,
      "reward": 0.7271508872509003,
      "reward_std": 0.08046222850680351,
      "reward_std_all": 0.08234849758446217,
      "rewards-max/forensics_sinkhorn_reward": 0.9647011160850525,
      "rewards-max/forensics_strict_edge_reward": 0.5862489938735962,
      "rewards-max/forensics_tiou_reward": 0.9441747665405273,
      "rewards/forensics_sinkhorn_reward": 0.6870145201683044,
      "rewards/forensics_strict_edge_reward": 0.06317300163209438,
      "rewards/forensics_tiou_reward": 0.4985201954841614,
      "step": 842
    },
    {
      "completion_length": 21.390625,
      "epoch": 3.5271966527196654,
      "grad_norm": 7.625837326049805,
      "kl": 0.34765625,
      "learning_rate": 1.1924686192468618e-07,
      "loss": -0.0309,
      "reward": 0.785200446844101,
      "reward_std": 0.03527220385149121,
      "reward_std_all": 0.032460304675623775,
      "rewards-max/forensics_sinkhorn_reward": 0.9847527742385864,
      "rewards-max/forensics_strict_edge_reward": 0.6816803216934204,
      "rewards-max/forensics_tiou_reward": 0.9850155413150787,
      "rewards/forensics_sinkhorn_reward": 0.7665051817893982,
      "rewards/forensics_strict_edge_reward": 0.05678229779005051,
      "rewards/forensics_tiou_reward": 0.5826897621154785,
      "step": 843
    },
    {
      "completion_length": 20.859375,
      "epoch": 3.5313807531380754,
      "grad_norm": 6.6798906326293945,
      "kl": 0.369140625,
      "learning_rate": 1.1820083682008368e-07,
      "loss": -0.0732,
      "reward": 0.7449203133583069,
      "reward_std": 0.054612721665762365,
      "reward_std_all": 0.05673543456941843,
      "rewards-max/forensics_sinkhorn_reward": 0.97989821434021,
      "rewards-max/forensics_strict_edge_reward": 0.5724648237228394,
      "rewards-max/forensics_tiou_reward": 0.9742261171340942,
      "rewards/forensics_sinkhorn_reward": 0.6986980438232422,
      "rewards/forensics_strict_edge_reward": 0.04525510594248772,
      "rewards/forensics_tiou_reward": 0.4826228618621826,
      "step": 844
    },
    {
      "completion_length": 20.296875,
      "epoch": 3.5355648535564854,
      "grad_norm": 8.055230140686035,
      "kl": 0.3916015625,
      "learning_rate": 1.1715481171548117e-07,
      "loss": -0.0764,
      "reward": 0.7807978987693787,
      "reward_std": 0.031036758795380592,
      "reward_std_all": 0.030631967820227146,
      "rewards-max/forensics_sinkhorn_reward": 0.9770084321498871,
      "rewards-max/forensics_strict_edge_reward": 0.5072369128465652,
      "rewards-max/forensics_tiou_reward": 0.9711781740188599,
      "rewards/forensics_sinkhorn_reward": 0.7802511751651764,
      "rewards/forensics_strict_edge_reward": 0.04321100376546383,
      "rewards/forensics_tiou_reward": 0.5566487014293671,
      "step": 845
    },
    {
      "completion_length": 22.109375,
      "epoch": 3.5397489539748954,
      "grad_norm": 6.755296230316162,
      "kl": 0.4169921875,
      "learning_rate": 1.1610878661087866e-07,
      "loss": -0.0455,
      "reward": 0.7815963923931122,
      "reward_std": 0.029314449056982994,
      "reward_std_all": 0.029361823573708534,
      "rewards-max/forensics_sinkhorn_reward": 0.986637145280838,
      "rewards-max/forensics_strict_edge_reward": 0.5992296040058136,
      "rewards-max/forensics_tiou_reward": 0.9927358031272888,
      "rewards/forensics_sinkhorn_reward": 0.7355653047561646,
      "rewards/forensics_strict_edge_reward": 0.0416741706430912,
      "rewards/forensics_tiou_reward": 0.4990478456020355,
      "step": 846
    },
    {
      "completion_length": 22.578125,
      "epoch": 3.5439330543933054,
      "grad_norm": 7.926720142364502,
      "kl": 0.37890625,
      "learning_rate": 1.1506276150627614e-07,
      "loss": 0.0821,
      "reward": 0.7787018120288849,
      "reward_std": 0.01775384647771716,
      "reward_std_all": 0.02546537248417735,
      "rewards-max/forensics_sinkhorn_reward": 0.9596348106861115,
      "rewards-max/forensics_strict_edge_reward": 0.05748993158340454,
      "rewards-max/forensics_tiou_reward": 0.914965033531189,
      "rewards/forensics_sinkhorn_reward": 0.7434799373149872,
      "rewards/forensics_strict_edge_reward": 0.006075514480471611,
      "rewards/forensics_tiou_reward": 0.5388328731060028,
      "step": 847
    },
    {
      "completion_length": 20.421875,
      "epoch": 3.5481171548117154,
      "grad_norm": 6.227687358856201,
      "kl": 0.3203125,
      "learning_rate": 1.1401673640167364e-07,
      "loss": -0.0409,
      "reward": 0.7839473485946655,
      "reward_std": 0.03706360689830035,
      "reward_std_all": 0.034487914643250406,
      "rewards-max/forensics_sinkhorn_reward": 0.9873031675815582,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9949080049991608,
      "rewards/forensics_sinkhorn_reward": 0.7727050185203552,
      "rewards/forensics_strict_edge_reward": 0.048353519290685654,
      "rewards/forensics_tiou_reward": 0.5543364882469177,
      "step": 848
    },
    {
      "completion_length": 19.34375,
      "epoch": 3.5523012552301254,
      "grad_norm": 6.514847278594971,
      "kl": 0.333984375,
      "learning_rate": 1.1297071129707113e-07,
      "loss": -0.0231,
      "reward": 0.7737645506858826,
      "reward_std": 0.036380300763994455,
      "reward_std_all": 0.035849787294864655,
      "rewards-max/forensics_sinkhorn_reward": 0.9794026017189026,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.978238046169281,
      "rewards/forensics_sinkhorn_reward": 0.7847700715065002,
      "rewards/forensics_strict_edge_reward": 0.07498354464769363,
      "rewards/forensics_tiou_reward": 0.5648369044065475,
      "step": 849
    },
    {
      "completion_length": 22.34375,
      "epoch": 3.5564853556485354,
      "grad_norm": 6.042262554168701,
      "kl": 0.30859375,
      "learning_rate": 1.1192468619246861e-07,
      "loss": -0.0805,
      "reward": 0.7843236029148102,
      "reward_std": 0.03101955633610487,
      "reward_std_all": 0.029779743403196335,
      "rewards-max/forensics_sinkhorn_reward": 0.9789455533027649,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9723805487155914,
      "rewards/forensics_sinkhorn_reward": 0.7634691894054413,
      "rewards/forensics_strict_edge_reward": 0.049892956390976906,
      "rewards/forensics_tiou_reward": 0.6147513389587402,
      "step": 850
    },
    {
      "completion_length": 19.609375,
      "epoch": 3.560669456066946,
      "grad_norm": 5.736618995666504,
      "kl": 0.3759765625,
      "learning_rate": 1.108786610878661e-07,
      "loss": -0.0139,
      "reward": 0.7547795176506042,
      "reward_std": 0.008396062068641186,
      "reward_std_all": 0.029925414361059666,
      "rewards-max/forensics_sinkhorn_reward": 0.9662537276744843,
      "rewards-max/forensics_strict_edge_reward": 0.47987377643585205,
      "rewards-max/forensics_tiou_reward": 0.9439395666122437,
      "rewards/forensics_sinkhorn_reward": 0.7312137484550476,
      "rewards/forensics_strict_edge_reward": 0.04743169620633125,
      "rewards/forensics_tiou_reward": 0.48344098031520844,
      "step": 851
    },
    {
      "completion_length": 21.375,
      "epoch": 3.564853556485356,
      "grad_norm": 7.211823463439941,
      "kl": 0.4677734375,
      "learning_rate": 1.0983263598326359e-07,
      "loss": -0.033,
      "reward": 0.790633499622345,
      "reward_std": 0.026628728723153472,
      "reward_std_all": 0.02504017297178507,
      "rewards-max/forensics_sinkhorn_reward": 0.9773403108119965,
      "rewards-max/forensics_strict_edge_reward": 0.5776711404323578,
      "rewards-max/forensics_tiou_reward": 0.9676825702190399,
      "rewards/forensics_sinkhorn_reward": 0.7906615734100342,
      "rewards/forensics_strict_edge_reward": 0.06068609654903412,
      "rewards/forensics_tiou_reward": 0.6176574528217316,
      "step": 852
    },
    {
      "completion_length": 21.21875,
      "epoch": 3.569037656903766,
      "grad_norm": 6.237295627593994,
      "kl": 0.3681640625,
      "learning_rate": 1.0878661087866109e-07,
      "loss": -0.0987,
      "reward": 0.7335759401321411,
      "reward_std": 0.02341845165938139,
      "reward_std_all": 0.04853636771440506,
      "rewards-max/forensics_sinkhorn_reward": 0.9857366681098938,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9920102655887604,
      "rewards/forensics_sinkhorn_reward": 0.7353686392307281,
      "rewards/forensics_strict_edge_reward": 0.09909800440073013,
      "rewards/forensics_tiou_reward": 0.571147620677948,
      "step": 853
    },
    {
      "completion_length": 21.515625,
      "epoch": 3.573221757322176,
      "grad_norm": 6.556199550628662,
      "kl": 0.357421875,
      "learning_rate": 1.0774058577405857e-07,
      "loss": -0.0002,
      "reward": 0.7683398723602295,
      "reward_std": 0.03376845782622695,
      "reward_std_all": 0.03954123891890049,
      "rewards-max/forensics_sinkhorn_reward": 0.9774373471736908,
      "rewards-max/forensics_strict_edge_reward": 0.6839397251605988,
      "rewards-max/forensics_tiou_reward": 0.9658563733100891,
      "rewards/forensics_sinkhorn_reward": 0.7924290895462036,
      "rewards/forensics_strict_edge_reward": 0.08436150476336479,
      "rewards/forensics_tiou_reward": 0.625551700592041,
      "step": 854
    },
    {
      "completion_length": 19.875,
      "epoch": 3.577405857740586,
      "grad_norm": 52.39748001098633,
      "kl": 2.70703125,
      "learning_rate": 1.0669456066945606e-07,
      "loss": 0.1668,
      "reward": 0.7812404632568359,
      "reward_std": 0.03224177658557892,
      "reward_std_all": 0.031002883799374104,
      "rewards-max/forensics_sinkhorn_reward": 0.9872876107692719,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9951553642749786,
      "rewards/forensics_sinkhorn_reward": 0.7906217575073242,
      "rewards/forensics_strict_edge_reward": 0.09927648678421974,
      "rewards/forensics_tiou_reward": 0.6214622855186462,
      "step": 855
    },
    {
      "completion_length": 23.109375,
      "epoch": 3.581589958158996,
      "grad_norm": 6.293537139892578,
      "kl": 0.47265625,
      "learning_rate": 1.0564853556485355e-07,
      "loss": 0.0259,
      "reward": 0.6824949383735657,
      "reward_std": 0.08865854516625404,
      "reward_std_all": 0.10735969990491867,
      "rewards-max/forensics_sinkhorn_reward": 0.9751774966716766,
      "rewards-max/forensics_strict_edge_reward": 0.4935835525393486,
