{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.9906790945406128,
  "eval_steps": 500,
  "global_step": 1500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 23.515625,
      "epoch": 0.002663115845539281,
      "grad_norm": 7.088331699371338,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": -0.0,
      "reward": 0.19094884395599365,
      "reward_std": 0.15431217849254608,
      "reward_std_all": 0.1522206813097,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9689922332763672,
      "rewards/forensics_tempsamp_iou_reward": 0.19094884395599365,
      "step": 1
    },
    {
      "completion_length": 18.84375,
      "epoch": 0.005326231691078562,
      "grad_norm": 5.943724155426025,
      "kl": 0.0009403228759765625,
      "learning_rate": 9.993333333333333e-07,
      "loss": 0.0,
      "reward": 0.22520336508750916,
      "reward_std": 0.13253309577703476,
      "reward_std_all": 0.14436061680316925,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9830808043479919,
      "rewards/forensics_tempsamp_iou_reward": 0.22520336508750916,
      "step": 2
    },
    {
      "completion_length": 21.59375,
      "epoch": 0.007989347536617843,
      "grad_norm": 7.673116207122803,
      "kl": 0.001697540283203125,
      "learning_rate": 9.986666666666667e-07,
      "loss": 0.0001,
      "reward": 0.1796959638595581,
      "reward_std": 0.1252819448709488,
      "reward_std_all": 0.13806790858507156,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6118069738149643,
      "rewards/forensics_tempsamp_iou_reward": 0.1796959638595581,
      "step": 3
    },
    {
      "completion_length": 16.921875,
      "epoch": 0.010652463382157125,
      "grad_norm": 5.027831554412842,
      "kl": 0.00433349609375,
      "learning_rate": 9.98e-07,
      "loss": 0.0002,
      "reward": 0.0615384615957737,
      "reward_std": 0.0369934132322669,
      "reward_std_all": 0.03971591405570507,
      "rewards-max/forensics_tempsamp_iou_reward": 0.27978067100048065,
      "rewards/forensics_tempsamp_iou_reward": 0.0615384615957737,
      "step": 4
    },
    {
      "completion_length": 18.9375,
      "epoch": 0.013315579227696404,
      "grad_norm": 6.808509349822998,
      "kl": 0.0093536376953125,
      "learning_rate": 9.973333333333332e-07,
      "loss": 0.0004,
      "reward": 0.16031770408153534,
      "reward_std": 0.11112917214632034,
      "reward_std_all": 0.1323230303823948,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9797854721546173,
      "rewards/forensics_tempsamp_iou_reward": 0.16031770408153534,
      "step": 5
    },
    {
      "completion_length": 15.5,
      "epoch": 0.015978695073235686,
      "grad_norm": 5.9808669090271,
      "kl": 0.02545166015625,
      "learning_rate": 9.966666666666667e-07,
      "loss": 0.001,
      "reward": 0.18893737345933914,
      "reward_std": 0.07233212608844042,
      "reward_std_all": 0.0909622311592102,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9192624986171722,
      "rewards/forensics_tempsamp_iou_reward": 0.18893737345933914,
      "step": 6
    },
    {
      "completion_length": 22.53125,
      "epoch": 0.018641810918774968,
      "grad_norm": 6.486148834228516,
      "kl": 0.02630615234375,
      "learning_rate": 9.959999999999999e-07,
      "loss": 0.0011,
      "reward": 0.23937319964170456,
      "reward_std": 0.08052830025553703,
      "reward_std_all": 0.08499033004045486,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9853268265724182,
      "rewards/forensics_tempsamp_iou_reward": 0.23937319964170456,
      "step": 7
    },
    {
      "completion_length": 13.1875,
      "epoch": 0.02130492676431425,
      "grad_norm": 6.053421497344971,
      "kl": 0.051513671875,
      "learning_rate": 9.953333333333332e-07,
      "loss": 0.0021,
      "reward": 0.1897440254688263,
      "reward_std": 0.11140463128685951,
      "reward_std_all": 0.09993976727128029,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9912352561950684,
      "rewards/forensics_tempsamp_iou_reward": 0.1897440254688263,
      "step": 8
    },
    {
      "completion_length": 13.40625,
      "epoch": 0.023968042609853527,
      "grad_norm": 5.214946746826172,
      "kl": 0.035400390625,
      "learning_rate": 9.946666666666666e-07,
      "loss": 0.0014,
      "reward": 0.2884043902158737,
      "reward_std": 0.14428375661373138,
      "reward_std_all": 0.15073469281196594,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9884532392024994,
      "rewards/forensics_tempsamp_iou_reward": 0.2884043902158737,
      "step": 9
    },
    {
      "completion_length": 15.1875,
      "epoch": 0.02663115845539281,
      "grad_norm": 5.9041361808776855,
      "kl": 0.06298828125,
      "learning_rate": 9.94e-07,
      "loss": 0.0025,
      "reward": 0.31761062890291214,
      "reward_std": 0.11726230010390282,
      "reward_std_all": 0.11845867708325386,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9608267843723297,
      "rewards/forensics_tempsamp_iou_reward": 0.31761062890291214,
      "step": 10
    },
    {
      "completion_length": 13.796875,
      "epoch": 0.02929427430093209,
      "grad_norm": 9.423796653747559,
      "kl": 0.2509765625,
      "learning_rate": 9.933333333333333e-07,
      "loss": 0.01,
      "reward": 0.19805604219436646,
      "reward_std": 0.07393567264080048,
      "reward_std_all": 0.07296672370284796,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9858593940734863,
      "rewards/forensics_tempsamp_iou_reward": 0.19805604219436646,
      "step": 11
    },
    {
      "completion_length": 13.171875,
      "epoch": 0.03195739014647137,
      "grad_norm": 5.711302280426025,
      "kl": 0.104248046875,
      "learning_rate": 9.926666666666666e-07,
      "loss": 0.0042,
      "reward": 0.2633499801158905,
      "reward_std": 0.10742969252169132,
      "reward_std_all": 0.09888152778148651,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9465329349040985,
      "rewards/forensics_tempsamp_iou_reward": 0.2633499801158905,
      "step": 12
    },
    {
      "completion_length": 13.0625,
      "epoch": 0.03462050599201065,
      "grad_norm": 8.342245101928711,
      "kl": 0.10986328125,
      "learning_rate": 9.92e-07,
      "loss": 0.0044,
      "reward": 0.25571566075086594,
      "reward_std": 0.04588590748608112,
      "reward_std_all": 0.05239627696573734,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7442927956581116,
      "rewards/forensics_tempsamp_iou_reward": 0.25571566075086594,
      "step": 13
    },
    {
      "completion_length": 19.671875,
      "epoch": 0.037283621837549935,
      "grad_norm": 5.663524627685547,
      "kl": 0.0517578125,
      "learning_rate": 9.913333333333333e-07,
      "loss": 0.0021,
      "reward": 0.25446950644254684,
      "reward_std": 0.037229339592158794,
      "reward_std_all": 0.04970581363886595,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9974352717399597,
      "rewards/forensics_tempsamp_iou_reward": 0.25446950644254684,
      "step": 14
    },
    {
      "completion_length": 13.296875,
      "epoch": 0.03994673768308921,
      "grad_norm": 6.024118423461914,
      "kl": 0.091064453125,
      "learning_rate": 9.906666666666667e-07,
      "loss": 0.0036,
      "reward": 0.21298304200172424,
      "reward_std": 0.05671155825257301,
      "reward_std_all": 0.06074565462768078,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9146897792816162,
      "rewards/forensics_tempsamp_iou_reward": 0.21298304200172424,
      "step": 15
    },
    {
      "completion_length": 16.296875,
      "epoch": 0.0426098535286285,
      "grad_norm": 6.585291862487793,
      "kl": 0.16748046875,
      "learning_rate": 9.9e-07,
      "loss": 0.0067,
      "reward": 0.2267034649848938,
      "reward_std": 0.05738389492034912,
      "reward_std_all": 0.0533223282545805,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9497798085212708,
      "rewards/forensics_tempsamp_iou_reward": 0.2267034649848938,
      "step": 16
    },
    {
      "completion_length": 13.171875,
      "epoch": 0.045272969374167776,
      "grad_norm": 6.749837875366211,
      "kl": 0.18408203125,
      "learning_rate": 9.893333333333332e-07,
      "loss": 0.0074,
      "reward": 0.3534667640924454,
      "reward_std": 0.040201522409915924,
      "reward_std_all": 0.042737290263175964,
      "rewards-max/forensics_tempsamp_iou_reward": 0.92357137799263,
      "rewards/forensics_tempsamp_iou_reward": 0.3534667640924454,
      "step": 17
    },
    {
      "completion_length": 14.015625,
      "epoch": 0.047936085219707054,
      "grad_norm": 6.347641944885254,
      "kl": 0.087158203125,
      "learning_rate": 9.886666666666665e-07,
      "loss": 0.0035,
      "reward": 0.3607080802321434,
      "reward_std": 0.0415964238345623,
      "reward_std_all": 0.04954117722809315,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9984221458435059,
      "rewards/forensics_tempsamp_iou_reward": 0.3607080802321434,
      "step": 18
    },
    {
      "completion_length": 14.921875,
      "epoch": 0.05059920106524634,
      "grad_norm": 7.698709964752197,
      "kl": 0.224365234375,
      "learning_rate": 9.88e-07,
      "loss": 0.009,
      "reward": 0.2181483432650566,
      "reward_std": 0.07566940784454346,
      "reward_std_all": 0.06806332617998123,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9065393507480621,
      "rewards/forensics_tempsamp_iou_reward": 0.2181483432650566,
      "step": 19
    },
    {
      "completion_length": 17.078125,
      "epoch": 0.05326231691078562,
      "grad_norm": 10.47898006439209,
      "kl": 0.333740234375,
      "learning_rate": 9.873333333333333e-07,
      "loss": 0.0133,
      "reward": 0.289050817489624,
      "reward_std": 0.05056094005703926,
      "reward_std_all": 0.053815918043255806,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9764252603054047,
      "rewards/forensics_tempsamp_iou_reward": 0.289050817489624,
      "step": 20
    },
    {
      "completion_length": 17.78125,
      "epoch": 0.0559254327563249,
      "grad_norm": 7.324185848236084,
      "kl": 0.08544921875,
      "learning_rate": 9.866666666666666e-07,
      "loss": 0.0034,
      "reward": 0.24502327665686607,
      "reward_std": 0.03916102834045887,
      "reward_std_all": 0.04613499157130718,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6711244881153107,
      "rewards/forensics_tempsamp_iou_reward": 0.24502327665686607,
      "step": 21
    },
    {
      "completion_length": 13.109375,
      "epoch": 0.05858854860186418,
      "grad_norm": 5.728130340576172,
      "kl": 0.23828125,
      "learning_rate": 9.86e-07,
      "loss": 0.0095,
      "reward": 0.2435423545539379,
      "reward_std": 0.07907659001648426,
      "reward_std_all": 0.07935903035104275,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6000000014901161,
      "rewards/forensics_tempsamp_iou_reward": 0.2435423545539379,
      "step": 22
    },
    {
      "completion_length": 13.96875,
      "epoch": 0.06125166444740346,
      "grad_norm": 9.543063163757324,
      "kl": 0.310546875,
      "learning_rate": 9.853333333333333e-07,
      "loss": 0.0124,
      "reward": 0.26715975999832153,
      "reward_std": 0.09463578462600708,
      "reward_std_all": 0.11438479274511337,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6908411979675293,
      "rewards/forensics_tempsamp_iou_reward": 0.26715975999832153,
      "step": 23
    },
    {
      "completion_length": 14.5625,
      "epoch": 0.06391478029294274,
      "grad_norm": 8.33825397491455,
      "kl": 0.38525390625,
      "learning_rate": 9.846666666666667e-07,
      "loss": 0.0154,
      "reward": 0.162636436522007,
      "reward_std": 0.08285702764987946,
      "reward_std_all": 0.07850391417741776,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9061160385608673,
      "rewards/forensics_tempsamp_iou_reward": 0.162636436522007,
      "step": 24
    },
    {
      "completion_length": 15.375,
      "epoch": 0.06657789613848203,
      "grad_norm": 10.362305641174316,
      "kl": 0.36328125,
      "learning_rate": 9.84e-07,
      "loss": 0.0146,
      "reward": 0.15724864974617958,
      "reward_std": 0.08334795664995909,
      "reward_std_all": 0.08031228184700012,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5419028475880623,
      "rewards/forensics_tempsamp_iou_reward": 0.15724864974617958,
      "step": 25
    },
    {
      "completion_length": 15.1875,
      "epoch": 0.0692410119840213,
      "grad_norm": 9.83596420288086,
      "kl": 0.33251953125,
      "learning_rate": 9.833333333333332e-07,
      "loss": 0.0133,
      "reward": 0.16061869263648987,
      "reward_std": 0.07892841845750809,
      "reward_std_all": 0.08119526319205761,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6346790939569473,
      "rewards/forensics_tempsamp_iou_reward": 0.16061869263648987,
      "step": 26
    },
    {
      "completion_length": 13.75,
      "epoch": 0.07190412782956059,
      "grad_norm": 5.389522075653076,
      "kl": 0.2119140625,
      "learning_rate": 9.826666666666667e-07,
      "loss": 0.0085,
      "reward": 0.1582697033882141,
      "reward_std": 0.08435793872922659,
      "reward_std_all": 0.09059223905205727,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5521611049771309,
      "rewards/forensics_tempsamp_iou_reward": 0.1582697033882141,
      "step": 27
    },
    {
      "completion_length": 13.734375,
      "epoch": 0.07456724367509987,
      "grad_norm": 14.765300750732422,
      "kl": 1.18359375,
      "learning_rate": 9.819999999999999e-07,
      "loss": 0.0474,
      "reward": 0.16238707304000854,
      "reward_std": 0.08085202053189278,
      "reward_std_all": 0.08072541281580925,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7467913329601288,
      "rewards/forensics_tempsamp_iou_reward": 0.16238707304000854,
      "step": 28
    },
    {
      "completion_length": 14.4375,
      "epoch": 0.07723035952063914,
      "grad_norm": 7.349244594573975,
      "kl": 0.32080078125,
      "learning_rate": 9.813333333333332e-07,
      "loss": 0.0128,
      "reward": 0.12410581484436989,
      "reward_std": 0.07647507265210152,
      "reward_std_all": 0.0769518343731761,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6105842739343643,
      "rewards/forensics_tempsamp_iou_reward": 0.12410581484436989,
      "step": 29
    },
    {
      "completion_length": 14.40625,
      "epoch": 0.07989347536617843,
      "grad_norm": 7.109793663024902,
      "kl": 0.30712890625,
      "learning_rate": 9.806666666666666e-07,
      "loss": 0.0123,
      "reward": 0.23740535229444504,
      "reward_std": 0.1082761138677597,
      "reward_std_all": 0.10711277648806572,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9832451641559601,
      "rewards/forensics_tempsamp_iou_reward": 0.23740535229444504,
      "step": 30
    },
    {
      "completion_length": 16.0625,
      "epoch": 0.08255659121171771,
      "grad_norm": 8.802209854125977,
      "kl": 0.21533203125,
      "learning_rate": 9.8e-07,
      "loss": 0.0086,
      "reward": 0.1316942647099495,
      "reward_std": 0.1143699660897255,
      "reward_std_all": 0.11071765795350075,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6714938879013062,
      "rewards/forensics_tempsamp_iou_reward": 0.1316942647099495,
      "step": 31
    },
    {
      "completion_length": 13.1875,
      "epoch": 0.085219707057257,
      "grad_norm": 5.975461006164551,
      "kl": 0.455078125,
      "learning_rate": 9.793333333333333e-07,
      "loss": 0.0182,
      "reward": 0.1961311399936676,
      "reward_std": 0.07602158561348915,
      "reward_std_all": 0.08584072440862656,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8074803650379181,
      "rewards/forensics_tempsamp_iou_reward": 0.1961311399936676,
      "step": 32
    },
    {
      "completion_length": 15.90625,
      "epoch": 0.08788282290279627,
      "grad_norm": 6.595985412597656,
      "kl": 0.23681640625,
      "learning_rate": 9.786666666666666e-07,
      "loss": 0.0095,
      "reward": 0.1266207955777645,
      "reward_std": 0.09696868062019348,
      "reward_std_all": 0.08946865797042847,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7918033003807068,
      "rewards/forensics_tempsamp_iou_reward": 0.1266207955777645,
      "step": 33
    },
    {
      "completion_length": 13.515625,
      "epoch": 0.09054593874833555,
      "grad_norm": 5.656022071838379,
      "kl": 0.3203125,
      "learning_rate": 9.78e-07,
      "loss": 0.0128,
      "reward": 0.16830937564373016,
      "reward_std": 0.10564752295613289,
      "reward_std_all": 0.10221978276968002,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7950195372104645,
      "rewards/forensics_tempsamp_iou_reward": 0.16830937564373016,
      "step": 34
    },
    {
      "completion_length": 15.4375,
      "epoch": 0.09320905459387484,
      "grad_norm": 7.868865489959717,
      "kl": 0.220458984375,
      "learning_rate": 9.773333333333333e-07,
      "loss": 0.0088,
      "reward": 0.2011426016688347,
      "reward_std": 0.08504900336265564,
      "reward_std_all": 0.09270066395401955,
      "rewards-max/forensics_tempsamp_iou_reward": 0.912519097328186,
      "rewards/forensics_tempsamp_iou_reward": 0.2011426016688347,
      "step": 35
    },
    {
      "completion_length": 14.375,
      "epoch": 0.09587217043941411,
      "grad_norm": 9.334800720214844,
      "kl": 0.296142578125,
      "learning_rate": 9.766666666666667e-07,
      "loss": 0.0118,
      "reward": 0.14110679551959038,
      "reward_std": 0.043478209525346756,
      "reward_std_all": 0.06225917860865593,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6325197815895081,
      "rewards/forensics_tempsamp_iou_reward": 0.14110679551959038,
      "step": 36
    },
    {
      "completion_length": 16.96875,
      "epoch": 0.0985352862849534,
      "grad_norm": 5.161166191101074,
      "kl": 0.326171875,
      "learning_rate": 9.759999999999998e-07,
      "loss": 0.013,
      "reward": 0.14843787252902985,
      "reward_std": 0.1169708352535963,
      "reward_std_all": 0.11002442054450512,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6106946766376495,
      "rewards/forensics_tempsamp_iou_reward": 0.14843787252902985,
      "step": 37
    },
    {
      "completion_length": 14.640625,
      "epoch": 0.10119840213049268,
      "grad_norm": 7.173987865447998,
      "kl": 0.41015625,
      "learning_rate": 9.753333333333334e-07,
      "loss": 0.0165,
      "reward": 0.2431386336684227,
      "reward_std": 0.1030188538134098,
      "reward_std_all": 0.11713224276900291,
      "rewards-max/forensics_tempsamp_iou_reward": 0.753044992685318,
      "rewards/forensics_tempsamp_iou_reward": 0.2431386336684227,
      "step": 38
    },
    {
      "completion_length": 16.203125,
      "epoch": 0.10386151797603196,
      "grad_norm": 6.448601722717285,
      "kl": 0.2890625,
      "learning_rate": 9.746666666666666e-07,
      "loss": 0.0115,
      "reward": 0.25878433883190155,
      "reward_std": 0.0983075238764286,
      "reward_std_all": 0.10515286773443222,
      "rewards-max/forensics_tempsamp_iou_reward": 0.874861866235733,
      "rewards/forensics_tempsamp_iou_reward": 0.25878433883190155,
      "step": 39
    },
    {
      "completion_length": 15.125,
      "epoch": 0.10652463382157124,
      "grad_norm": 7.113555431365967,
      "kl": 0.341796875,
      "learning_rate": 9.74e-07,
      "loss": 0.0137,
      "reward": 0.22234388440847397,
      "reward_std": 0.11804297938942909,
      "reward_std_all": 0.11076504737138748,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9855466783046722,
      "rewards/forensics_tempsamp_iou_reward": 0.22234388440847397,
      "step": 40
    },
    {
      "completion_length": 15.3125,
      "epoch": 0.10918774966711052,
      "grad_norm": 7.235232353210449,
      "kl": 0.1845703125,
      "learning_rate": 9.733333333333333e-07,
      "loss": 0.0074,
      "reward": 0.23318371921777725,
      "reward_std": 0.09195732697844505,
      "reward_std_all": 0.08873707428574562,
      "rewards-max/forensics_tempsamp_iou_reward": 0.861786276102066,
      "rewards/forensics_tempsamp_iou_reward": 0.23318371921777725,
      "step": 41
    },
    {
      "completion_length": 13.875,
      "epoch": 0.1118508655126498,
      "grad_norm": 4.686288356781006,
      "kl": 0.1494140625,
      "learning_rate": 9.726666666666666e-07,
      "loss": 0.006,