      "rewards-max/forensics_tiou_reward": 0.9639614820480347,
      "rewards/forensics_sinkhorn_reward": 0.660162478685379,
      "rewards/forensics_strict_edge_reward": 0.04164979886263609,
      "rewards/forensics_tiou_reward": 0.4177294969558716,
      "step": 856
    },
    {
      "completion_length": 20.859375,
      "epoch": 3.585774058577406,
      "grad_norm": 8.402742385864258,
      "kl": 0.6484375,
      "learning_rate": 1.0460251046025103e-07,
      "loss": 0.0204,
      "reward": 0.793113112449646,
      "reward_std": 0.02350048179505393,
      "reward_std_all": 0.02212631213478744,
      "rewards-max/forensics_sinkhorn_reward": 0.9756025671958923,
      "rewards-max/forensics_strict_edge_reward": 0.49686121940612793,
      "rewards-max/forensics_tiou_reward": 0.9640983045101166,
      "rewards/forensics_sinkhorn_reward": 0.8039531409740448,
      "rewards/forensics_strict_edge_reward": 0.06059381924569607,
      "rewards/forensics_tiou_reward": 0.7110955715179443,
      "step": 857
    },
    {
      "completion_length": 23.40625,
      "epoch": 3.589958158995816,
      "grad_norm": 5.996176242828369,
      "kl": 0.314453125,
      "learning_rate": 1.0355648535564853e-07,
      "loss": -0.0106,
      "reward": 0.7584109008312225,
      "reward_std": 0.03953291568905115,
      "reward_std_all": 0.042586054652929306,
      "rewards-max/forensics_sinkhorn_reward": 0.9651203751564026,
      "rewards-max/forensics_strict_edge_reward": 0.5892792046070099,
      "rewards-max/forensics_tiou_reward": 0.9423749446868896,
      "rewards/forensics_sinkhorn_reward": 0.7040466368198395,
      "rewards/forensics_strict_edge_reward": 0.04645971953868866,
      "rewards/forensics_tiou_reward": 0.5382167398929596,
      "step": 858
    },
    {
      "completion_length": 22.125,
      "epoch": 3.594142259414226,
      "grad_norm": 7.100523471832275,
      "kl": 0.3828125,
      "learning_rate": 1.0251046025104603e-07,
      "loss": 0.0324,
      "reward": 0.7717085778713226,
      "reward_std": 0.04386649024672806,
      "reward_std_all": 0.04188571032136679,
      "rewards-max/forensics_sinkhorn_reward": 0.9785946607589722,
      "rewards-max/forensics_strict_edge_reward": 0.6065306663513184,
      "rewards-max/forensics_tiou_reward": 0.9717656970024109,
      "rewards/forensics_sinkhorn_reward": 0.7373590767383575,
      "rewards/forensics_strict_edge_reward": 0.06142057850956917,
      "rewards/forensics_tiou_reward": 0.5815621018409729,
      "step": 859
    },
    {
      "completion_length": 21.328125,
      "epoch": 3.598326359832636,
      "grad_norm": 6.391192436218262,
      "kl": 0.3828125,
      "learning_rate": 1.0146443514644352e-07,
      "loss": -0.0113,
      "reward": 0.7513794302940369,
      "reward_std": 0.027093371376395226,
      "reward_std_all": 0.043821665458381176,
      "rewards-max/forensics_sinkhorn_reward": 0.9527095556259155,
      "rewards-max/forensics_strict_edge_reward": 0.32062673568725586,
      "rewards-max/forensics_tiou_reward": 0.9055466055870056,
      "rewards/forensics_sinkhorn_reward": 0.7064478695392609,
      "rewards/forensics_strict_edge_reward": 0.028105035424232483,
      "rewards/forensics_tiou_reward": 0.49586138129234314,
      "step": 860
    },
    {
      "completion_length": 20.65625,
      "epoch": 3.602510460251046,
      "grad_norm": 7.600602626800537,
      "kl": 0.4013671875,
      "learning_rate": 1.00418410041841e-07,
      "loss": 0.0189,
      "reward": 0.7396951019763947,
      "reward_std": 0.05713210068643093,
      "reward_std_all": 0.0640023685991764,
      "rewards-max/forensics_sinkhorn_reward": 0.972918689250946,
      "rewards-max/forensics_strict_edge_reward": 0.6448149085044861,
      "rewards-max/forensics_tiou_reward": 0.9570133686065674,
      "rewards/forensics_sinkhorn_reward": 0.7303118407726288,
      "rewards/forensics_strict_edge_reward": 0.04970650002360344,
      "rewards/forensics_tiou_reward": 0.5174512565135956,
      "step": 861
    },
    {
      "completion_length": 21.234375,
      "epoch": 3.606694560669456,
      "grad_norm": 7.203541278839111,
      "kl": 0.48828125,
      "learning_rate": 9.937238493723849e-08,
      "loss": -0.0187,
      "reward": 0.7607592642307281,
      "reward_std": 0.05510133132338524,
      "reward_std_all": 0.05611497536301613,
      "rewards-max/forensics_sinkhorn_reward": 0.9817054271697998,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9815137684345245,
      "rewards/forensics_sinkhorn_reward": 0.7859169244766235,
      "rewards/forensics_strict_edge_reward": 0.05768861994147301,
      "rewards/forensics_tiou_reward": 0.6454792320728302,
      "step": 862
    },
    {
      "completion_length": 21.125,
      "epoch": 3.6108786610878663,
      "grad_norm": 8.556885719299316,
      "kl": 0.4169921875,
      "learning_rate": 9.832635983263599e-08,
      "loss": 0.141,
      "reward": 0.7545671761035919,
      "reward_std": 0.049796829000115395,
      "reward_std_all": 0.056219181045889854,
      "rewards-max/forensics_sinkhorn_reward": 0.980170726776123,
      "rewards-max/forensics_strict_edge_reward": 0.5488366484642029,
      "rewards-max/forensics_tiou_reward": 0.9718975722789764,
      "rewards/forensics_sinkhorn_reward": 0.7654818296432495,
      "rewards/forensics_strict_edge_reward": 0.07805514335632324,
      "rewards/forensics_tiou_reward": 0.5747528374195099,
      "step": 863
    },
    {
      "completion_length": 20.78125,
      "epoch": 3.6150627615062763,
      "grad_norm": 7.122236728668213,
      "kl": 0.361328125,
      "learning_rate": 9.728033472803348e-08,
      "loss": -0.0491,
      "reward": 0.7994965314865112,
      "reward_std": 0.012496836483478546,
      "reward_std_all": 0.012560777831822634,
      "rewards-max/forensics_sinkhorn_reward": 0.984750360250473,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9877823293209076,
      "rewards/forensics_sinkhorn_reward": 0.8463111519813538,
      "rewards/forensics_strict_edge_reward": 0.07124008052051067,
      "rewards/forensics_tiou_reward": 0.6880960166454315,
      "step": 864
    },
    {
      "completion_length": 21.90625,
      "epoch": 3.6192468619246863,
      "grad_norm": 6.812478065490723,
      "kl": 0.3271484375,
      "learning_rate": 9.623430962343096e-08,
      "loss": 0.067,
      "reward": 0.7475108504295349,
      "reward_std": 0.03114207088947296,
      "reward_std_all": 0.04832272604107857,
      "rewards-max/forensics_sinkhorn_reward": 0.9728695452213287,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9635051786899567,
      "rewards/forensics_sinkhorn_reward": 0.6602641046047211,
      "rewards/forensics_strict_edge_reward": 0.0422752033919096,
      "rewards/forensics_tiou_reward": 0.4905662387609482,
      "step": 865
    },
    {
      "completion_length": 21.3125,
      "epoch": 3.6234309623430963,
      "grad_norm": 6.467611312866211,
      "kl": 0.37109375,
      "learning_rate": 9.518828451882845e-08,
      "loss": -0.0946,
      "reward": 0.8032860457897186,
      "reward_std": 0.0055163620272651315,
      "reward_std_all": 0.0065496559254825115,
      "rewards-max/forensics_sinkhorn_reward": 0.9844024479389191,
      "rewards-max/forensics_strict_edge_reward": 0.8464817404747009,
      "rewards-max/forensics_tiou_reward": 0.9884862303733826,
      "rewards/forensics_sinkhorn_reward": 0.8231526017189026,
      "rewards/forensics_strict_edge_reward": 0.0907612033188343,
      "rewards/forensics_tiou_reward": 0.6389492154121399,
      "step": 866
    },
    {
      "completion_length": 23.03125,
      "epoch": 3.6276150627615062,
      "grad_norm": 6.1647820472717285,
      "kl": 0.33984375,
      "learning_rate": 9.414225941422594e-08,
      "loss": -0.015,
      "reward": 0.7778642773628235,
      "reward_std": 0.03591958060860634,
      "reward_std_all": 0.03489107824862003,
      "rewards-max/forensics_sinkhorn_reward": 0.953989714384079,
      "rewards-max/forensics_strict_edge_reward": 0.3088050112128258,
      "rewards-max/forensics_tiou_reward": 0.9004086852073669,
      "rewards/forensics_sinkhorn_reward": 0.7440447807312012,
      "rewards/forensics_strict_edge_reward": 0.024340699426829815,
      "rewards/forensics_tiou_reward": 0.5518870651721954,
      "step": 867
    },
    {
      "completion_length": 20.828125,
      "epoch": 3.6317991631799162,
      "grad_norm": 9.549535751342773,
      "kl": 0.3212890625,
      "learning_rate": 9.309623430962342e-08,
      "loss": -0.007,
      "reward": 0.7217711508274078,
      "reward_std": 0.041668918915092945,
      "reward_std_all": 0.06259272806346416,
      "rewards-max/forensics_sinkhorn_reward": 0.9618065357208252,
      "rewards-max/forensics_strict_edge_reward": 0.16429969295859337,
      "rewards-max/forensics_tiou_reward": 0.9364049434661865,
      "rewards/forensics_sinkhorn_reward": 0.6612739264965057,
      "rewards/forensics_strict_edge_reward": 0.00789602193981409,
      "rewards/forensics_tiou_reward": 0.40415677428245544,
      "step": 868
    },
    {
      "completion_length": 21.28125,
      "epoch": 3.6359832635983262,
      "grad_norm": 6.220776557922363,
      "kl": 0.396484375,
      "learning_rate": 9.205020920502092e-08,
      "loss": 0.0446,
      "reward": 0.7719596028327942,
      "reward_std": 0.03830297291278839,
      "reward_std_all": 0.04035197012126446,
      "rewards-max/forensics_sinkhorn_reward": 0.986011803150177,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.9928981065750122,
      "rewards/forensics_sinkhorn_reward": 0.7511630654335022,
      "rewards/forensics_strict_edge_reward": 0.08508909866213799,
      "rewards/forensics_tiou_reward": 0.5525599420070648,
      "step": 869
    },
    {
      "completion_length": 21.640625,
      "epoch": 3.6401673640167362,
      "grad_norm": 7.289152145385742,
      "kl": 0.556640625,
      "learning_rate": 9.100418410041841e-08,
      "loss": 0.0761,
      "reward": 0.7556264698505402,
      "reward_std": 0.06292005069553852,
      "reward_std_all": 0.05958045460283756,
      "rewards-max/forensics_sinkhorn_reward": 0.9843832552433014,
      "rewards-max/forensics_strict_edge_reward": 0.3720540702342987,
      "rewards-max/forensics_tiou_reward": 0.9855262339115143,
      "rewards/forensics_sinkhorn_reward": 0.7283509969711304,
      "rewards/forensics_strict_edge_reward": 0.03737936355173588,
      "rewards/forensics_tiou_reward": 0.522524505853653,