      "reward": 0.23480743169784546,
      "reward_std": 0.06616377457976341,
      "reward_std_all": 0.08484265953302383,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6624104082584381,
      "rewards/forensics_tempsamp_iou_reward": 0.23480743169784546,
      "step": 42
    },
    {
      "completion_length": 15.25,
      "epoch": 0.11451398135818908,
      "grad_norm": 7.2848801612854,
      "kl": 0.15576171875,
      "learning_rate": 9.72e-07,
      "loss": 0.0062,
      "reward": 0.20370050892233849,
      "reward_std": 0.06845470517873764,
      "reward_std_all": 0.09679914638400078,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6335540860891342,
      "rewards/forensics_tempsamp_iou_reward": 0.20370050892233849,
      "step": 43
    },
    {
      "completion_length": 19.046875,
      "epoch": 0.11717709720372836,
      "grad_norm": 4.32876443862915,
      "kl": 0.14794921875,
      "learning_rate": 9.713333333333333e-07,
      "loss": 0.0059,
      "reward": 0.2197330817580223,
      "reward_std": 0.07214505039155483,
      "reward_std_all": 0.0729035846889019,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8047433495521545,
      "rewards/forensics_tempsamp_iou_reward": 0.2197330817580223,
      "step": 44
    },
    {
      "completion_length": 23.609375,
      "epoch": 0.11984021304926765,
      "grad_norm": 5.477384090423584,
      "kl": 0.2041015625,
      "learning_rate": 9.706666666666667e-07,
      "loss": 0.0082,
      "reward": 0.20918290317058563,
      "reward_std": 0.08446913212537766,
      "reward_std_all": 0.09190873801708221,
      "rewards-max/forensics_tempsamp_iou_reward": 0.894727885723114,
      "rewards/forensics_tempsamp_iou_reward": 0.20918290317058563,
      "step": 45
    },
    {
      "completion_length": 17.265625,
      "epoch": 0.12250332889480692,
      "grad_norm": 6.022874355316162,
      "kl": 0.1201171875,
      "learning_rate": 9.7e-07,
      "loss": 0.0048,
      "reward": 0.18210644274950027,
      "reward_std": 0.06839561648666859,
      "reward_std_all": 0.0798393152654171,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6941632181406021,
      "rewards/forensics_tempsamp_iou_reward": 0.18210644274950027,
      "step": 46
    },
    {
      "completion_length": 16.140625,
      "epoch": 0.12516644474034622,
      "grad_norm": 5.616611480712891,
      "kl": 0.14306640625,
      "learning_rate": 9.693333333333334e-07,
      "loss": 0.0057,
      "reward": 0.3742285519838333,
      "reward_std": 0.10073758661746979,
      "reward_std_all": 0.12026233598589897,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8855536878108978,
      "rewards/forensics_tempsamp_iou_reward": 0.3742285519838333,
      "step": 47
    },
    {
      "completion_length": 15.578125,
      "epoch": 0.1278295605858855,
      "grad_norm": 5.380115032196045,
      "kl": 0.13330078125,
      "learning_rate": 9.686666666666667e-07,
      "loss": 0.0053,
      "reward": 0.2445119321346283,
      "reward_std": 0.08651064708828926,
      "reward_std_all": 0.08749470487236977,
      "rewards-max/forensics_tempsamp_iou_reward": 0.945084273815155,
      "rewards/forensics_tempsamp_iou_reward": 0.2445119321346283,
      "step": 48
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.13049267643142476,
      "grad_norm": 5.036190509796143,
      "kl": 0.150634765625,
      "learning_rate": 9.679999999999999e-07,
      "loss": 0.006,
      "reward": 0.2521722614765167,
      "reward_std": 0.1090473085641861,
      "reward_std_all": 0.10142131522297859,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9589980840682983,
      "rewards/forensics_tempsamp_iou_reward": 0.2521722614765167,
      "step": 49
    },
    {
      "completion_length": 14.546875,
      "epoch": 0.13315579227696406,
      "grad_norm": 5.549219131469727,
      "kl": 0.22509765625,
      "learning_rate": 9.673333333333332e-07,
      "loss": 0.009,
      "reward": 0.16152755916118622,
      "reward_std": 0.05116772651672363,
      "reward_std_all": 0.06199003383517265,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6644930690526962,
      "rewards/forensics_tempsamp_iou_reward": 0.16152755916118622,
      "step": 50
    },
    {
      "completion_length": 22.921875,
      "epoch": 0.13581890812250333,
      "grad_norm": 6.753683090209961,
      "kl": 0.2158203125,
      "learning_rate": 9.666666666666666e-07,
      "loss": 0.0086,
      "reward": 0.30770185589790344,
      "reward_std": 0.13436809927225113,
      "reward_std_all": 0.13740132004022598,
      "rewards-max/forensics_tempsamp_iou_reward": 0.975558340549469,
      "rewards/forensics_tempsamp_iou_reward": 0.30770185589790344,
      "step": 51
    },
    {
      "completion_length": 14.015625,
      "epoch": 0.1384820239680426,
      "grad_norm": 5.931222438812256,
      "kl": 0.18017578125,
      "learning_rate": 9.66e-07,
      "loss": 0.0072,
      "reward": 0.28954213857650757,
      "reward_std": 0.10206332057714462,
      "reward_std_all": 0.1136152409017086,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9559803903102875,
      "rewards/forensics_tempsamp_iou_reward": 0.28954213857650757,
      "step": 52
    },
    {
      "completion_length": 13.375,
      "epoch": 0.1411451398135819,
      "grad_norm": 6.138474464416504,
      "kl": 0.173828125,
      "learning_rate": 9.653333333333333e-07,
      "loss": 0.007,
      "reward": 0.23571226000785828,
      "reward_std": 0.049601515755057335,
      "reward_std_all": 0.05089063011109829,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9302355349063873,
      "rewards/forensics_tempsamp_iou_reward": 0.23571226000785828,
      "step": 53
    },
    {
      "completion_length": 16.171875,
      "epoch": 0.14380825565912117,
      "grad_norm": 5.8631696701049805,
      "kl": 0.203125,
      "learning_rate": 9.646666666666666e-07,
      "loss": 0.0082,
      "reward": 0.1556570641696453,
      "reward_std": 0.06749366596341133,
      "reward_std_all": 0.07464632019400597,
      "rewards-max/forensics_tempsamp_iou_reward": 0.79182168841362,
      "rewards/forensics_tempsamp_iou_reward": 0.1556570641696453,
      "step": 54
    },
    {
      "completion_length": 13.703125,
      "epoch": 0.14647137150466044,
      "grad_norm": 20.07195472717285,
      "kl": 0.82373046875,
      "learning_rate": 9.64e-07,
      "loss": 0.0329,
      "reward": 0.25826241075992584,
      "reward_std": 0.058312175795435905,
      "reward_std_all": 0.06388839334249496,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9656310975551605,
      "rewards/forensics_tempsamp_iou_reward": 0.25826241075992584,
      "step": 55
    },
    {
      "completion_length": 13.984375,
      "epoch": 0.14913448735019974,
      "grad_norm": 7.750335693359375,
      "kl": 0.2470703125,
      "learning_rate": 9.633333333333334e-07,
      "loss": 0.0099,
      "reward": 0.32210955023765564,
      "reward_std": 0.09972422942519188,
      "reward_std_all": 0.08842066302895546,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9620413780212402,
      "rewards/forensics_tempsamp_iou_reward": 0.32210955023765564,
      "step": 56
    },
    {
      "completion_length": 13.453125,
      "epoch": 0.151797603195739,
      "grad_norm": 5.5533952713012695,
      "kl": 0.129150390625,
      "learning_rate": 9.626666666666667e-07,
      "loss": 0.0052,
      "reward": 0.2867632806301117,
      "reward_std": 0.04263828322291374,
      "reward_std_all": 0.052736859768629074,
      "rewards-max/forensics_tempsamp_iou_reward": 0.874849259853363,
      "rewards/forensics_tempsamp_iou_reward": 0.2867632806301117,
      "step": 57
    },
    {
      "completion_length": 14.390625,
      "epoch": 0.15446071904127828,
      "grad_norm": 7.913843631744385,
      "kl": 0.23486328125,
      "learning_rate": 9.619999999999999e-07,
      "loss": 0.0094,
      "reward": 0.15610992163419724,
      "reward_std": 0.033259784802794456,
      "reward_std_all": 0.04845619574189186,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5499204248189926,
      "rewards/forensics_tempsamp_iou_reward": 0.15610992163419724,
      "step": 58
    },
    {
      "completion_length": 13.109375,
      "epoch": 0.15712383488681758,
      "grad_norm": 7.123469352722168,
      "kl": 0.12890625,
      "learning_rate": 9.613333333333334e-07,
      "loss": 0.0052,
      "reward": 0.1770532950758934,
      "reward_std": 0.04598281718790531,
      "reward_std_all": 0.05166001617908478,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5536144077777863,
      "rewards/forensics_tempsamp_iou_reward": 0.1770532950758934,
      "step": 59
    },
    {
      "completion_length": 13.640625,
      "epoch": 0.15978695073235685,
      "grad_norm": 5.598912715911865,
      "kl": 0.19970703125,
      "learning_rate": 9.606666666666666e-07,
      "loss": 0.008,
      "reward": 0.31978069245815277,
      "reward_std": 0.08422122150659561,
      "reward_std_all": 0.08442802354693413,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9359161853790283,
      "rewards/forensics_tempsamp_iou_reward": 0.31978069245815277,
      "step": 60
    },
    {
      "completion_length": 13.046875,
      "epoch": 0.16245006657789615,
      "grad_norm": 7.625974655151367,
      "kl": 0.2880859375,
      "learning_rate": 9.6e-07,
      "loss": 0.0115,
      "reward": 0.23556632548570633,
      "reward_std": 0.031898293644189835,
      "reward_std_all": 0.06275777518749237,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6622329950332642,
      "rewards/forensics_tempsamp_iou_reward": 0.23556632548570633,
      "step": 61
    },
    {
      "completion_length": 13.125,
      "epoch": 0.16511318242343542,
      "grad_norm": 5.834689140319824,
      "kl": 0.28466796875,
      "learning_rate": 9.593333333333333e-07,
      "loss": 0.0114,
      "reward": 0.20121043920516968,
      "reward_std": 0.10971649922430515,
      "reward_std_all": 0.10070986580103636,
      "rewards-max/forensics_tempsamp_iou_reward": 0.594324067234993,
      "rewards/forensics_tempsamp_iou_reward": 0.20121043920516968,
      "step": 62
    },
    {
      "completion_length": 13.171875,
      "epoch": 0.1677762982689747,
      "grad_norm": 7.698456764221191,
      "kl": 0.22412109375,
      "learning_rate": 9.586666666666666e-07,
      "loss": 0.009,
      "reward": 0.2215561717748642,
      "reward_std": 0.07189656049013138,
      "reward_std_all": 0.07808399572968483,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9284722208976746,
      "rewards/forensics_tempsamp_iou_reward": 0.2215561717748642,
      "step": 63
    },
    {
      "completion_length": 15.453125,
      "epoch": 0.170439414114514,
      "grad_norm": 7.534857273101807,
      "kl": 0.23828125,
      "learning_rate": 9.58e-07,
      "loss": 0.0095,
      "reward": 0.20885630697011948,
      "reward_std": 0.057121168822050095,
      "reward_std_all": 0.056585367769002914,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9038722515106201,
      "rewards/forensics_tempsamp_iou_reward": 0.20885630697011948,
      "step": 64
    },
    {
      "completion_length": 19.078125,
      "epoch": 0.17310252996005326,
      "grad_norm": 5.132640361785889,
      "kl": 0.2041015625,
      "learning_rate": 9.573333333333333e-07,
      "loss": 0.0082,
      "reward": 0.33736786246299744,
      "reward_std": 0.10473544895648956,
      "reward_std_all": 0.10712818801403046,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9756328165531158,
      "rewards/forensics_tempsamp_iou_reward": 0.33736786246299744,
      "step": 65
    },
    {
      "completion_length": 16.515625,
      "epoch": 0.17576564580559254,
      "grad_norm": 4.801682949066162,
      "kl": 0.2138671875,
      "learning_rate": 9.566666666666667e-07,
      "loss": 0.0085,
      "reward": 0.3039391040802002,
      "reward_std": 0.0786285400390625,
      "reward_std_all": 0.0819549709558487,
      "rewards-max/forensics_tempsamp_iou_reward": 0.909617006778717,
      "rewards/forensics_tempsamp_iou_reward": 0.3039391040802002,
      "step": 66
    },
    {
      "completion_length": 14.234375,
      "epoch": 0.17842876165113183,
      "grad_norm": 5.187739849090576,
      "kl": 0.28759765625,
      "learning_rate": 9.559999999999998e-07,
      "loss": 0.0115,
      "reward": 0.20847509056329727,
      "reward_std": 0.090633574873209,
      "reward_std_all": 0.08738664165139198,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7989651560783386,
      "rewards/forensics_tempsamp_iou_reward": 0.20847509056329727,
      "step": 67
    },
    {
      "completion_length": 14.203125,
      "epoch": 0.1810918774966711,
      "grad_norm": 6.204514026641846,
      "kl": 0.349609375,
      "learning_rate": 9.553333333333334e-07,
      "loss": 0.014,
      "reward": 0.26887719333171844,
      "reward_std": 0.11773523315787315,
      "reward_std_all": 0.11960087344050407,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6711794883012772,
      "rewards/forensics_tempsamp_iou_reward": 0.26887719333171844,
      "step": 68
    },
    {
      "completion_length": 21.9375,
      "epoch": 0.18375499334221038,
      "grad_norm": 5.547693729400635,
      "kl": 0.3984375,
      "learning_rate": 9.546666666666665e-07,
      "loss": 0.0159,
      "reward": 0.2422208935022354,
      "reward_std": 0.1019977405667305,
      "reward_std_all": 0.11814407259225845,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7665832340717316,
      "rewards/forensics_tempsamp_iou_reward": 0.2422208935022354,
      "step": 69
    },
    {
      "completion_length": 13.46875,
      "epoch": 0.18641810918774968,
      "grad_norm": 4.576892852783203,
      "kl": 0.357421875,
      "learning_rate": 9.539999999999999e-07,
      "loss": 0.0143,
      "reward": 0.2521452531218529,
      "reward_std": 0.08648072183132172,
      "reward_std_all": 0.09655945375561714,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7614348530769348,
      "rewards/forensics_tempsamp_iou_reward": 0.2521452531218529,
      "step": 70
    },
    {
      "completion_length": 14.765625,
      "epoch": 0.18908122503328895,
      "grad_norm": 4.764678478240967,
      "kl": 0.193359375,
      "learning_rate": 9.533333333333333e-07,
      "loss": 0.0077,
      "reward": 0.2183549404144287,
      "reward_std": 0.0704986471682787,
      "reward_std_all": 0.09707920253276825,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5816263258457184,
      "rewards/forensics_tempsamp_iou_reward": 0.2183549404144287,
      "step": 71
    },
    {
      "completion_length": 18.921875,
      "epoch": 0.19174434087882822,
      "grad_norm": 5.879024505615234,
      "kl": 0.24365234375,
      "learning_rate": 9.526666666666666e-07,
      "loss": 0.0097,
      "reward": 0.20592822134494781,
      "reward_std": 0.10825841873884201,
      "reward_std_all": 0.12217932567000389,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7376545965671539,
      "rewards/forensics_tempsamp_iou_reward": 0.20592822134494781,
      "step": 72
    },
    {
      "completion_length": 13.734375,
      "epoch": 0.19440745672436752,
      "grad_norm": 5.539093017578125,
      "kl": 0.2802734375,
      "learning_rate": 9.52e-07,
      "loss": 0.0112,
      "reward": 0.2647741585969925,
      "reward_std": 0.12045466899871826,
      "reward_std_all": 0.11787928268313408,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8431521058082581,
      "rewards/forensics_tempsamp_iou_reward": 0.2647741585969925,
      "step": 73
    },
    {
      "completion_length": 13.4375,
      "epoch": 0.1970705725699068,
      "grad_norm": 16.452478408813477,
      "kl": 0.9140625,
      "learning_rate": 9.513333333333333e-07,
      "loss": 0.0367,
      "reward": 0.16160015761852264,
      "reward_std": 0.05897035636007786,
      "reward_std_all": 0.0676724724471569,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4919307380914688,
      "rewards/forensics_tempsamp_iou_reward": 0.16160015761852264,
      "step": 74
    },
    {
      "completion_length": 19.359375,
      "epoch": 0.19973368841544606,
      "grad_norm": 4.726822376251221,
      "kl": 0.4716796875,
      "learning_rate": 9.506666666666667e-07,
      "loss": 0.0189,
      "reward": 0.15296966210007668,
      "reward_std": 0.11857909709215164,
      "reward_std_all": 0.11011850461363792,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6793999373912811,
      "rewards/forensics_tempsamp_iou_reward": 0.15296966210007668,
      "step": 75
    },
    {
      "completion_length": 14.75,
      "epoch": 0.20239680426098536,
      "grad_norm": 9.326299667358398,
      "kl": 0.7431640625,
      "learning_rate": 9.499999999999999e-07,
      "loss": 0.0298,
      "reward": 0.24690590798854828,
      "reward_std": 0.11005421727895737,
      "reward_std_all": 0.10932905226945877,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8891667127609253,
      "rewards/forensics_tempsamp_iou_reward": 0.24690590798854828,
      "step": 76
    },
    {
      "completion_length": 20.15625,
      "epoch": 0.20505992010652463,
      "grad_norm": 5.683047771453857,
      "kl": 0.259765625,
      "learning_rate": 9.493333333333334e-07,
      "loss": 0.0104,
      "reward": 0.21578582376241684,
      "reward_std": 0.11306342855095863,
      "reward_std_all": 0.11319580674171448,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8083170354366302,
      "rewards/forensics_tempsamp_iou_reward": 0.21578582376241684,
      "step": 77
    },
    {
      "completion_length": 19.484375,
      "epoch": 0.20772303595206393,
      "grad_norm": 4.982259750366211,
      "kl": 0.2236328125,
      "learning_rate": 9.486666666666666e-07,
      "loss": 0.0089,
      "reward": 0.144111767411232,
      "reward_std": 0.1003182977437973,
      "reward_std_all": 0.09596959128975868,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9167299568653107,
      "rewards/forensics_tempsamp_iou_reward": 0.144111767411232,
      "step": 78
    },
    {
      "completion_length": 16.4375,
      "epoch": 0.2103861517976032,
      "grad_norm": 5.993364334106445,
      "kl": 0.15234375,
      "learning_rate": 9.479999999999999e-07,
      "loss": 0.0061,
      "reward": 0.1476036086678505,
      "reward_std": 0.06906198337674141,
      "reward_std_all": 0.07578720338642597,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4720841944217682,
      "rewards/forensics_tempsamp_iou_reward": 0.1476036086678505,
      "step": 79
    },
    {
      "completion_length": 19.046875,
      "epoch": 0.21304926764314247,
      "grad_norm": 6.602160930633545,
      "kl": 0.2314453125,
      "learning_rate": 9.473333333333333e-07,
      "loss": 0.0093,
      "reward": 0.16378570348024368,
      "reward_std": 0.08532186970114708,
      "reward_std_all": 0.10351583361625671,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7128966748714447,
      "rewards/forensics_tempsamp_iou_reward": 0.16378570348024368,
      "step": 80
    },
    {
      "completion_length": 23.546875,
      "epoch": 0.21571238348868177,
      "grad_norm": 5.272195816040039,
      "kl": 0.2900390625,
      "learning_rate": 9.466666666666666e-07,
      "loss": 0.0116,
      "reward": 0.2660663574934006,
      "reward_std": 0.1498829871416092,
      "reward_std_all": 0.14580368995666504,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7725517749786377,
      "rewards/forensics_tempsamp_iou_reward": 0.2660663574934006,
      "step": 81
    },
    {
      "completion_length": 18.96875,
      "epoch": 0.21837549933422104,
      "grad_norm": 7.502697467803955,
      "kl": 0.5771484375,
      "learning_rate": 9.459999999999999e-07,
      "loss": 0.0232,
      "reward": 0.22021149098873138,
      "reward_std": 0.14724963903427124,
      "reward_std_all": 0.16357062384486198,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8562606871128082,
      "rewards/forensics_tempsamp_iou_reward": 0.22021149098873138,
      "step": 82
    },
    {
      "completion_length": 19.84375,
      "epoch": 0.2210386151797603,
      "grad_norm": 6.437880516052246,
      "kl": 0.20703125,
      "learning_rate": 9.453333333333333e-07,
      "loss": 0.0083,
      "reward": 0.20941344648599625,
      "reward_std": 0.0800582729279995,
      "reward_std_all": 0.10975702106952667,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5586113929748535,
      "rewards/forensics_tempsamp_iou_reward": 0.20941344648599625,
      "step": 83
    },
    {
      "completion_length": 17.328125,