      "step": 870
    },
    {
      "completion_length": 23.71875,
      "epoch": 3.6443514644351467,
      "grad_norm": 8.254105567932129,
      "kl": 0.33984375,
      "learning_rate": 8.99581589958159e-08,
      "loss": -0.0014,
      "reward": 0.7597565054893494,
      "reward_std": 0.055921645835042,
      "reward_std_all": 0.0536324679851532,
      "rewards-max/forensics_sinkhorn_reward": 0.975153237581253,
      "rewards-max/forensics_strict_edge_reward": 0.37205515056848526,
      "rewards-max/forensics_tiou_reward": 0.9661868214607239,
      "rewards/forensics_sinkhorn_reward": 0.7597532570362091,
      "rewards/forensics_strict_edge_reward": 0.035700018517673016,
      "rewards/forensics_tiou_reward": 0.5733521282672882,
      "step": 871
    },
    {
      "completion_length": 20.140625,
      "epoch": 3.6485355648535567,
      "grad_norm": 7.1885175704956055,
      "kl": 0.40234375,
      "learning_rate": 8.891213389121338e-08,
      "loss": 0.0406,
      "reward": 0.7851685285568237,
      "reward_std": 0.031524356454610825,
      "reward_std_all": 0.03001427836716175,
      "rewards-max/forensics_sinkhorn_reward": 0.9766497611999512,
      "rewards-max/forensics_strict_edge_reward": 0.4903825744986534,
      "rewards-max/forensics_tiou_reward": 0.9663173258304596,
      "rewards/forensics_sinkhorn_reward": 0.7710455060005188,
      "rewards/forensics_strict_edge_reward": 0.0347547335550189,
      "rewards/forensics_tiou_reward": 0.5699832737445831,
      "step": 872
    },
    {
      "completion_length": 21.265625,
      "epoch": 3.6527196652719667,
      "grad_norm": 5.998320579528809,
      "kl": 0.419921875,
      "learning_rate": 8.786610878661087e-08,
      "loss": 0.0815,
      "reward": 0.7706986367702484,
      "reward_std": 0.04105710983276367,
      "reward_std_all": 0.042388392612338066,
      "rewards-max/forensics_sinkhorn_reward": 0.9680118262767792,
      "rewards-max/forensics_strict_edge_reward": 0.412410169839859,
      "rewards-max/forensics_tiou_reward": 0.9435513317584991,
      "rewards/forensics_sinkhorn_reward": 0.7355424761772156,
      "rewards/forensics_strict_edge_reward": 0.0324600525200367,
      "rewards/forensics_tiou_reward": 0.5137965381145477,
      "step": 873
    },
    {
      "completion_length": 21.28125,
      "epoch": 3.6569037656903767,
      "grad_norm": 5.767929553985596,
      "kl": 0.3681640625,
      "learning_rate": 8.682008368200837e-08,
      "loss": 0.0507,
      "reward": 0.7706647515296936,
      "reward_std": 0.04670395702123642,
      "reward_std_all": 0.04499511793255806,
      "rewards-max/forensics_sinkhorn_reward": 0.978893518447876,
      "rewards-max/forensics_strict_edge_reward": 0.6284831166267395,
      "rewards-max/forensics_tiou_reward": 0.9732743203639984,
      "rewards/forensics_sinkhorn_reward": 0.7617741525173187,
      "rewards/forensics_strict_edge_reward": 0.05186423193663359,
      "rewards/forensics_tiou_reward": 0.5317265093326569,
      "step": 874
    },
    {
      "completion_length": 23.0,
      "epoch": 3.6610878661087867,
      "grad_norm": 5.412528038024902,
      "kl": 0.419921875,
      "learning_rate": 8.577405857740585e-08,
      "loss": -0.0698,
      "reward": 0.7790212631225586,
      "reward_std": 0.02901752246543765,
      "reward_std_all": 0.030404019635170698,
      "rewards-max/forensics_sinkhorn_reward": 0.9772390127182007,
      "rewards-max/forensics_strict_edge_reward": 0.4966079294681549,
      "rewards-max/forensics_tiou_reward": 0.9695866703987122,
      "rewards/forensics_sinkhorn_reward": 0.7534878253936768,
      "rewards/forensics_strict_edge_reward": 0.046401677653193474,
      "rewards/forensics_tiou_reward": 0.5578988790512085,
      "step": 875
    },
    {
      "completion_length": 23.703125,
      "epoch": 3.6652719665271967,
      "grad_norm": 8.599923133850098,
      "kl": 0.330078125,
      "learning_rate": 8.472803347280334e-08,
      "loss": 0.0308,
      "reward": 0.7580432891845703,
      "reward_std": 0.041326697915792465,
      "reward_std_all": 0.0446377657353878,
      "rewards-max/forensics_sinkhorn_reward": 0.9842247068881989,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9881595075130463,
      "rewards/forensics_sinkhorn_reward": 0.7310466170310974,
      "rewards/forensics_strict_edge_reward": 0.09270398691296577,
      "rewards/forensics_tiou_reward": 0.608019232749939,
      "step": 876
    },
    {
      "completion_length": 21.859375,
      "epoch": 3.6694560669456067,
      "grad_norm": 5.025030612945557,
      "kl": 0.4189453125,
      "learning_rate": 8.368200836820083e-08,
      "loss": 0.0352,
      "reward": 0.7581458389759064,
      "reward_std": 0.04753698967397213,
      "reward_std_all": 0.048303086310625076,
      "rewards-max/forensics_sinkhorn_reward": 0.9865192174911499,
      "rewards-max/forensics_strict_edge_reward": 0.8751733303070068,
      "rewards-max/forensics_tiou_reward": 0.9936908781528473,
      "rewards/forensics_sinkhorn_reward": 0.7517117261886597,
      "rewards/forensics_strict_edge_reward": 0.06530969589948654,
      "rewards/forensics_tiou_reward": 0.5794811248779297,
      "step": 877
    },
    {
      "completion_length": 21.09375,
      "epoch": 3.6736401673640167,
      "grad_norm": 6.1806416511535645,
      "kl": 0.3017578125,
      "learning_rate": 8.263598326359831e-08,
      "loss": -0.092,
      "reward": 0.7873682081699371,
      "reward_std": 0.018095793668180704,
      "reward_std_all": 0.019743729382753372,
      "rewards-max/forensics_sinkhorn_reward": 0.9884220063686371,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9979516267776489,
      "rewards/forensics_sinkhorn_reward": 0.8227468729019165,
      "rewards/forensics_strict_edge_reward": 0.0708276554942131,
      "rewards/forensics_tiou_reward": 0.6297380328178406,
      "step": 878
    },
    {
      "completion_length": 21.96875,
      "epoch": 3.6778242677824267,
      "grad_norm": 8.14848804473877,
      "kl": 0.3369140625,
      "learning_rate": 8.158995815899581e-08,
      "loss": 0.0524,
      "reward": 0.7639562785625458,
      "reward_std": 0.06453662365674973,
      "reward_std_all": 0.05836103484034538,
      "rewards-max/forensics_sinkhorn_reward": 0.985606461763382,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9920602142810822,
      "rewards/forensics_sinkhorn_reward": 0.766172468662262,
      "rewards/forensics_strict_edge_reward": 0.13536332547664642,
      "rewards/forensics_tiou_reward": 0.6103930771350861,
      "step": 879
    },
    {
      "completion_length": 18.875,
      "epoch": 3.6820083682008367,
      "grad_norm": 8.084939956665039,
      "kl": 0.361328125,
      "learning_rate": 8.05439330543933e-08,
      "loss": 0.0187,
      "reward": 0.7703819274902344,
      "reward_std": 0.02786724455654621,
      "reward_std_all": 0.03494709171354771,
      "rewards-max/forensics_sinkhorn_reward": 0.9768639802932739,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9705743789672852,
      "rewards/forensics_sinkhorn_reward": 0.7545058131217957,
      "rewards/forensics_strict_edge_reward": 0.04238693043589592,
      "rewards/forensics_tiou_reward": 0.5314140617847443,
      "step": 880
    },
    {
      "completion_length": 20.640625,
      "epoch": 3.6861924686192467,
      "grad_norm": 6.6174211502075195,
      "kl": 0.3994140625,
      "learning_rate": 7.949790794979078e-08,
      "loss": -0.0503,
      "reward": 0.7969388663768768,
      "reward_std": 0.011756077175959945,
      "reward_std_all": 0.012508607236668468,
      "rewards-max/forensics_sinkhorn_reward": 0.9786114990711212,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9723331928253174,
      "rewards/forensics_sinkhorn_reward": 0.7937192320823669,
      "rewards/forensics_strict_edge_reward": 0.05312778428196907,
      "rewards/forensics_tiou_reward": 0.5714493095874786,
      "step": 881
    },
    {
      "completion_length": 20.96875,
      "epoch": 3.6903765690376567,
      "grad_norm": 5.950526237487793,
      "kl": 0.580078125,
      "learning_rate": 7.845188284518828e-08,
      "loss": 0.0392,
      "reward": 0.7739195227622986,
      "reward_std": 0.039675742387771606,
      "reward_std_all": 0.038585469126701355,
      "rewards-max/forensics_sinkhorn_reward": 0.9518237113952637,
      "rewards-max/forensics_strict_edge_reward": 0.3443178366869688,
      "rewards-max/forensics_tiou_reward": 0.8972444832324982,
      "rewards/forensics_sinkhorn_reward": 0.7388255000114441,
      "rewards/forensics_strict_edge_reward": 0.042297631909605116,
      "rewards/forensics_tiou_reward": 0.49723775684833527,
      "step": 882
    },
    {
      "completion_length": 20.34375,
      "epoch": 3.694560669456067,
      "grad_norm": 5.794158935546875,
      "kl": 0.353515625,
      "learning_rate": 7.740585774058577e-08,
      "loss": -0.0632,
      "reward": 0.7938783764839172,
      "reward_std": 0.01902886270545423,
      "reward_std_all": 0.019009796902537346,
      "rewards-max/forensics_sinkhorn_reward": 0.9405443668365479,
      "rewards-max/forensics_strict_edge_reward": 0.21986748091876507,
      "rewards-max/forensics_tiou_reward": 0.8727991878986359,
      "rewards/forensics_sinkhorn_reward": 0.7741972506046295,
      "rewards/forensics_strict_edge_reward": 0.023731912369839847,
      "rewards/forensics_tiou_reward": 0.5579538941383362,
      "step": 883
    },
    {
      "completion_length": 25.3125,
      "epoch": 3.698744769874477,
      "grad_norm": 6.803926944732666,
      "kl": 0.3935546875,
      "learning_rate": 7.635983263598327e-08,
      "loss": 0.031,
      "reward": 0.7485273480415344,
      "reward_std": 0.056365231052041054,
      "reward_std_all": 0.05753487907350063,
      "rewards-max/forensics_sinkhorn_reward": 0.9679552912712097,
      "rewards-max/forensics_strict_edge_reward": 0.38803622126579285,
      "rewards-max/forensics_tiou_reward": 0.945787787437439,
      "rewards/forensics_sinkhorn_reward": 0.665643721818924,
      "rewards/forensics_strict_edge_reward": 0.03379737213253975,
      "rewards/forensics_tiou_reward": 0.4091191589832306,
      "step": 884
    },
    {
      "completion_length": 21.59375,
      "epoch": 3.702928870292887,
      "grad_norm": 15.903660774230957,
      "kl": 1.4423828125,
      "learning_rate": 7.531380753138076e-08,
      "loss": 0.1912,
      "reward": 0.7695009708404541,
      "reward_std": 0.04493527486920357,
      "reward_std_all": 0.04322293400764465,
      "rewards-max/forensics_sinkhorn_reward": 0.952721357345581,
      "rewards-max/forensics_strict_edge_reward": 0.23506871983408928,
      "rewards-max/forensics_tiou_reward": 0.9087719023227692,
      "rewards/forensics_sinkhorn_reward": 0.7095358073711395,
      "rewards/forensics_strict_edge_reward": 0.0206580120138824,
      "rewards/forensics_tiou_reward": 0.4911951422691345,
      "step": 885
    },
    {
      "completion_length": 22.90625,
      "epoch": 3.707112970711297,
      "grad_norm": 5.529568195343018,
      "kl": 0.3505859375,
      "learning_rate": 7.426778242677824e-08,
      "loss": 0.0271,
      "reward": 0.786264181137085,
      "reward_std": 0.03470964357256889,
      "reward_std_all": 0.0317159341648221,
      "rewards-max/forensics_sinkhorn_reward": 0.9729107916355133,
      "rewards-max/forensics_strict_edge_reward": 0.5731385350227356,
      "rewards-max/forensics_tiou_reward": 0.9592877328395844,
      "rewards/forensics_sinkhorn_reward": 0.7888064384460449,
      "rewards/forensics_strict_edge_reward": 0.07161924056708813,
      "rewards/forensics_tiou_reward": 0.6114684343338013,
      "step": 886
    },
    {
      "completion_length": 22.09375,
      "epoch": 3.711297071129707,
      "grad_norm": 8.373762130737305,
      "kl": 0.548828125,
      "learning_rate": 7.322175732217573e-08,
      "loss": -0.0148,
      "reward": 0.7596898376941681,
      "reward_std": 0.03403059672564268,
      "reward_std_all": 0.0425318144261837,
      "rewards-max/forensics_sinkhorn_reward": 0.9586130082607269,
      "rewards-max/forensics_strict_edge_reward": 0.34921763837337494,
      "rewards-max/forensics_tiou_reward": 0.9258637428283691,
      "rewards/forensics_sinkhorn_reward": 0.7178095281124115,
      "rewards/forensics_strict_edge_reward": 0.026693038176745176,
      "rewards/forensics_tiou_reward": 0.5011775195598602,
      "step": 887
    },
    {
      "completion_length": 21.359375,
      "epoch": 3.715481171548117,
      "grad_norm": 6.430829048156738,
      "kl": 0.4609375,
      "learning_rate": 7.217573221757322e-08,
      "loss": 0.0353,
      "reward": 0.7635276615619659,
      "reward_std": 0.021849549375474453,
      "reward_std_all": 0.033767903223633766,
      "rewards-max/forensics_sinkhorn_reward": 0.9686779379844666,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9542789459228516,
      "rewards/forensics_sinkhorn_reward": 0.7496678233146667,
      "rewards/forensics_strict_edge_reward": 0.08063905872404575,
      "rewards/forensics_tiou_reward": 0.5855255722999573,
      "step": 888
    },
    {
      "completion_length": 21.859375,
      "epoch": 3.719665271966527,
      "grad_norm": 5.325613975524902,
      "kl": 0.287109375,
      "learning_rate": 7.112970711297071e-08,
      "loss": -0.088,
      "reward": 0.7990563809871674,
      "reward_std": 0.011389000865165144,
      "reward_std_all": 0.012187893269583583,
      "rewards-max/forensics_sinkhorn_reward": 0.9732378721237183,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.958406537771225,
      "rewards/forensics_sinkhorn_reward": 0.8010808825492859,
      "rewards/forensics_strict_edge_reward": 0.043497174978256226,
      "rewards/forensics_tiou_reward": 0.6375670433044434,
      "step": 889
    },
    {
      "completion_length": 22.625,
      "epoch": 3.723849372384937,
      "grad_norm": 7.929487228393555,
      "kl": 0.439453125,
      "learning_rate": 7.00836820083682e-08,
      "loss": 0.0255,
      "reward": 0.7836456894874573,
      "reward_std": 0.026909911073744297,
      "reward_std_all": 0.026805981993675232,
      "rewards-max/forensics_sinkhorn_reward": 0.9797812104225159,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9756677150726318,
      "rewards/forensics_sinkhorn_reward": 0.7677489519119263,
      "rewards/forensics_strict_edge_reward": 0.05836751963943243,
      "rewards/forensics_tiou_reward": 0.5836620628833771,
      "step": 890
    },
    {
      "completion_length": 22.359375,
      "epoch": 3.7280334728033475,
      "grad_norm": 5.429123401641846,
      "kl": 0.4541015625,
      "learning_rate": 6.903765690376569e-08,
      "loss": 0.0906,
      "reward": 0.773957759141922,
      "reward_std": 0.055207643657922745,
      "reward_std_all": 0.049695439636707306,
      "rewards-max/forensics_sinkhorn_reward": 0.9761244654655457,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9662552177906036,
      "rewards/forensics_sinkhorn_reward": 0.7594234645366669,
      "rewards/forensics_strict_edge_reward": 0.0592382587492466,
      "rewards/forensics_tiou_reward": 0.6041634678840637,
      "step": 891
    },
    {
      "completion_length": 22.375,
      "epoch": 3.7322175732217575,
      "grad_norm": 6.137353420257568,
      "kl": 0.4033203125,
      "learning_rate": 6.799163179916317e-08,
      "loss": 0.1433,
      "reward": 0.7728733718395233,
      "reward_std": 0.0369331706315279,
      "reward_std_all": 0.03695731144398451,
      "rewards-max/forensics_sinkhorn_reward": 0.962954044342041,
      "rewards-max/forensics_strict_edge_reward": 0.41141651570796967,
      "rewards-max/forensics_tiou_reward": 0.9372681379318237,
      "rewards/forensics_sinkhorn_reward": 0.7211558520793915,
      "rewards/forensics_strict_edge_reward": 0.03774504456669092,
      "rewards/forensics_tiou_reward": 0.514416515827179,
      "step": 892
    },
    {
      "completion_length": 22.546875,
      "epoch": 3.7364016736401675,
      "grad_norm": 6.130643367767334,
      "kl": 0.3349609375,
      "learning_rate": 6.694560669456066e-08,
      "loss": -0.066,
      "reward": 0.7881884574890137,
      "reward_std": 0.02477646851912141,
      "reward_std_all": 0.02468746481463313,
      "rewards-max/forensics_sinkhorn_reward": 0.964708536863327,
      "rewards-max/forensics_strict_edge_reward": 0.35364001989364624,
      "rewards-max/forensics_tiou_reward": 0.9347608685493469,
      "rewards/forensics_sinkhorn_reward": 0.7125232517719269,
      "rewards/forensics_strict_edge_reward": 0.027481209486722946,
      "rewards/forensics_tiou_reward": 0.49516770243644714,
      "step": 893
    },
    {
      "completion_length": 21.0,
      "epoch": 3.7405857740585775,
      "grad_norm": 7.528328895568848,
      "kl": 0.365234375,
      "learning_rate": 6.589958158995816e-08,
      "loss": 0.0567,
      "reward": 0.7737675309181213,
      "reward_std": 0.04970436356961727,
      "reward_std_all": 0.04565174877643585,
      "rewards-max/forensics_sinkhorn_reward": 0.9834705889225006,
      "rewards-max/forensics_strict_edge_reward": 0.5439433008432388,
      "rewards-max/forensics_tiou_reward": 0.9849909543991089,
      "rewards/forensics_sinkhorn_reward": 0.7759446203708649,
      "rewards/forensics_strict_edge_reward": 0.054997390136122704,
      "rewards/forensics_tiou_reward": 0.615102082490921,
      "step": 894
    },
    {
      "completion_length": 19.84375,
      "epoch": 3.7447698744769875,
      "grad_norm": 5.303900241851807,
      "kl": 0.3662109375,
      "learning_rate": 6.485355648535565e-08,
      "loss": 0.0471,
      "reward": 0.7953224182128906,
      "reward_std": 0.013789401622489095,
      "reward_std_all": 0.014351377729326487,
      "rewards-max/forensics_sinkhorn_reward": 0.9850244224071503,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.988519936800003,
      "rewards/forensics_sinkhorn_reward": 0.8081231415271759,
      "rewards/forensics_strict_edge_reward": 0.07303699478507042,
      "rewards/forensics_tiou_reward": 0.6481592655181885,
      "step": 895
    },
    {
      "completion_length": 23.21875,
      "epoch": 3.7489539748953975,
      "grad_norm": 6.38375997543335,
      "kl": 0.384765625,
      "learning_rate": 6.380753138075313e-08,
      "loss": -0.0265,
      "reward": 0.7796012759208679,
      "reward_std": 0.009357830742374063,
      "reward_std_all": 0.018225962994620204,
      "rewards-max/forensics_sinkhorn_reward": 0.9712633490562439,
      "rewards-max/forensics_strict_edge_reward": 0.546856515109539,
      "rewards-max/forensics_tiou_reward": 0.9560700058937073,
      "rewards/forensics_sinkhorn_reward": 0.7769063413143158,
      "rewards/forensics_strict_edge_reward": 0.038248843513429165,
      "rewards/forensics_tiou_reward": 0.6275678277015686,
      "step": 896
    },
    {
      "completion_length": 20.171875,
      "epoch": 3.7531380753138075,
      "grad_norm": 7.273210048675537,
      "kl": 0.44140625,
      "learning_rate": 6.276150627615062e-08,
      "loss": -0.0155,
      "reward": 0.7838308215141296,
      "reward_std": 0.025339381769299507,
      "reward_std_all": 0.02589194942265749,
      "rewards-max/forensics_sinkhorn_reward": 0.9726966023445129,
      "rewards-max/forensics_strict_edge_reward": 0.48156414926052094,
      "rewards-max/forensics_tiou_reward": 0.9565418660640717,
      "rewards/forensics_sinkhorn_reward": 0.7812590897083282,
      "rewards/forensics_strict_edge_reward": 0.06192254927009344,
      "rewards/forensics_tiou_reward": 0.583928257226944,
      "step": 897
    },
    {
      "completion_length": 20.8125,
      "epoch": 3.7573221757322175,
      "grad_norm": 9.170605659484863,
      "kl": 0.4052734375,
      "learning_rate": 6.171548117154812e-08,
      "loss": 0.0958,
      "reward": 0.7682062089443207,
      "reward_std": 0.021406730636954308,
      "reward_std_all": 0.031731558963656425,
      "rewards-max/forensics_sinkhorn_reward": 0.9870971143245697,
      "rewards-max/forensics_strict_edge_reward": 0.6115650832653046,
      "rewards-max/forensics_tiou_reward": 0.9934929013252258,
      "rewards/forensics_sinkhorn_reward": 0.7619370818138123,
      "rewards/forensics_strict_edge_reward": 0.06099823396652937,
      "rewards/forensics_tiou_reward": 0.566888153553009,
      "step": 898
    },
    {
      "completion_length": 21.546875,
      "epoch": 3.7615062761506275,
      "grad_norm": 7.769206523895264,
      "kl": 0.4755859375,
      "learning_rate": 6.06694560669456e-08,
      "loss": 0.029,
      "reward": 0.7987031638622284,
      "reward_std": 0.009094479028135538,
      "reward_std_all": 0.01030259719118476,
      "rewards-max/forensics_sinkhorn_reward": 0.9848673343658447,
      "rewards-max/forensics_strict_edge_reward": 0.8029135167598724,
      "rewards-max/forensics_tiou_reward": 0.9884189963340759,