      "epoch": 0.2237017310252996,
      "grad_norm": 9.93847942352295,
      "kl": 0.5361328125,
      "learning_rate": 9.446666666666666e-07,
      "loss": 0.0214,
      "reward": 0.2569105550646782,
      "reward_std": 0.1396554820239544,
      "reward_std_all": 0.13298838958144188,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9489587843418121,
      "rewards/forensics_tempsamp_iou_reward": 0.2569105550646782,
      "step": 84
    },
    {
      "completion_length": 18.5625,
      "epoch": 0.22636484687083888,
      "grad_norm": 5.382565975189209,
      "kl": 0.18310546875,
      "learning_rate": 9.439999999999999e-07,
      "loss": 0.0073,
      "reward": 0.2685755640268326,
      "reward_std": 0.13743677735328674,
      "reward_std_all": 0.12646939232945442,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9184412360191345,
      "rewards/forensics_tempsamp_iou_reward": 0.2685755640268326,
      "step": 85
    },
    {
      "completion_length": 25.6875,
      "epoch": 0.22902796271637815,
      "grad_norm": 4.74600076675415,
      "kl": 0.18310546875,
      "learning_rate": 9.433333333333333e-07,
      "loss": 0.0073,
      "reward": 0.1980077475309372,
      "reward_std": 0.07419377472251654,
      "reward_std_all": 0.09254834800958633,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8871527910232544,
      "rewards/forensics_tempsamp_iou_reward": 0.1980077475309372,
      "step": 86
    },
    {
      "completion_length": 16.9375,
      "epoch": 0.23169107856191745,
      "grad_norm": 7.167601108551025,
      "kl": 0.22705078125,
      "learning_rate": 9.426666666666666e-07,
      "loss": 0.0091,
      "reward": 0.25954072177410126,
      "reward_std": 0.09730536118149757,
      "reward_std_all": 0.1116539016366005,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9485063254833221,
      "rewards/forensics_tempsamp_iou_reward": 0.25954072177410126,
      "step": 87
    },
    {
      "completion_length": 14.203125,
      "epoch": 0.23435419440745672,
      "grad_norm": 5.628681182861328,
      "kl": 0.267578125,
      "learning_rate": 9.419999999999999e-07,
      "loss": 0.0107,
      "reward": 0.22046614438295364,
      "reward_std": 0.050913210958242416,
      "reward_std_all": 0.054445596411824226,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7200243771076202,
      "rewards/forensics_tempsamp_iou_reward": 0.22046614438295364,
      "step": 88
    },
    {
      "completion_length": 13.265625,
      "epoch": 0.237017310252996,
      "grad_norm": 5.5373430252075195,
      "kl": 0.1630859375,
      "learning_rate": 9.413333333333333e-07,
      "loss": 0.0065,
      "reward": 0.30160142481327057,
      "reward_std": 0.08257251232862473,
      "reward_std_all": 0.07739536836743355,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9335725009441376,
      "rewards/forensics_tempsamp_iou_reward": 0.30160142481327057,
      "step": 89
    },
    {
      "completion_length": 15.15625,
      "epoch": 0.2396804260985353,
      "grad_norm": 5.905013561248779,
      "kl": 0.27734375,
      "learning_rate": 9.406666666666666e-07,
      "loss": 0.0111,
      "reward": 0.35983937978744507,
      "reward_std": 0.10430191457271576,
      "reward_std_all": 0.09947337210178375,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9703260660171509,
      "rewards/forensics_tempsamp_iou_reward": 0.35983937978744507,
      "step": 90
    },
    {
      "completion_length": 14.609375,
      "epoch": 0.24234354194407456,
      "grad_norm": 6.894356727600098,
      "kl": 0.20068359375,
      "learning_rate": 9.399999999999999e-07,
      "loss": 0.008,
      "reward": 0.31670863926410675,
      "reward_std": 0.07159960642457008,
      "reward_std_all": 0.06662116572260857,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9320175647735596,
      "rewards/forensics_tempsamp_iou_reward": 0.31670863926410675,
      "step": 91
    },
    {
      "completion_length": 13.796875,
      "epoch": 0.24500665778961384,
      "grad_norm": 5.362201690673828,
      "kl": 0.2451171875,
      "learning_rate": 9.393333333333334e-07,
      "loss": 0.0098,
      "reward": 0.3139301836490631,
      "reward_std": 0.10886117443442345,
      "reward_std_all": 0.09652956202626228,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9174178540706635,
      "rewards/forensics_tempsamp_iou_reward": 0.3139301836490631,
      "step": 92
    },
    {
      "completion_length": 15.109375,
      "epoch": 0.24766977363515313,
      "grad_norm": 3.75072979927063,
      "kl": 0.14892578125,
      "learning_rate": 9.386666666666666e-07,
      "loss": 0.006,
      "reward": 0.15232504531741142,
      "reward_std": 0.05746118351817131,
      "reward_std_all": 0.05212743580341339,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6330998986959457,
      "rewards/forensics_tempsamp_iou_reward": 0.15232504531741142,
      "step": 93
    },
    {
      "completion_length": 15.59375,
      "epoch": 0.25033288948069243,
      "grad_norm": 6.5391950607299805,
      "kl": 0.2119140625,
      "learning_rate": 9.379999999999998e-07,
      "loss": 0.0085,
      "reward": 0.2610376477241516,
      "reward_std": 0.09712900221347809,
      "reward_std_all": 0.10007153078913689,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8929563164710999,
      "rewards/forensics_tempsamp_iou_reward": 0.2610376477241516,
      "step": 94
    },
    {
      "completion_length": 14.671875,
      "epoch": 0.2529960053262317,
      "grad_norm": 4.8389692306518555,
      "kl": 0.2001953125,
      "learning_rate": 9.373333333333333e-07,
      "loss": 0.008,
      "reward": 0.30080394446849823,
      "reward_std": 0.0947638563811779,
      "reward_std_all": 0.08933298289775848,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8738246262073517,
      "rewards/forensics_tempsamp_iou_reward": 0.30080394446849823,
      "step": 95
    },
    {
      "completion_length": 14.046875,
      "epoch": 0.255659121171771,
      "grad_norm": 5.321652412414551,
      "kl": 0.19482421875,
      "learning_rate": 9.366666666666666e-07,
      "loss": 0.0078,
      "reward": 0.25696173310279846,
      "reward_std": 0.08194573223590851,
      "reward_std_all": 0.08487870544195175,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9013574421405792,
      "rewards/forensics_tempsamp_iou_reward": 0.25696173310279846,
      "step": 96
    },
    {
      "completion_length": 17.671875,
      "epoch": 0.2583222370173103,
      "grad_norm": 3.6727635860443115,
      "kl": 0.20068359375,
      "learning_rate": 9.36e-07,
      "loss": 0.008,
      "reward": 0.15900332480669022,
      "reward_std": 0.07143077254295349,
      "reward_std_all": 0.07597742602229118,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8545317351818085,
      "rewards/forensics_tempsamp_iou_reward": 0.15900332480669022,
      "step": 97
    },
    {
      "completion_length": 16.953125,
      "epoch": 0.2609853528628495,
      "grad_norm": 4.986686706542969,
      "kl": 0.2119140625,
      "learning_rate": 9.353333333333333e-07,
      "loss": 0.0085,
      "reward": 0.20611050724983215,
      "reward_std": 0.09378250688314438,
      "reward_std_all": 0.0895674005150795,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9049619138240814,
      "rewards/forensics_tempsamp_iou_reward": 0.20611050724983215,
      "step": 98
    },
    {
      "completion_length": 21.921875,
      "epoch": 0.2636484687083888,
      "grad_norm": 5.687963962554932,
      "kl": 0.22900390625,
      "learning_rate": 9.346666666666666e-07,
      "loss": 0.0092,
      "reward": 0.1706714704632759,
      "reward_std": 0.0841316357254982,
      "reward_std_all": 0.07597276568412781,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7114229798316956,
      "rewards/forensics_tempsamp_iou_reward": 0.1706714704632759,
      "step": 99
    },
    {
      "completion_length": 16.828125,
      "epoch": 0.2663115845539281,
      "grad_norm": 5.046832084655762,
      "kl": 0.23291015625,
      "learning_rate": 9.34e-07,
      "loss": 0.0093,
      "reward": 0.23906538635492325,
      "reward_std": 0.07171819359064102,
      "reward_std_all": 0.07305681705474854,
      "rewards-max/forensics_tempsamp_iou_reward": 0.610586479306221,
      "rewards/forensics_tempsamp_iou_reward": 0.23906538635492325,
      "step": 100
    },
    {
      "completion_length": 17.09375,
      "epoch": 0.26897470039946736,
      "grad_norm": 5.602178573608398,
      "kl": 0.21240234375,
      "learning_rate": 9.333333333333333e-07,
      "loss": 0.0085,
      "reward": 0.21550114452838898,
      "reward_std": 0.0793910101056099,
      "reward_std_all": 0.0721622183918953,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8678732812404633,
      "rewards/forensics_tempsamp_iou_reward": 0.21550114452838898,
      "step": 101
    },
    {
      "completion_length": 14.234375,
      "epoch": 0.27163781624500666,
      "grad_norm": 6.608160018920898,
      "kl": 0.29736328125,
      "learning_rate": 9.326666666666666e-07,
      "loss": 0.0119,
      "reward": 0.18505951017141342,
      "reward_std": 0.099893007427454,
      "reward_std_all": 0.09438126161694527,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8897272050380707,
      "rewards/forensics_tempsamp_iou_reward": 0.18505951017141342,
      "step": 102
    },
    {
      "completion_length": 15.859375,
      "epoch": 0.27430093209054596,
      "grad_norm": 5.705989360809326,
      "kl": 0.31103515625,
      "learning_rate": 9.32e-07,
      "loss": 0.0125,
      "reward": 0.13494648784399033,
      "reward_std": 0.059762462973594666,
      "reward_std_all": 0.07015790045261383,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6481481194496155,
      "rewards/forensics_tempsamp_iou_reward": 0.13494648784399033,
      "step": 103
    },
    {
      "completion_length": 15.546875,
      "epoch": 0.2769640479360852,
      "grad_norm": 6.921576499938965,
      "kl": 0.373046875,
      "learning_rate": 9.313333333333333e-07,
      "loss": 0.0149,
      "reward": 0.2192130908370018,
      "reward_std": 0.052112625911831856,
      "reward_std_all": 0.06915021315217018,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6506838798522949,