      "rewards/forensics_sinkhorn_reward": 0.7829923629760742,
      "rewards/forensics_strict_edge_reward": 0.04780944809317589,
      "rewards/forensics_tiou_reward": 0.5658355355262756,
      "step": 899
    },
    {
      "completion_length": 20.8125,
      "epoch": 3.7656903765690375,
      "grad_norm": 6.374766826629639,
      "kl": 0.3525390625,
      "learning_rate": 5.962343096234309e-08,
      "loss": 0.0384,
      "reward": 0.7733091115951538,
      "reward_std": 0.04023545212112367,
      "reward_std_all": 0.03913666575681418,
      "rewards-max/forensics_sinkhorn_reward": 0.982498288154602,
      "rewards-max/forensics_strict_edge_reward": 0.4811747968196869,
      "rewards-max/forensics_tiou_reward": 0.9788790941238403,
      "rewards/forensics_sinkhorn_reward": 0.7383466958999634,
      "rewards/forensics_strict_edge_reward": 0.030861565843224525,
      "rewards/forensics_tiou_reward": 0.5269446969032288,
      "step": 900
    },
    {
      "completion_length": 21.015625,
      "epoch": 3.7698744769874475,
      "grad_norm": 7.377803802490234,
      "kl": 0.4189453125,
      "learning_rate": 5.8577405857740584e-08,
      "loss": 0.026,
      "reward": 0.7630637586116791,
      "reward_std": 0.04809105861932039,
      "reward_std_all": 0.04757104627788067,
      "rewards-max/forensics_sinkhorn_reward": 0.9746378362178802,
      "rewards-max/forensics_strict_edge_reward": 0.36345113813877106,
      "rewards-max/forensics_tiou_reward": 0.9576417505741119,
      "rewards/forensics_sinkhorn_reward": 0.7596631348133087,
      "rewards/forensics_strict_edge_reward": 0.025913486257195473,
      "rewards/forensics_tiou_reward": 0.5709406137466431,
      "step": 901
    },
    {
      "completion_length": 19.4375,
      "epoch": 3.7740585774058575,
      "grad_norm": 7.268104553222656,
      "kl": 0.3740234375,
      "learning_rate": 5.753138075313807e-08,
      "loss": -0.0132,
      "reward": 0.7784149646759033,
      "reward_std": 0.021199456183239818,
      "reward_std_all": 0.02482433430850506,
      "rewards-max/forensics_sinkhorn_reward": 0.9787925779819489,
      "rewards-max/forensics_strict_edge_reward": 0.8187680542469025,
      "rewards-max/forensics_tiou_reward": 0.9760490953922272,
      "rewards/forensics_sinkhorn_reward": 0.789418637752533,
      "rewards/forensics_strict_edge_reward": 0.07868117466568947,
      "rewards/forensics_tiou_reward": 0.5760338604450226,
      "step": 902
    },
    {
      "completion_length": 21.359375,
      "epoch": 3.778242677824268,
      "grad_norm": 6.557408809661865,
      "kl": 0.3369140625,
      "learning_rate": 5.6485355648535564e-08,
      "loss": -0.0346,
      "reward": 0.7874855399131775,
      "reward_std": 0.03162119071930647,
      "reward_std_all": 0.02945199515670538,
      "rewards-max/forensics_sinkhorn_reward": 0.9772136509418488,
      "rewards-max/forensics_strict_edge_reward": 0.43032702803611755,
      "rewards-max/forensics_tiou_reward": 0.9604094326496124,
      "rewards/forensics_sinkhorn_reward": 0.7792631983757019,
      "rewards/forensics_strict_edge_reward": 0.04180179536342621,
      "rewards/forensics_tiou_reward": 0.6229110658168793,
      "step": 903
    },
    {
      "completion_length": 22.25,
      "epoch": 3.782426778242678,
      "grad_norm": 8.234094619750977,
      "kl": 0.3037109375,
      "learning_rate": 5.543933054393305e-08,
      "loss": -0.0322,
      "reward": 0.7727275788784027,
      "reward_std": 0.03179319575428963,
      "reward_std_all": 0.03294277470558882,
      "rewards-max/forensics_sinkhorn_reward": 0.9769859910011292,
      "rewards-max/forensics_strict_edge_reward": 0.638425350189209,
      "rewards-max/forensics_tiou_reward": 0.9687303006649017,
      "rewards/forensics_sinkhorn_reward": 0.771912693977356,
      "rewards/forensics_strict_edge_reward": 0.06505362316966057,
      "rewards/forensics_tiou_reward": 0.6213214993476868,
      "step": 904
    },
    {
      "completion_length": 20.78125,
      "epoch": 3.786610878661088,
      "grad_norm": 6.7507548332214355,
      "kl": 0.3427734375,
      "learning_rate": 5.439330543933054e-08,
      "loss": -0.0149,
      "reward": 0.7590585947036743,
      "reward_std": 0.03202237328514457,
      "reward_std_all": 0.03961179871112108,
      "rewards-max/forensics_sinkhorn_reward": 0.9789204597473145,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9734581708908081,
      "rewards/forensics_sinkhorn_reward": 0.6897910535335541,
      "rewards/forensics_strict_edge_reward": 0.04659699648618698,
      "rewards/forensics_tiou_reward": 0.4592123329639435,
      "step": 905
    },
    {
      "completion_length": 20.453125,
      "epoch": 3.790794979079498,
      "grad_norm": 7.238613128662109,
      "kl": 0.39453125,
      "learning_rate": 5.334728033472803e-08,
      "loss": 0.0295,
      "reward": 0.7743119895458221,
      "reward_std": 0.02542463317513466,
      "reward_std_all": 0.028545977547764778,
      "rewards-max/forensics_sinkhorn_reward": 0.9702469408512115,
      "rewards-max/forensics_strict_edge_reward": 0.5192448049783707,
      "rewards-max/forensics_tiou_reward": 0.9469376802444458,
      "rewards/forensics_sinkhorn_reward": 0.7476673722267151,
      "rewards/forensics_strict_edge_reward": 0.032027105800807476,
      "rewards/forensics_tiou_reward": 0.5621857941150665,
      "step": 906
    },
    {
      "completion_length": 23.546875,
      "epoch": 3.794979079497908,
      "grad_norm": 6.023604393005371,
      "kl": 0.4033203125,
      "learning_rate": 5.2301255230125516e-08,
      "loss": 0.0599,
      "reward": 0.7725533246994019,
      "reward_std": 0.03655780479311943,
      "reward_std_all": 0.03680396359413862,
      "rewards-max/forensics_sinkhorn_reward": 0.9731765389442444,
      "rewards-max/forensics_strict_edge_reward": 0.3031960055232048,
      "rewards-max/forensics_tiou_reward": 0.957500547170639,
      "rewards/forensics_sinkhorn_reward": 0.712145984172821,
      "rewards/forensics_strict_edge_reward": 0.028284598141908646,
      "rewards/forensics_tiou_reward": 0.5392096936702728,
      "step": 907
    },
    {
      "completion_length": 21.921875,
      "epoch": 3.799163179916318,
      "grad_norm": 6.780850887298584,
      "kl": 0.3271484375,
      "learning_rate": 5.1255230125523015e-08,
      "loss": -0.09,
      "reward": 0.7715269923210144,
      "reward_std": 0.046367661154363304,
      "reward_std_all": 0.04422419739421457,
      "rewards-max/forensics_sinkhorn_reward": 0.9734910130500793,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9525603950023651,
      "rewards/forensics_sinkhorn_reward": 0.7676040828227997,
      "rewards/forensics_strict_edge_reward": 0.029627835378050804,
      "rewards/forensics_tiou_reward": 0.5766097009181976,
      "step": 908
    },
    {
      "completion_length": 21.15625,
      "epoch": 3.803347280334728,
      "grad_norm": 6.296234130859375,
      "kl": 0.458984375,
      "learning_rate": 5.02092050209205e-08,
      "loss": -0.0329,
      "reward": 0.7936193943023682,
      "reward_std": 0.021566685289144516,
      "reward_std_all": 0.020728925243020058,
      "rewards-max/forensics_sinkhorn_reward": 0.976821094751358,
      "rewards-max/forensics_strict_edge_reward": 0.5543488264083862,
      "rewards-max/forensics_tiou_reward": 0.9633354842662811,
      "rewards/forensics_sinkhorn_reward": 0.8000734448432922,
      "rewards/forensics_strict_edge_reward": 0.04962198622524738,
      "rewards/forensics_tiou_reward": 0.6520968973636627,
      "step": 909
    },
    {
      "completion_length": 20.390625,
      "epoch": 3.8075313807531384,
      "grad_norm": 7.600253105163574,
      "kl": 0.48046875,
      "learning_rate": 4.9163179916317995e-08,
      "loss": -0.0943,
      "reward": 0.8002680242061615,
      "reward_std": 0.011073954403400421,
      "reward_std_all": 0.011348814005032182,
      "rewards-max/forensics_sinkhorn_reward": 0.9834206104278564,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9834804534912109,
      "rewards/forensics_sinkhorn_reward": 0.8135630488395691,
      "rewards/forensics_strict_edge_reward": 0.08809396252036095,
      "rewards/forensics_tiou_reward": 0.6662209033966064,
      "step": 910
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.8117154811715483,
      "grad_norm": 5.432400703430176,
      "kl": 0.3291015625,
      "learning_rate": 4.811715481171548e-08,
      "loss": 0.0089,
      "reward": 0.7817391753196716,
      "reward_std": 0.02521622460335493,
      "reward_std_all": 0.026630804874002934,
      "rewards-max/forensics_sinkhorn_reward": 0.9687249958515167,
      "rewards-max/forensics_strict_edge_reward": 0.4245914816856384,
      "rewards-max/forensics_tiou_reward": 0.946835458278656,
      "rewards/forensics_sinkhorn_reward": 0.7681543827056885,
      "rewards/forensics_strict_edge_reward": 0.03966616280376911,
      "rewards/forensics_tiou_reward": 0.5772763788700104,
      "step": 911
    },
    {
      "completion_length": 21.953125,
      "epoch": 3.8158995815899583,
      "grad_norm": 6.277388095855713,
      "kl": 0.5263671875,
      "learning_rate": 4.707112970711297e-08,
      "loss": -0.0162,
      "reward": 0.7721473574638367,
      "reward_std": 0.013626741711050272,
      "reward_std_all": 0.025008161435835063,
      "rewards-max/forensics_sinkhorn_reward": 0.9572634994983673,
      "rewards-max/forensics_strict_edge_reward": 0.3208436332643032,
      "rewards-max/forensics_tiou_reward": 0.9163148701190948,
      "rewards/forensics_sinkhorn_reward": 0.7711840569972992,
      "rewards/forensics_strict_edge_reward": 0.02259895158931613,
      "rewards/forensics_tiou_reward": 0.554050475358963,
      "step": 912
    },
    {
      "completion_length": 22.484375,
      "epoch": 3.8200836820083683,
      "grad_norm": 5.685723781585693,
      "kl": 0.51171875,
      "learning_rate": 4.602510460251046e-08,
      "loss": 0.0296,
      "reward": 0.769888162612915,
      "reward_std": 0.04334385599941015,
      "reward_std_all": 0.04391372762620449,
      "rewards-max/forensics_sinkhorn_reward": 0.9827496111392975,
      "rewards-max/forensics_strict_edge_reward": 0.6448149085044861,
      "rewards-max/forensics_tiou_reward": 0.9829088151454926,
      "rewards/forensics_sinkhorn_reward": 0.7819450497627258,
      "rewards/forensics_strict_edge_reward": 0.050578197464346886,
      "rewards/forensics_tiou_reward": 0.5847512483596802,
      "step": 913
    },
    {
      "completion_length": 21.59375,
      "epoch": 3.8242677824267783,
      "grad_norm": 6.452849864959717,
      "kl": 0.3935546875,
      "learning_rate": 4.497907949790795e-08,
      "loss": 0.0592,
      "reward": 0.7671347856521606,
      "reward_std": 0.046695027500391006,
      "reward_std_all": 0.045313555747270584,
      "rewards-max/forensics_sinkhorn_reward": 0.9809447228908539,
      "rewards-max/forensics_strict_edge_reward": 0.5561007261276245,
      "rewards-max/forensics_tiou_reward": 0.9770415127277374,
      "rewards/forensics_sinkhorn_reward": 0.7304202318191528,
      "rewards/forensics_strict_edge_reward": 0.025565327145159245,
      "rewards/forensics_tiou_reward": 0.5123473405838013,
      "step": 914
    },
    {
      "completion_length": 20.6875,
      "epoch": 3.8284518828451883,
      "grad_norm": 5.720353603363037,
      "kl": 0.3076171875,
      "learning_rate": 4.393305439330543e-08,
      "loss": 0.0795,
      "reward": 0.7731289863586426,
      "reward_std": 0.056341566145420074,
      "reward_std_all": 0.050665585324168205,
      "rewards-max/forensics_sinkhorn_reward": 0.9837536513805389,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9863038957118988,
      "rewards/forensics_sinkhorn_reward": 0.7697893977165222,
      "rewards/forensics_strict_edge_reward": 0.03977447748184204,
      "rewards/forensics_tiou_reward": 0.5772079527378082,
      "step": 915
    },
    {
      "completion_length": 21.171875,
      "epoch": 3.8326359832635983,
      "grad_norm": 7.341462135314941,
      "kl": 0.3798828125,
      "learning_rate": 4.2887029288702926e-08,
      "loss": -0.0854,
      "reward": 0.7517715394496918,
      "reward_std": 0.023517904803156853,
      "reward_std_all": 0.03895224258303642,
      "rewards-max/forensics_sinkhorn_reward": 0.9639405906200409,
      "rewards-max/forensics_strict_edge_reward": 0.4874049425125122,
      "rewards-max/forensics_tiou_reward": 0.9343878924846649,
      "rewards/forensics_sinkhorn_reward": 0.7325447797775269,
      "rewards/forensics_strict_edge_reward": 0.06012299936264753,
      "rewards/forensics_tiou_reward": 0.5414828360080719,
      "step": 916
    },
    {
      "completion_length": 22.75,
      "epoch": 3.8368200836820083,
      "grad_norm": 6.593169212341309,
      "kl": 0.3525390625,
      "learning_rate": 4.184100418410041e-08,
      "loss": 0.0021,
      "reward": 0.7763587534427643,
      "reward_std": 0.051039488054811954,
      "reward_std_all": 0.045757452957332134,
      "rewards-max/forensics_sinkhorn_reward": 0.9712321162223816,
      "rewards-max/forensics_strict_edge_reward": 0.5533042848110199,
      "rewards-max/forensics_tiou_reward": 0.955597311258316,
      "rewards/forensics_sinkhorn_reward": 0.7525222897529602,
      "rewards/forensics_strict_edge_reward": 0.0721457414329052,
      "rewards/forensics_tiou_reward": 0.6006427109241486,
      "step": 917
    },
    {
      "completion_length": 24.359375,
      "epoch": 3.8410041841004183,
      "grad_norm": 6.639044284820557,
      "kl": 0.3779296875,
      "learning_rate": 4.0794979079497906e-08,
      "loss": 0.0585,
      "reward": 0.7830823659896851,
      "reward_std": 0.0356265427544713,
      "reward_std_all": 0.03324328502640128,
      "rewards-max/forensics_sinkhorn_reward": 0.9786992073059082,
      "rewards-max/forensics_strict_edge_reward": 0.3885071277618408,
      "rewards-max/forensics_tiou_reward": 0.9701057970523834,
      "rewards/forensics_sinkhorn_reward": 0.7674503922462463,
      "rewards/forensics_strict_edge_reward": 0.03038989193737507,
      "rewards/forensics_tiou_reward": 0.6194583773612976,
      "step": 918
    },
    {
      "completion_length": 21.0,
      "epoch": 3.8451882845188283,
      "grad_norm": 9.724174499511719,
      "kl": 0.48046875,
      "learning_rate": 3.974895397489539e-08,
      "loss": -0.023,
      "reward": 0.7769968509674072,
      "reward_std": 0.034247429110109806,
      "reward_std_all": 0.03377912752330303,
      "rewards-max/forensics_sinkhorn_reward": 0.9839899241924286,
      "rewards-max/forensics_strict_edge_reward": 0.5842200070619583,
      "rewards-max/forensics_tiou_reward": 0.9865109324455261,
      "rewards/forensics_sinkhorn_reward": 0.7489953637123108,
      "rewards/forensics_strict_edge_reward": 0.052348085679113865,
      "rewards/forensics_tiou_reward": 0.49787674844264984,
      "step": 919
    },
    {
      "completion_length": 21.34375,
      "epoch": 3.8493723849372383,
      "grad_norm": 7.035332679748535,
      "kl": 0.4072265625,
      "learning_rate": 3.8702928870292885e-08,
      "loss": 0.0464,
      "reward": 0.7474292516708374,
      "reward_std": 0.023741175420582294,
      "reward_std_all": 0.04350261762738228,
      "rewards-max/forensics_sinkhorn_reward": 0.9734739661216736,
      "rewards-max/forensics_strict_edge_reward": 0.499885693192482,
      "rewards-max/forensics_tiou_reward": 0.9574880003929138,
      "rewards/forensics_sinkhorn_reward": 0.7157348096370697,
      "rewards/forensics_strict_edge_reward": 0.054318346083164215,
      "rewards/forensics_tiou_reward": 0.5115878134965897,
      "step": 920
    },
    {
      "completion_length": 20.875,
      "epoch": 3.8535564853556483,
      "grad_norm": 6.901394844055176,
      "kl": 0.474609375,
      "learning_rate": 3.765690376569038e-08,
      "loss": -0.002,
      "reward": 0.7820601761341095,
      "reward_std": 0.028743933886289597,
      "reward_std_all": 0.028460470959544182,
      "rewards-max/forensics_sinkhorn_reward": 0.9811565279960632,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9797742664813995,
      "rewards/forensics_sinkhorn_reward": 0.7500917613506317,
      "rewards/forensics_strict_edge_reward": 0.09332128241658211,
      "rewards/forensics_tiou_reward": 0.5581042468547821,
      "step": 921
    },
    {
      "completion_length": 21.90625,
      "epoch": 3.8577405857740583,
      "grad_norm": 7.845523834228516,
      "kl": 0.3173828125,
      "learning_rate": 3.6610878661087864e-08,
      "loss": 0.0296,
      "reward": 0.752355307340622,
      "reward_std": 0.06275140680372715,
      "reward_std_all": 0.06011340767145157,
      "rewards-max/forensics_sinkhorn_reward": 0.9769957065582275,
      "rewards-max/forensics_strict_edge_reward": 0.5844750255346298,
      "rewards-max/forensics_tiou_reward": 0.9703932106494904,
      "rewards/forensics_sinkhorn_reward": 0.7084117531776428,
      "rewards/forensics_strict_edge_reward": 0.06278932839632034,
      "rewards/forensics_tiou_reward": 0.500402182340622,
      "step": 922
    },
    {
      "completion_length": 21.0,
      "epoch": 3.8619246861924688,
      "grad_norm": 6.228543758392334,
      "kl": 0.37109375,
      "learning_rate": 3.556485355648536e-08,
      "loss": 0.0409,
      "reward": 0.7894739508628845,
      "reward_std": 0.01731693744659424,
      "reward_std_all": 0.018579866737127304,
      "rewards-max/forensics_sinkhorn_reward": 0.9849860966205597,
      "rewards-max/forensics_strict_edge_reward": 0.8106843829154968,
      "rewards-max/forensics_tiou_reward": 0.988239049911499,
      "rewards/forensics_sinkhorn_reward": 0.7839265167713165,
      "rewards/forensics_strict_edge_reward": 0.06063554063439369,
      "rewards/forensics_tiou_reward": 0.5957899689674377,
      "step": 923
    },
    {
      "completion_length": 21.984375,
      "epoch": 3.8661087866108788,
      "grad_norm": 6.1257643699646,
      "kl": 0.375,
      "learning_rate": 3.4518828451882844e-08,
      "loss": -0.019,
      "reward": 0.7948763966560364,
      "reward_std": 0.019133077701553702,
      "reward_std_all": 0.018598771188408136,
      "rewards-max/forensics_sinkhorn_reward": 0.9813195765018463,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9805096387863159,
      "rewards/forensics_sinkhorn_reward": 0.7569499313831329,
      "rewards/forensics_strict_edge_reward": 0.0756860226392746,
      "rewards/forensics_tiou_reward": 0.6207950711250305,
      "step": 924
    },
    {
      "completion_length": 21.4375,
      "epoch": 3.8702928870292888,
      "grad_norm": 6.930722713470459,
      "kl": 0.40234375,
      "learning_rate": 3.347280334728033e-08,
      "loss": 0.0608,
      "reward": 0.7875148355960846,
      "reward_std": 0.028322887141257524,
      "reward_std_all": 0.02685336349532008,
      "rewards-max/forensics_sinkhorn_reward": 0.9695613086223602,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9524143934249878,
      "rewards/forensics_sinkhorn_reward": 0.7953079044818878,
      "rewards/forensics_strict_edge_reward": 0.05537444353103638,
      "rewards/forensics_tiou_reward": 0.6058211326599121,
      "step": 925
    },
    {
      "completion_length": 21.265625,
      "epoch": 3.8744769874476988,
      "grad_norm": 11.077869415283203,
      "kl": 0.6650390625,
      "learning_rate": 3.242677824267782e-08,
      "loss": -0.0012,
      "reward": 0.7773360311985016,
      "reward_std": 0.0157228687312454,
      "reward_std_all": 0.024900189600884914,
      "rewards-max/forensics_sinkhorn_reward": 0.9693482518196106,
      "rewards-max/forensics_strict_edge_reward": 0.5814422816038132,
      "rewards-max/forensics_tiou_reward": 0.9553859531879425,
      "rewards/forensics_sinkhorn_reward": 0.7850537598133087,
      "rewards/forensics_strict_edge_reward": 0.07881507463753223,
      "rewards/forensics_tiou_reward": 0.6332695782184601,
      "step": 926
    },
    {
      "completion_length": 19.953125,
      "epoch": 3.8786610878661087,
      "grad_norm": 8.76701831817627,
      "kl": 0.3701171875,
      "learning_rate": 3.138075313807531e-08,
      "loss": -0.1164,
      "reward": 0.7498525679111481,
      "reward_std": 0.024413038976490498,
      "reward_std_all": 0.03758912347257137,
      "rewards-max/forensics_sinkhorn_reward": 0.9784781634807587,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9731724858283997,
      "rewards/forensics_sinkhorn_reward": 0.7224189639091492,
      "rewards/forensics_strict_edge_reward": 0.039688488468527794,
      "rewards/forensics_tiou_reward": 0.49942803382873535,
      "step": 927
    },
    {
      "completion_length": 20.703125,
      "epoch": 3.8828451882845187,
      "grad_norm": 8.224967002868652,
      "kl": 0.3779296875,
      "learning_rate": 3.03347280334728e-08,
      "loss": 0.022,
      "reward": 0.7830944657325745,
      "reward_std": 0.041525471955537796,