      "rewards/forensics_tempsamp_iou_reward": 0.2192130908370018,
      "step": 104
    },
    {
      "completion_length": 13.578125,
      "epoch": 0.2796271637816245,
      "grad_norm": 7.87930154800415,
      "kl": 0.5517578125,
      "learning_rate": 9.306666666666666e-07,
      "loss": 0.0221,
      "reward": 0.29780636727809906,
      "reward_std": 0.13495466113090515,
      "reward_std_all": 0.13850955665111542,
      "rewards-max/forensics_tempsamp_iou_reward": 0.971704512834549,
      "rewards/forensics_tempsamp_iou_reward": 0.29780636727809906,
      "step": 105
    },
    {
      "completion_length": 14.46875,
      "epoch": 0.2822902796271638,
      "grad_norm": 6.942934513092041,
      "kl": 0.34375,
      "learning_rate": 9.3e-07,
      "loss": 0.0138,
      "reward": 0.30753906071186066,
      "reward_std": 0.15825152397155762,
      "reward_std_all": 0.1503434106707573,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8981092274188995,
      "rewards/forensics_tempsamp_iou_reward": 0.30753906071186066,
      "step": 106
    },
    {
      "completion_length": 14.703125,
      "epoch": 0.28495339547270304,
      "grad_norm": 5.075705528259277,
      "kl": 0.28125,
      "learning_rate": 9.293333333333333e-07,
      "loss": 0.0112,
      "reward": 0.20368745177984238,
      "reward_std": 0.13304156064987183,
      "reward_std_all": 0.1274181418120861,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8939670622348785,
      "rewards/forensics_tempsamp_iou_reward": 0.20368745177984238,
      "step": 107
    },
    {
      "completion_length": 15.34375,
      "epoch": 0.28761651131824234,
      "grad_norm": 5.548544406890869,
      "kl": 0.3017578125,
      "learning_rate": 9.286666666666666e-07,
      "loss": 0.012,
      "reward": 0.18722356483340263,
      "reward_std": 0.07312743738293648,
      "reward_std_all": 0.06785384565591812,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6339828670024872,
      "rewards/forensics_tempsamp_iou_reward": 0.18722356483340263,
      "step": 108
    },
    {
      "completion_length": 14.890625,
      "epoch": 0.29027962716378164,
      "grad_norm": 4.341124057769775,
      "kl": 0.19580078125,
      "learning_rate": 9.28e-07,
      "loss": 0.0078,
      "reward": 0.2817186564207077,
      "reward_std": 0.09170985966920853,
      "reward_std_all": 0.0950218103826046,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9340835809707642,
      "rewards/forensics_tempsamp_iou_reward": 0.2817186564207077,
      "step": 109
    },
    {
      "completion_length": 16.515625,
      "epoch": 0.2929427430093209,
      "grad_norm": 5.029008865356445,
      "kl": 0.24072265625,
      "learning_rate": 9.273333333333333e-07,
      "loss": 0.0096,
      "reward": 0.1696917787194252,
      "reward_std": 0.061668455600738525,
      "reward_std_all": 0.061648592352867126,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7021028697490692,
      "rewards/forensics_tempsamp_iou_reward": 0.1696917787194252,
      "step": 110
    },
    {
      "completion_length": 14.484375,
      "epoch": 0.2956058588548602,
      "grad_norm": 6.615656852722168,
      "kl": 0.3349609375,
      "learning_rate": 9.266666666666665e-07,
      "loss": 0.0134,
      "reward": 0.21611837297677994,
      "reward_std": 0.08911320939660072,
      "reward_std_all": 0.08351297676563263,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7072358429431915,
      "rewards/forensics_tempsamp_iou_reward": 0.21611837297677994,
      "step": 111
    },
    {
      "completion_length": 15.671875,
      "epoch": 0.2982689747003995,
      "grad_norm": 6.725018501281738,
      "kl": 0.40625,
      "learning_rate": 9.26e-07,
      "loss": 0.0162,
      "reward": 0.23858139663934708,
      "reward_std": 0.12489001080393791,
      "reward_std_all": 0.12563449889421463,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7996523380279541,
      "rewards/forensics_tempsamp_iou_reward": 0.23858139663934708,
      "step": 112
    },
    {
      "completion_length": 13.84375,
      "epoch": 0.3009320905459387,
      "grad_norm": 6.31990385055542,
      "kl": 0.318359375,
      "learning_rate": 9.253333333333333e-07,
      "loss": 0.0127,
      "reward": 0.21396566182374954,
      "reward_std": 0.09293917194008827,
      "reward_std_all": 0.08981053531169891,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5979599356651306,
      "rewards/forensics_tempsamp_iou_reward": 0.21396566182374954,
      "step": 113
    },
    {
      "completion_length": 16.578125,
      "epoch": 0.303595206391478,
      "grad_norm": 7.156739711761475,
      "kl": 0.4453125,
      "learning_rate": 9.246666666666666e-07,
      "loss": 0.0178,
      "reward": 0.26015617698431015,
      "reward_std": 0.09991745837032795,
      "reward_std_all": 0.116641815751791,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6953338980674744,
      "rewards/forensics_tempsamp_iou_reward": 0.26015617698431015,
      "step": 114
    },
    {
      "completion_length": 14.515625,
      "epoch": 0.3062583222370173,
      "grad_norm": 6.240458011627197,
      "kl": 0.3759765625,
      "learning_rate": 9.24e-07,
      "loss": 0.015,
      "reward": 0.17859406769275665,
      "reward_std": 0.06893993727862835,
      "reward_std_all": 0.08307220228016376,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4129476547241211,
      "rewards/forensics_tempsamp_iou_reward": 0.17859406769275665,
      "step": 115
    },
    {
      "completion_length": 17.265625,
      "epoch": 0.30892143808255657,
      "grad_norm": 6.294478416442871,
      "kl": 0.35107421875,
      "learning_rate": 9.233333333333333e-07,
      "loss": 0.014,
      "reward": 0.14314351975917816,
      "reward_std": 0.10095968469977379,
      "reward_std_all": 0.10823024436831474,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6407486796379089,
      "rewards/forensics_tempsamp_iou_reward": 0.14314351975917816,
      "step": 116
    },
    {
      "completion_length": 30.390625,
      "epoch": 0.31158455392809586,
      "grad_norm": 4.527400970458984,
      "kl": 0.22314453125,
      "learning_rate": 9.226666666666666e-07,
      "loss": 0.0089,
      "reward": 0.15618764609098434,
      "reward_std": 0.11492087692022324,
      "reward_std_all": 0.12082172930240631,
      "rewards-max/forensics_tempsamp_iou_reward": 0.711214542388916,
      "rewards/forensics_tempsamp_iou_reward": 0.15618764609098434,
      "step": 117
    },
    {
      "completion_length": 18.0,
      "epoch": 0.31424766977363516,
      "grad_norm": 6.059361934661865,
      "kl": 0.24462890625,
      "learning_rate": 9.22e-07,
      "loss": 0.0098,
      "reward": 0.13481050729751587,
      "reward_std": 0.07596230693161488,
      "reward_std_all": 0.07708273082971573,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5532298982143402,
      "rewards/forensics_tempsamp_iou_reward": 0.13481050729751587,
      "step": 118
    },
    {
      "completion_length": 17.59375,
      "epoch": 0.3169107856191744,
      "grad_norm": 6.524276256561279,
      "kl": 0.26171875,
      "learning_rate": 9.213333333333333e-07,
      "loss": 0.0105,
      "reward": 0.15900613367557526,
      "reward_std": 0.09719059243798256,
      "reward_std_all": 0.10354545339941978,
      "rewards-max/forensics_tempsamp_iou_reward": 0.621153861284256,
      "rewards/forensics_tempsamp_iou_reward": 0.15900613367557526,
      "step": 119
    },
    {
      "completion_length": 20.484375,
      "epoch": 0.3195739014647137,
      "grad_norm": 6.109954833984375,
      "kl": 0.3515625,
      "learning_rate": 9.206666666666666e-07,
      "loss": 0.0141,
      "reward": 0.2523393929004669,
      "reward_std": 0.12779353559017181,
      "reward_std_all": 0.13348908722400665,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9032883644104004,
      "rewards/forensics_tempsamp_iou_reward": 0.2523393929004669,
      "step": 120
    },
    {
      "completion_length": 19.859375,
      "epoch": 0.322237017310253,
      "grad_norm": 5.845400810241699,
      "kl": 0.3095703125,
      "learning_rate": 9.2e-07,
      "loss": 0.0124,
      "reward": 0.16163105890154839,
      "reward_std": 0.0999923013150692,
      "reward_std_all": 0.09490956738591194,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5386574119329453,
      "rewards/forensics_tempsamp_iou_reward": 0.16163105890154839,
      "step": 121
    },
    {
      "completion_length": 19.375,
      "epoch": 0.3249001331557923,
      "grad_norm": 6.164781093597412,
      "kl": 0.3076171875,
      "learning_rate": 9.193333333333333e-07,
      "loss": 0.0123,
      "reward": 0.16797707974910736,
      "reward_std": 0.08040831610560417,
      "reward_std_all": 0.09772976487874985,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6086001694202423,
      "rewards/forensics_tempsamp_iou_reward": 0.16797707974910736,
      "step": 122
    },
    {
      "completion_length": 20.75,
      "epoch": 0.32756324900133155,
      "grad_norm": 5.084867000579834,
      "kl": 0.3330078125,
      "learning_rate": 9.186666666666666e-07,
      "loss": 0.0133,
      "reward": 0.13830333948135376,
      "reward_std": 0.06961050443351269,
      "reward_std_all": 0.06779090501368046,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6350366175174713,
      "rewards/forensics_tempsamp_iou_reward": 0.13830333948135376,
      "step": 123
    },
    {
      "completion_length": 21.78125,
      "epoch": 0.33022636484687085,
      "grad_norm": 6.689507961273193,
      "kl": 0.431640625,
      "learning_rate": 9.18e-07,
      "loss": 0.0173,
      "reward": 0.22974997013807297,
      "reward_std": 0.13001257181167603,
      "reward_std_all": 0.12526678666472435,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9082722663879395,
      "rewards/forensics_tempsamp_iou_reward": 0.22974997013807297,
      "step": 124
    },
    {
      "completion_length": 19.125,
      "epoch": 0.33288948069241014,
      "grad_norm": 7.299772262573242,
      "kl": 0.724609375,
      "learning_rate": 9.173333333333333e-07,
      "loss": 0.029,
      "reward": 0.13876403868198395,
      "reward_std": 0.07134007103741169,
      "reward_std_all": 0.0665673203766346,