      "reward_std_all": 0.03757136035710573,
      "rewards-max/forensics_sinkhorn_reward": 0.9889904856681824,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8129213750362396,
      "rewards/forensics_strict_edge_reward": 0.12323819845914841,
      "rewards/forensics_tiou_reward": 0.6652583181858063,
      "step": 928
    },
    {
      "completion_length": 21.546875,
      "epoch": 3.8870292887029287,
      "grad_norm": 6.226418972015381,
      "kl": 0.326171875,
      "learning_rate": 2.9288702928870292e-08,
      "loss": -0.0833,
      "reward": 0.7774869799613953,
      "reward_std": 0.037910377606749535,
      "reward_std_all": 0.03607029654085636,
      "rewards-max/forensics_sinkhorn_reward": 0.980808436870575,
      "rewards-max/forensics_strict_edge_reward": 0.6403803080320358,
      "rewards-max/forensics_tiou_reward": 0.9693146049976349,
      "rewards/forensics_sinkhorn_reward": 0.7719563543796539,
      "rewards/forensics_strict_edge_reward": 0.09533953201025724,
      "rewards/forensics_tiou_reward": 0.6214714050292969,
      "step": 929
    },
    {
      "completion_length": 21.234375,
      "epoch": 3.891213389121339,
      "grad_norm": 6.202601909637451,
      "kl": 0.3203125,
      "learning_rate": 2.8242677824267782e-08,
      "loss": -0.0652,
      "reward": 0.7787300050258636,
      "reward_std": 0.04499875754117966,
      "reward_std_all": 0.04150782711803913,
      "rewards-max/forensics_sinkhorn_reward": 0.987095445394516,
      "rewards-max/forensics_strict_edge_reward": 0.6317985653877258,
      "rewards-max/forensics_tiou_reward": 0.9933193624019623,
      "rewards/forensics_sinkhorn_reward": 0.7526876926422119,
      "rewards/forensics_strict_edge_reward": 0.06925630941987038,
      "rewards/forensics_tiou_reward": 0.593100905418396,
      "step": 930
    },
    {
      "completion_length": 21.78125,
      "epoch": 3.895397489539749,
      "grad_norm": 7.490749359130859,
      "kl": 0.46484375,
      "learning_rate": 2.719665271966527e-08,
      "loss": 0.0252,
      "reward": 0.6901578009128571,
      "reward_std": 0.06876347586512566,
      "reward_std_all": 0.09699785336852074,
      "rewards-max/forensics_sinkhorn_reward": 0.9629877805709839,
      "rewards-max/forensics_strict_edge_reward": 0.2972450405359268,
      "rewards-max/forensics_tiou_reward": 0.9340173900127411,
      "rewards/forensics_sinkhorn_reward": 0.6469197273254395,
      "rewards/forensics_strict_edge_reward": 0.03418272826820612,
      "rewards/forensics_tiou_reward": 0.46833835542201996,
      "step": 931
    },
    {
      "completion_length": 22.3125,
      "epoch": 3.899581589958159,
      "grad_norm": 9.415227890014648,
      "kl": 0.3369140625,
      "learning_rate": 2.6150627615062758e-08,
      "loss": -0.0973,
      "reward": 0.7677834928035736,
      "reward_std": 0.02698484784923494,
      "reward_std_all": 0.03155331173911691,
      "rewards-max/forensics_sinkhorn_reward": 0.9599414467811584,
      "rewards-max/forensics_strict_edge_reward": 0.3897523954510689,
      "rewards-max/forensics_tiou_reward": 0.9247543513774872,
      "rewards/forensics_sinkhorn_reward": 0.7494311928749084,
      "rewards/forensics_strict_edge_reward": 0.039933339692652225,
      "rewards/forensics_tiou_reward": 0.577781617641449,
      "step": 932
    },
    {
      "completion_length": 23.125,
      "epoch": 3.903765690376569,
      "grad_norm": 6.336202621459961,
      "kl": 0.29296875,
      "learning_rate": 2.510460251046025e-08,
      "loss": -0.0945,
      "reward": 0.7968009114265442,
      "reward_std": 0.00958889041794464,
      "reward_std_all": 0.011286854045465589,
      "rewards-max/forensics_sinkhorn_reward": 0.9624807238578796,
      "rewards-max/forensics_strict_edge_reward": 0.4374866336584091,
      "rewards-max/forensics_tiou_reward": 0.928275853395462,
      "rewards/forensics_sinkhorn_reward": 0.8060524165630341,
      "rewards/forensics_strict_edge_reward": 0.026884596794843674,
      "rewards/forensics_tiou_reward": 0.6353336274623871,
      "step": 933
    },
    {
      "completion_length": 21.609375,
      "epoch": 3.907949790794979,
      "grad_norm": 7.002731800079346,
      "kl": 0.4716796875,
      "learning_rate": 2.405857740585774e-08,
      "loss": 0.0348,
      "reward": 0.7436177432537079,
      "reward_std": 0.03386403899639845,
      "reward_std_all": 0.047310082241892815,
      "rewards-max/forensics_sinkhorn_reward": 0.9770370423793793,
      "rewards-max/forensics_strict_edge_reward": 0.411779360845685,
      "rewards-max/forensics_tiou_reward": 0.96800297498703,
      "rewards/forensics_sinkhorn_reward": 0.6750070452690125,
      "rewards/forensics_strict_edge_reward": 0.04103494934679475,
      "rewards/forensics_tiou_reward": 0.44782282412052155,
      "step": 934
    },
    {
      "completion_length": 22.953125,
      "epoch": 3.912133891213389,
      "grad_norm": 7.173625946044922,
      "kl": 0.375,
      "learning_rate": 2.301255230125523e-08,
      "loss": 0.0676,
      "reward": 0.7896874248981476,
      "reward_std": 0.023834384162910283,
      "reward_std_all": 0.023199791088700294,
      "rewards-max/forensics_sinkhorn_reward": 0.9699629247188568,
      "rewards-max/forensics_strict_edge_reward": 0.5207925289869308,
      "rewards-max/forensics_tiou_reward": 0.9473315179347992,
      "rewards/forensics_sinkhorn_reward": 0.7464084327220917,
      "rewards/forensics_strict_edge_reward": 0.06456232070922852,
      "rewards/forensics_tiou_reward": 0.5897029042243958,
      "step": 935
    },
    {
      "completion_length": 20.03125,
      "epoch": 3.916317991631799,
      "grad_norm": 6.17042875289917,
      "kl": 0.376953125,
      "learning_rate": 2.1966527196652717e-08,
      "loss": -0.0733,
      "reward": 0.7999315857887268,
      "reward_std": 0.0098682971438393,
      "reward_std_all": 0.010828549973666668,
      "rewards-max/forensics_sinkhorn_reward": 0.9821667671203613,
      "rewards-max/forensics_strict_edge_reward": 0.5512756779789925,
      "rewards-max/forensics_tiou_reward": 0.9771135747432709,
      "rewards/forensics_sinkhorn_reward": 0.7970440983772278,
      "rewards/forensics_strict_edge_reward": 0.07080802135169506,
      "rewards/forensics_tiou_reward": 0.6023351550102234,
      "step": 936
    },
    {
      "completion_length": 21.78125,
      "epoch": 3.920502092050209,
      "grad_norm": 6.7115020751953125,
      "kl": 0.45703125,
      "learning_rate": 2.0920502092050206e-08,
      "loss": -0.0064,
      "reward": 0.7710431218147278,
      "reward_std": 0.03325022105127573,
      "reward_std_all": 0.03673857403919101,
      "rewards-max/forensics_sinkhorn_reward": 0.986284464597702,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.992226630449295,
      "rewards/forensics_sinkhorn_reward": 0.7387131452560425,
      "rewards/forensics_strict_edge_reward": 0.05545416101813316,
      "rewards/forensics_tiou_reward": 0.5122753381729126,
      "step": 937
    },
    {
      "completion_length": 22.640625,
      "epoch": 3.924686192468619,
      "grad_norm": 8.076581954956055,
      "kl": 0.3623046875,
      "learning_rate": 1.9874476987447696e-08,
      "loss": -0.0483,
      "reward": 0.77743199467659,
      "reward_std": 0.039412698009982705,
      "reward_std_all": 0.038705334067344666,
      "rewards-max/forensics_sinkhorn_reward": 0.9831924736499786,
      "rewards-max/forensics_strict_edge_reward": 0.6976466774940491,
      "rewards-max/forensics_tiou_reward": 0.9831058084964752,
      "rewards/forensics_sinkhorn_reward": 0.76934415102005,
      "rewards/forensics_strict_edge_reward": 0.04317731782793999,
      "rewards/forensics_tiou_reward": 0.5707019567489624,
      "step": 938
    },
    {
      "completion_length": 21.328125,
      "epoch": 3.928870292887029,
      "grad_norm": 8.816224098205566,
      "kl": 0.416015625,
      "learning_rate": 1.882845188284519e-08,
      "loss": 0.0551,
      "reward": 0.7466166019439697,
      "reward_std": 0.061847299337387085,
      "reward_std_all": 0.061798345297575,
      "rewards-max/forensics_sinkhorn_reward": 0.898235410451889,
      "rewards-max/forensics_strict_edge_reward": 0.4387461101869121,
      "rewards-max/forensics_tiou_reward": 0.7953980267047882,
      "rewards/forensics_sinkhorn_reward": 0.7074858248233795,
      "rewards/forensics_strict_edge_reward": 0.10389283438416896,
      "rewards/forensics_tiou_reward": 0.48838573694229126,
      "step": 939
    },
    {
      "completion_length": 21.75,
      "epoch": 3.933054393305439,
      "grad_norm": 5.867214202880859,
      "kl": 0.400390625,
      "learning_rate": 1.778242677824268e-08,
      "loss": 0.0224,
      "reward": 0.7895545363426208,
      "reward_std": 0.02018030663020909,
      "reward_std_all": 0.020780745893716812,
      "rewards-max/forensics_sinkhorn_reward": 0.9696927666664124,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.9563805758953094,
      "rewards/forensics_sinkhorn_reward": 0.7496807277202606,
      "rewards/forensics_strict_edge_reward": 0.046576157212257385,
      "rewards/forensics_tiou_reward": 0.5098908245563507,
      "step": 940
    },
    {
      "completion_length": 20.828125,
      "epoch": 3.937238493723849,
      "grad_norm": 15.817780494689941,
      "kl": 0.33203125,
      "learning_rate": 1.6736401673640165e-08,
      "loss": -0.0224,
      "reward": 0.7899361252784729,
      "reward_std": 0.02473103115335107,
      "reward_std_all": 0.023743835277855396,
      "rewards-max/forensics_sinkhorn_reward": 0.9845629036426544,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9868596494197845,
      "rewards/forensics_sinkhorn_reward": 0.8197455704212189,
      "rewards/forensics_strict_edge_reward": 0.05599597003310919,
      "rewards/forensics_tiou_reward": 0.6576626896858215,
      "step": 941
    },
    {
      "completion_length": 24.25,
      "epoch": 3.941422594142259,
      "grad_norm": 7.633510112762451,
      "kl": 0.3232421875,
      "learning_rate": 1.5690376569037655e-08,
      "loss": -0.0736,
      "reward": 0.7791362106800079,
      "reward_std": 0.04677162319421768,
      "reward_std_all": 0.041909532621502876,
      "rewards-max/forensics_sinkhorn_reward": 0.9714894592761993,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9599566757678986,
      "rewards/forensics_sinkhorn_reward": 0.7773130238056183,
      "rewards/forensics_strict_edge_reward": 0.04910360462963581,
      "rewards/forensics_tiou_reward": 0.6007055640220642,
      "step": 942
    },
    {
      "completion_length": 20.0,
      "epoch": 3.9456066945606696,
      "grad_norm": 8.04802417755127,
      "kl": 0.3486328125,
      "learning_rate": 1.4644351464435146e-08,
      "loss": 0.006,
      "reward": 0.7373646795749664,
      "reward_std": 0.06983952224254608,
      "reward_std_all": 0.06998076476156712,
      "rewards-max/forensics_sinkhorn_reward": 0.9772091507911682,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9689030051231384,
      "rewards/forensics_sinkhorn_reward": 0.7397075891494751,
      "rewards/forensics_strict_edge_reward": 0.08080769889056683,
      "rewards/forensics_tiou_reward": 0.5181977152824402,
      "step": 943
    },
    {
      "completion_length": 21.8125,
      "epoch": 3.9497907949790796,
      "grad_norm": 6.457942008972168,
      "kl": 0.423828125,
      "learning_rate": 1.3598326359832636e-08,
      "loss": 0.0785,
      "reward": 0.7874321341514587,
      "reward_std": 0.023327446077018976,
      "reward_std_all": 0.023187921848148108,
      "rewards-max/forensics_sinkhorn_reward": 0.9657376706600189,
      "rewards-max/forensics_strict_edge_reward": 0.40060924738645554,
      "rewards-max/forensics_tiou_reward": 0.942051112651825,
      "rewards/forensics_sinkhorn_reward": 0.7769008278846741,
      "rewards/forensics_strict_edge_reward": 0.050156403332948685,
      "rewards/forensics_tiou_reward": 0.6286597549915314,
      "step": 944
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.9539748953974896,
      "grad_norm": 7.857550621032715,
      "kl": 0.294921875,
      "learning_rate": 1.2552301255230125e-08,
      "loss": 0.0899,
      "reward": 0.7181634604930878,
      "reward_std": 0.03671797551214695,
      "reward_std_all": 0.0637246835976839,
      "rewards-max/forensics_sinkhorn_reward": 0.9659228920936584,
      "rewards-max/forensics_strict_edge_reward": 0.3028031438589096,
      "rewards-max/forensics_tiou_reward": 0.9361901581287384,
      "rewards/forensics_sinkhorn_reward": 0.6440702974796295,
      "rewards/forensics_strict_edge_reward": 0.019172320142388344,
      "rewards/forensics_tiou_reward": 0.4460858255624771,
      "step": 945
    },
    {
      "completion_length": 21.59375,
      "epoch": 3.9581589958158996,
      "grad_norm": 6.300907135009766,
      "kl": 0.4228515625,
      "learning_rate": 1.1506276150627615e-08,
      "loss": 0.0273,
      "reward": 0.7725259363651276,
      "reward_std": 0.05356236547231674,
      "reward_std_all": 0.04893426317721605,
      "rewards-max/forensics_sinkhorn_reward": 0.9818710088729858,
      "rewards-max/forensics_strict_edge_reward": 0.6516932249069214,
      "rewards-max/forensics_tiou_reward": 0.9786377251148224,
      "rewards/forensics_sinkhorn_reward": 0.7294978499412537,
      "rewards/forensics_strict_edge_reward": 0.041306874714791775,
      "rewards/forensics_tiou_reward": 0.5006209015846252,
      "step": 946
    },
    {
      "completion_length": 22.625,
      "epoch": 3.9623430962343096,
      "grad_norm": 8.780258178710938,
      "kl": 0.3466796875,
      "learning_rate": 1.0460251046025103e-08,
      "loss": 0.0844,
      "reward": 0.7546687126159668,
      "reward_std": 0.03879040107131004,
      "reward_std_all": 0.04601414315402508,
      "rewards-max/forensics_sinkhorn_reward": 0.9675062894821167,
      "rewards-max/forensics_strict_edge_reward": 0.37032441794872284,
      "rewards-max/forensics_tiou_reward": 0.948256254196167,
      "rewards/forensics_sinkhorn_reward": 0.7099356651306152,
      "rewards/forensics_strict_edge_reward": 0.04184754565358162,
      "rewards/forensics_tiou_reward": 0.5217084437608719,
      "step": 947
    },
    {
      "completion_length": 20.65625,
      "epoch": 3.9665271966527196,
      "grad_norm": 8.308713912963867,
      "kl": 0.3408203125,
      "learning_rate": 9.414225941422594e-09,
      "loss": 0.085,
      "reward": 0.7572545409202576,
      "reward_std": 0.07042917888611555,
      "reward_std_all": 0.06480337958782911,
      "rewards-max/forensics_sinkhorn_reward": 0.9847083985805511,
      "rewards-max/forensics_strict_edge_reward": 0.509157819673419,
      "rewards-max/forensics_tiou_reward": 0.9864329397678375,
      "rewards/forensics_sinkhorn_reward": 0.7364358305931091,
      "rewards/forensics_strict_edge_reward": 0.04960900597507134,
      "rewards/forensics_tiou_reward": 0.5398341715335846,
      "step": 948
    },
    {
      "completion_length": 20.296875,
      "epoch": 3.9707112970711296,
      "grad_norm": 6.166962623596191,
      "kl": 0.57421875,
      "learning_rate": 8.368200836820083e-09,
      "loss": 0.1262,
      "reward": 0.7826721966266632,
      "reward_std": 0.04073979798704386,
      "reward_std_all": 0.0369992908090353,
      "rewards-max/forensics_sinkhorn_reward": 0.9752934277057648,
      "rewards-max/forensics_strict_edge_reward": 0.5948129594326019,
      "rewards-max/forensics_tiou_reward": 0.9667876958847046,
      "rewards/forensics_sinkhorn_reward": 0.7726403772830963,
      "rewards/forensics_strict_edge_reward": 0.07756075076758862,
      "rewards/forensics_tiou_reward": 0.5954851508140564,
      "step": 949
    },
    {
      "completion_length": 21.53125,
      "epoch": 3.97489539748954,
      "grad_norm": 8.013472557067871,
      "kl": 0.431640625,
      "learning_rate": 7.322175732217573e-09,
      "loss": 0.1207,
      "reward": 0.7442228198051453,
      "reward_std": 0.05256544705480337,
      "reward_std_all": 0.055699657648801804,
      "rewards-max/forensics_sinkhorn_reward": 0.9719943404197693,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9585098028182983,
      "rewards/forensics_sinkhorn_reward": 0.6963924169540405,
      "rewards/forensics_strict_edge_reward": 0.06442537158727646,
      "rewards/forensics_tiou_reward": 0.48100195825099945,
      "step": 950
    },
    {
      "completion_length": 19.578125,
      "epoch": 3.97907949790795,
      "grad_norm": 7.1639485359191895,
      "kl": 0.3037109375,
      "learning_rate": 6.276150627615063e-09,
      "loss": 0.0183,
      "reward": 0.7882141768932343,
      "reward_std": 0.031040939036756754,
      "reward_std_all": 0.028668958693742752,
      "rewards-max/forensics_sinkhorn_reward": 0.9781162440776825,
      "rewards-max/forensics_strict_edge_reward": 0.5238595008850098,
      "rewards-max/forensics_tiou_reward": 0.9706136584281921,
      "rewards/forensics_sinkhorn_reward": 0.789732962846756,
      "rewards/forensics_strict_edge_reward": 0.054853059351444244,
      "rewards/forensics_tiou_reward": 0.621190220117569,
      "step": 951
    },
    {
      "completion_length": 20.484375,
      "epoch": 3.98326359832636,
      "grad_norm": 7.346302509307861,
      "kl": 0.4365234375,
      "learning_rate": 5.2301255230125516e-09,
      "loss": -0.0233,
      "reward": 0.7810947299003601,
      "reward_std": 0.04250320326536894,
      "reward_std_all": 0.038792697712779045,
      "rewards-max/forensics_sinkhorn_reward": 0.985634446144104,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9905195534229279,
      "rewards/forensics_sinkhorn_reward": 0.7556941509246826,
      "rewards/forensics_strict_edge_reward": 0.08673610910773277,
      "rewards/forensics_tiou_reward": 0.5784274935722351,
      "step": 952
    },
    {
      "completion_length": 20.25,
      "epoch": 3.98744769874477,
      "grad_norm": 8.31574535369873,
      "kl": 0.5966796875,
      "learning_rate": 4.184100418410041e-09,
      "loss": 0.0458,
      "reward": 0.7363417148590088,
      "reward_std": 0.057556452229619026,
      "reward_std_all": 0.06763745099306107,
      "rewards-max/forensics_sinkhorn_reward": 0.9694985449314117,
      "rewards-max/forensics_strict_edge_reward": 0.5034982860088348,
      "rewards-max/forensics_tiou_reward": 0.9462548792362213,
      "rewards/forensics_sinkhorn_reward": 0.6909812092781067,
      "rewards/forensics_strict_edge_reward": 0.04877879470586777,
      "rewards/forensics_tiou_reward": 0.5070501863956451,
      "step": 953
    },
    {
      "completion_length": 22.34375,
      "epoch": 3.99163179916318,
      "grad_norm": 7.595902919769287,
      "kl": 0.5224609375,
      "learning_rate": 3.1380753138075314e-09,
      "loss": 0.0395,
      "reward": 0.784244954586029,
      "reward_std": 0.035500879399478436,
      "reward_std_all": 0.03277947846800089,
      "rewards-max/forensics_sinkhorn_reward": 0.9844273924827576,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9880495667457581,
      "rewards/forensics_sinkhorn_reward": 0.7481694221496582,
      "rewards/forensics_strict_edge_reward": 0.08172859624028206,
      "rewards/forensics_tiou_reward": 0.6560864746570587,
      "step": 954
    },
    {
      "completion_length": 21.421875,
      "epoch": 3.99581589958159,
      "grad_norm": 6.689349174499512,
      "kl": 0.48828125,
      "learning_rate": 2.0920502092050206e-09,
      "loss": -0.0095,
      "reward": 0.7788329720497131,
      "reward_std": 0.04784225672483444,
      "reward_std_all": 0.043351469561457634,
      "rewards-max/forensics_sinkhorn_reward": 0.982201099395752,
      "rewards-max/forensics_strict_edge_reward": 0.6360974311828613,
      "rewards-max/forensics_tiou_reward": 0.9747461080551147,
      "rewards/forensics_sinkhorn_reward": 0.8245843648910522,
      "rewards/forensics_strict_edge_reward": 0.060034507885575294,
      "rewards/forensics_tiou_reward": 0.6710928380489349,
      "step": 955
    },
    {
      "completion_length": 20.85714340209961,
      "epoch": 4.0,
      "grad_norm": 8.828447341918945,
      "kl": 0.68359375,
      "learning_rate": 1.0460251046025103e-09,
      "loss": -0.0055,
      "reward": 0.8037140667438507,
      "reward_std": 0.03589537367224693,
      "reward_std_all": 0.01819137390702963,
      "rewards-max/forensics_sinkhorn_reward": 0.8985647559165955,
      "rewards-max/forensics_strict_edge_reward": 0.04078339133411646,
      "rewards-max/forensics_tiou_reward": 0.7634011805057526,
      "rewards/forensics_sinkhorn_reward": 0.7571329474449158,
      "rewards/forensics_strict_edge_reward": 0.006552925333380699,
      "rewards/forensics_tiou_reward": 0.5028095990419388,
      "step": 956
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 956,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 50,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}