      "rewards-max/forensics_tempsamp_iou_reward": 0.44828571379184723,
      "rewards/forensics_tempsamp_iou_reward": 0.13876403868198395,
      "step": 125
    },
    {
      "completion_length": 19.859375,
      "epoch": 0.3355525965379494,
      "grad_norm": 7.863882541656494,
      "kl": 0.380859375,
      "learning_rate": 9.166666666666665e-07,
      "loss": 0.0152,
      "reward": 0.14180779457092285,
      "reward_std": 0.09396367892622948,
      "reward_std_all": 0.09383592754602432,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5177953988313675,
      "rewards/forensics_tempsamp_iou_reward": 0.14180779457092285,
      "step": 126
    },
    {
      "completion_length": 24.59375,
      "epoch": 0.3382157123834887,
      "grad_norm": 5.125712871551514,
      "kl": 0.4091796875,
      "learning_rate": 9.16e-07,
      "loss": 0.0164,
      "reward": 0.22993513941764832,
      "reward_std": 0.123363446444273,
      "reward_std_all": 0.11484868824481964,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8855730891227722,
      "rewards/forensics_tempsamp_iou_reward": 0.22993513941764832,
      "step": 127
    },
    {
      "completion_length": 29.046875,
      "epoch": 0.340878828229028,
      "grad_norm": 6.600436687469482,
      "kl": 0.4189453125,
      "learning_rate": 9.153333333333332e-07,
      "loss": 0.0168,
      "reward": 0.19715754687786102,
      "reward_std": 0.11293575167655945,
      "reward_std_all": 0.1126326285302639,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6389091610908508,
      "rewards/forensics_tempsamp_iou_reward": 0.19715754687786102,
      "step": 128
    },
    {
      "completion_length": 31.640625,
      "epoch": 0.34354194407456723,
      "grad_norm": 5.314303874969482,
      "kl": 0.568359375,
      "learning_rate": 9.146666666666666e-07,
      "loss": 0.0227,
      "reward": 0.22010958939790726,
      "reward_std": 0.10821885243058205,
      "reward_std_all": 0.12178098410367966,
      "rewards-max/forensics_tempsamp_iou_reward": 0.800067275762558,
      "rewards/forensics_tempsamp_iou_reward": 0.22010958939790726,
      "step": 129
    },
    {
      "completion_length": 25.421875,
      "epoch": 0.34620505992010653,
      "grad_norm": 4.695012092590332,
      "kl": 0.451171875,
      "learning_rate": 9.14e-07,
      "loss": 0.0181,
      "reward": 0.24024264514446259,
      "reward_std": 0.13143527507781982,
      "reward_std_all": 0.12382430583238602,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7928743958473206,
      "rewards/forensics_tempsamp_iou_reward": 0.24024264514446259,
      "step": 130
    },
    {
      "completion_length": 20.796875,
      "epoch": 0.3488681757656458,
      "grad_norm": 4.526851177215576,
      "kl": 0.4375,
      "learning_rate": 9.133333333333333e-07,
      "loss": 0.0175,
      "reward": 0.21351895481348038,
      "reward_std": 0.09958484396338463,
      "reward_std_all": 0.11155426129698753,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6917673647403717,
      "rewards/forensics_tempsamp_iou_reward": 0.21351895481348038,
      "step": 131
    },
    {
      "completion_length": 30.140625,
      "epoch": 0.35153129161118507,
      "grad_norm": 5.371805191040039,
      "kl": 0.373046875,
      "learning_rate": 9.126666666666666e-07,
      "loss": 0.0149,
      "reward": 0.21769164502620697,
      "reward_std": 0.13897250220179558,
      "reward_std_all": 0.12814044207334518,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9167590737342834,
      "rewards/forensics_tempsamp_iou_reward": 0.21769164502620697,
      "step": 132
    },
    {
      "completion_length": 20.671875,
      "epoch": 0.35419440745672437,
      "grad_norm": 5.109017848968506,
      "kl": 0.4619140625,
      "learning_rate": 9.12e-07,
      "loss": 0.0185,
      "reward": 0.23867379128932953,
      "reward_std": 0.11350443214178085,
      "reward_std_all": 0.1146710067987442,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8823099434375763,
      "rewards/forensics_tempsamp_iou_reward": 0.23867379128932953,
      "step": 133
    },
    {
      "completion_length": 16.546875,
      "epoch": 0.35685752330226367,
      "grad_norm": 257.24066162109375,
      "kl": 3.779296875,
      "learning_rate": 9.113333333333333e-07,
      "loss": 0.1511,
      "reward": 0.11385460570454597,
      "reward_std": 0.06893457844853401,
      "reward_std_all": 0.08234760165214539,
      "rewards-max/forensics_tempsamp_iou_reward": 0.48752665519714355,
      "rewards/forensics_tempsamp_iou_reward": 0.11385460570454597,
      "step": 134
    },
    {
      "completion_length": 20.609375,
      "epoch": 0.3595206391478029,
      "grad_norm": 5.601149559020996,
      "kl": 0.4072265625,
      "learning_rate": 9.106666666666666e-07,
      "loss": 0.0163,
      "reward": 0.28641802072525024,
      "reward_std": 0.14814352244138718,
      "reward_std_all": 0.15330715477466583,
      "rewards-max/forensics_tempsamp_iou_reward": 0.869685173034668,
      "rewards/forensics_tempsamp_iou_reward": 0.28641802072525024,
      "step": 135
    },
    {
      "completion_length": 18.765625,
      "epoch": 0.3621837549933422,
      "grad_norm": 5.39362096786499,
      "kl": 0.3466796875,
      "learning_rate": 9.1e-07,
      "loss": 0.0138,
      "reward": 0.20718453079462051,
      "reward_std": 0.11442086845636368,
      "reward_std_all": 0.12207755073904991,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9411888420581818,
      "rewards/forensics_tempsamp_iou_reward": 0.20718453079462051,
      "step": 136
    },
    {
      "completion_length": 21.75,
      "epoch": 0.3648468708388815,
      "grad_norm": 5.1544060707092285,
      "kl": 0.3359375,
      "learning_rate": 9.093333333333333e-07,
      "loss": 0.0134,
      "reward": 0.1365998350083828,
      "reward_std": 0.06827186420559883,
      "reward_std_all": 0.06857496500015259,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4907572716474533,
      "rewards/forensics_tempsamp_iou_reward": 0.1365998350083828,
      "step": 137
    },
    {
      "completion_length": 18.078125,
      "epoch": 0.36750998668442075,
      "grad_norm": 5.655546188354492,
      "kl": 0.23046875,
      "learning_rate": 9.086666666666666e-07,
      "loss": 0.0092,
      "reward": 0.2451622039079666,
      "reward_std": 0.08348044380545616,
      "reward_std_all": 0.08639823272824287,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6716978847980499,
      "rewards/forensics_tempsamp_iou_reward": 0.2451622039079666,
      "step": 138
    },
    {
      "completion_length": 15.765625,
      "epoch": 0.37017310252996005,
      "grad_norm": 5.101880073547363,
      "kl": 0.232421875,
      "learning_rate": 9.08e-07,
      "loss": 0.0093,
      "reward": 0.21984105557203293,
      "reward_std": 0.07742179930210114,
      "reward_std_all": 0.09936180710792542,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8425950706005096,
      "rewards/forensics_tempsamp_iou_reward": 0.21984105557203293,
      "step": 139
    },
    {
      "completion_length": 14.609375,
      "epoch": 0.37283621837549935,
      "grad_norm": 4.859246253967285,
      "kl": 0.2353515625,
      "learning_rate": 9.073333333333333e-07,
      "loss": 0.0094,
      "reward": 0.20968245714902878,
      "reward_std": 0.08566254377365112,
      "reward_std_all": 0.1021578349173069,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6801565885543823,
      "rewards/forensics_tempsamp_iou_reward": 0.20968245714902878,
      "step": 140
    },
    {
      "completion_length": 17.34375,
      "epoch": 0.3754993342210386,
      "grad_norm": 4.803069114685059,
      "kl": 0.251953125,
      "learning_rate": 9.066666666666665e-07,
      "loss": 0.0101,
      "reward": 0.2586066946387291,
      "reward_std": 0.11358427628874779,
      "reward_std_all": 0.11514517664909363,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7679294645786285,
      "rewards/forensics_tempsamp_iou_reward": 0.2586066946387291,
      "step": 141
    },
    {
      "completion_length": 14.796875,
      "epoch": 0.3781624500665779,
      "grad_norm": 5.143069267272949,
      "kl": 0.23046875,
      "learning_rate": 9.06e-07,
      "loss": 0.0092,
      "reward": 0.27895503491163254,
      "reward_std": 0.09311359375715256,
      "reward_std_all": 0.09251859784126282,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8834776282310486,
      "rewards/forensics_tempsamp_iou_reward": 0.27895503491163254,
      "step": 142
    },
    {
      "completion_length": 13.625,
      "epoch": 0.3808255659121172,
      "grad_norm": 5.821961879730225,
      "kl": 0.16943359375,
      "learning_rate": 9.053333333333332e-07,
      "loss": 0.0068,
      "reward": 0.30427631735801697,
      "reward_std": 0.07415646500885487,
      "reward_std_all": 0.10554145649075508,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7529230415821075,
      "rewards/forensics_tempsamp_iou_reward": 0.30427631735801697,
      "step": 143
    },
    {
      "completion_length": 16.3125,
      "epoch": 0.38348868175765644,
      "grad_norm": 4.303036689758301,
      "kl": 0.283203125,
      "learning_rate": 9.046666666666666e-07,
      "loss": 0.0113,
      "reward": 0.18027542531490326,
      "reward_std": 0.08721234276890755,
      "reward_std_all": 0.08402284234762192,
      "rewards-max/forensics_tempsamp_iou_reward": 0.82245072722435,
      "rewards/forensics_tempsamp_iou_reward": 0.18027542531490326,
      "step": 144
    },
    {
      "completion_length": 20.453125,
      "epoch": 0.38615179760319573,
      "grad_norm": 5.678048610687256,
      "kl": 0.21240234375,
      "learning_rate": 9.039999999999999e-07,
      "loss": 0.0085,
      "reward": 0.24798190593719482,
      "reward_std": 0.08878878876566887,
      "reward_std_all": 0.08570936694741249,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8006222546100616,
      "rewards/forensics_tempsamp_iou_reward": 0.24798190593719482,
      "step": 145
    },
    {
      "completion_length": 14.8125,
      "epoch": 0.38881491344873503,
      "grad_norm": 5.962091445922852,
      "kl": 0.157470703125,
      "learning_rate": 9.033333333333333e-07,
      "loss": 0.0063,
      "reward": 0.16951275616884232,
      "reward_std": 0.07589556649327278,
      "reward_std_all": 0.08315017819404602,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7030538022518158,
      "rewards/forensics_tempsamp_iou_reward": 0.16951275616884232,
      "step": 146
    },
    {
      "completion_length": 15.34375,
      "epoch": 0.3914780292942743,
      "grad_norm": 5.565754413604736,
      "kl": 0.2314453125,
      "learning_rate": 9.026666666666665e-07,
      "loss": 0.0093,
      "reward": 0.2722817286849022,
      "reward_std": 0.07503166049718857,
      "reward_std_all": 0.08459698408842087,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8595075905323029,
      "rewards/forensics_tempsamp_iou_reward": 0.2722817286849022,
      "step": 147
    },
    {
      "completion_length": 14.9375,
      "epoch": 0.3941411451398136,
      "grad_norm": 5.0496721267700195,
      "kl": 0.27880859375,
      "learning_rate": 9.02e-07,
      "loss": 0.0112,
      "reward": 0.2758677527308464,
      "reward_std": 0.10495080053806305,
      "reward_std_all": 0.1088892724364996,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6317821443080902,
      "rewards/forensics_tempsamp_iou_reward": 0.2758677527308464,
      "step": 148
    },
    {
      "completion_length": 16.140625,
      "epoch": 0.3968042609853529,
      "grad_norm": 5.023301601409912,
      "kl": 0.2451171875,
      "learning_rate": 9.013333333333333e-07,
      "loss": 0.0098,
      "reward": 0.2994181513786316,
      "reward_std": 0.08994585648179054,
      "reward_std_all": 0.10376648232340813,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9751749336719513,
      "rewards/forensics_tempsamp_iou_reward": 0.2994181513786316,
      "step": 149
    },
    {
      "completion_length": 18.21875,
      "epoch": 0.3994673768308921,
      "grad_norm": 4.680842399597168,
      "kl": 0.16064453125,
      "learning_rate": 9.006666666666666e-07,
      "loss": 0.0064,
      "reward": 0.3874604105949402,
      "reward_std": 0.08363637700676918,
      "reward_std_all": 0.08847526833415031,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9229276180267334,
      "rewards/forensics_tempsamp_iou_reward": 0.3874604105949402,
      "step": 150
    },
    {
      "completion_length": 15.734375,
      "epoch": 0.4021304926764314,
      "grad_norm": 5.451523780822754,
      "kl": 0.25244140625,
      "learning_rate": 9e-07,
      "loss": 0.0101,
      "reward": 0.3268778622150421,
      "reward_std": 0.10060706734657288,
      "reward_std_all": 0.10609489679336548,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9615958631038666,
      "rewards/forensics_tempsamp_iou_reward": 0.3268778622150421,
      "step": 151
    },
    {
      "completion_length": 17.171875,
      "epoch": 0.4047936085219707,
      "grad_norm": 5.143345355987549,
      "kl": 0.2646484375,
      "learning_rate": 8.993333333333333e-07,
      "loss": 0.0106,
      "reward": 0.25344591587781906,
      "reward_std": 0.09150472283363342,
      "reward_std_all": 0.09778424352407455,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8967573940753937,
      "rewards/forensics_tempsamp_iou_reward": 0.25344591587781906,
      "step": 152
    },
    {
      "completion_length": 17.578125,
      "epoch": 0.40745672436750996,
      "grad_norm": 4.41171932220459,
      "kl": 0.26904296875,
      "learning_rate": 8.986666666666666e-07,
      "loss": 0.0107,
      "reward": 0.24137066304683685,
      "reward_std": 0.07788874208927155,
      "reward_std_all": 0.0817299447953701,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9628454446792603,
      "rewards/forensics_tempsamp_iou_reward": 0.24137066304683685,
      "step": 153
    },
    {
      "completion_length": 14.09375,
      "epoch": 0.41011984021304926,
      "grad_norm": 6.655637741088867,
      "kl": 0.3193359375,
      "learning_rate": 8.98e-07,
      "loss": 0.0128,
      "reward": 0.20405440777540207,
      "reward_std": 0.04227925464510918,
      "reward_std_all": 0.05959223955869675,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7230685651302338,
      "rewards/forensics_tempsamp_iou_reward": 0.20405440777540207,
      "step": 154
    },
    {
      "completion_length": 17.234375,
      "epoch": 0.41278295605858856,
      "grad_norm": 7.089432716369629,
      "kl": 0.337890625,
      "learning_rate": 8.973333333333333e-07,
      "loss": 0.0135,
      "reward": 0.2942505329847336,
      "reward_std": 0.11224822327494621,
      "reward_std_all": 0.10564891248941422,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9408971071243286,
      "rewards/forensics_tempsamp_iou_reward": 0.2942505329847336,
      "step": 155
    },
    {
      "completion_length": 14.625,
      "epoch": 0.41544607190412786,
      "grad_norm": 5.7838592529296875,
      "kl": 0.365234375,
      "learning_rate": 8.966666666666666e-07,
      "loss": 0.0146,
      "reward": 0.18601958453655243,
      "reward_std": 0.05438989493995905,
      "reward_std_all": 0.05722907464951277,
      "rewards-max/forensics_tempsamp_iou_reward": 0.581805557012558,
      "rewards/forensics_tempsamp_iou_reward": 0.18601958453655243,
      "step": 156
    },
    {
      "completion_length": 16.125,
      "epoch": 0.4181091877496671,
      "grad_norm": 5.655130386352539,
      "kl": 0.3564453125,
      "learning_rate": 8.96e-07,
      "loss": 0.0143,
      "reward": 0.20829788595438004,
      "reward_std": 0.06614774093031883,
      "reward_std_all": 0.058919522911310196,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9505723714828491,
      "rewards/forensics_tempsamp_iou_reward": 0.20829788595438004,
      "step": 157
    },
    {
      "completion_length": 15.078125,
      "epoch": 0.4207723035952064,
      "grad_norm": 6.783344745635986,
      "kl": 0.39453125,
      "learning_rate": 8.953333333333332e-07,
      "loss": 0.0158,
      "reward": 0.17855935543775558,
      "reward_std": 0.10850276798009872,
      "reward_std_all": 0.10073276981711388,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5998384207487106,
      "rewards/forensics_tempsamp_iou_reward": 0.17855935543775558,
      "step": 158
    },
    {
      "completion_length": 15.078125,
      "epoch": 0.4234354194407457,
      "grad_norm": 7.0881147384643555,
      "kl": 0.427734375,
      "learning_rate": 8.946666666666667e-07,
      "loss": 0.0171,
      "reward": 0.30222564190626144,
      "reward_std": 0.07648799195885658,
      "reward_std_all": 0.08777261525392532,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9370359480381012,
      "rewards/forensics_tempsamp_iou_reward": 0.30222564190626144,
      "step": 159
    },
    {
      "completion_length": 17.578125,
      "epoch": 0.42609853528628494,
      "grad_norm": 5.260518550872803,
      "kl": 0.32421875,
      "learning_rate": 8.939999999999999e-07,
      "loss": 0.0129,
      "reward": 0.29316993057727814,
      "reward_std": 0.0718254093080759,
      "reward_std_all": 0.06816758960485458,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8671257197856903,
      "rewards/forensics_tempsamp_iou_reward": 0.29316993057727814,
      "step": 160
    },
    {
      "completion_length": 16.671875,
      "epoch": 0.42876165113182424,
      "grad_norm": 5.144521236419678,
      "kl": 0.4013671875,
      "learning_rate": 8.933333333333333e-07,
      "loss": 0.0161,
      "reward": 0.2444058507680893,
      "reward_std": 0.10119420662522316,
      "reward_std_all": 0.09557737410068512,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7799166440963745,
      "rewards/forensics_tempsamp_iou_reward": 0.2444058507680893,
      "step": 161
    },
    {
      "completion_length": 15.8125,
      "epoch": 0.43142476697736354,
      "grad_norm": 5.658234119415283,
      "kl": 0.4638671875,
      "learning_rate": 8.926666666666666e-07,
      "loss": 0.0185,
      "reward": 0.2835589349269867,
      "reward_std": 0.05565483681857586,
      "reward_std_all": 0.07474526017904282,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9389864802360535,
      "rewards/forensics_tempsamp_iou_reward": 0.2835589349269867,
      "step": 162
    },
    {
      "completion_length": 15.5,
      "epoch": 0.4340878828229028,
      "grad_norm": 4.369124889373779,
      "kl": 0.3583984375,
      "learning_rate": 8.92e-07,
      "loss": 0.0143,
      "reward": 0.2763614058494568,
      "reward_std": 0.13068822398781776,
      "reward_std_all": 0.1281115487217903,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9142156839370728,
      "rewards/forensics_tempsamp_iou_reward": 0.2763614058494568,
      "step": 163
    },
    {
      "completion_length": 15.09375,
      "epoch": 0.4367509986684421,
      "grad_norm": 6.987516403198242,
      "kl": 0.4111328125,
      "learning_rate": 8.913333333333332e-07,
      "loss": 0.0165,
      "reward": 0.11625808849930763,
      "reward_std": 0.0494169145822525,
      "reward_std_all": 0.047199320048093796,
      "rewards-max/forensics_tempsamp_iou_reward": 0.49985118210315704,
      "rewards/forensics_tempsamp_iou_reward": 0.11625808849930763,
      "step": 164
    },
    {
      "completion_length": 13.421875,
      "epoch": 0.4394141145139814,
      "grad_norm": 5.450296401977539,
      "kl": 0.328125,
      "learning_rate": 8.906666666666667e-07,
      "loss": 0.0131,
      "reward": 0.3197374939918518,
      "reward_std": 0.07743780687451363,
      "reward_std_all": 0.07421955838799477,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7557641863822937,
      "rewards/forensics_tempsamp_iou_reward": 0.3197374939918518,
      "step": 165
    },
    {
      "completion_length": 13.71875,
      "epoch": 0.4420772303595206,
      "grad_norm": 7.033100605010986,
      "kl": 0.2265625,
      "learning_rate": 8.9e-07,
      "loss": 0.0091,
      "reward": 0.38241682946681976,
      "reward_std": 0.11938118934631348,
      "reward_std_all": 0.13148964941501617,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8862903416156769,
      "rewards/forensics_tempsamp_iou_reward": 0.38241682946681976,
      "step": 166
    },
    {
      "completion_length": 14.765625,
      "epoch": 0.4447403462050599,
      "grad_norm": 4.685544967651367,
      "kl": 0.1923828125,
      "learning_rate": 8.893333333333333e-07,
      "loss": 0.0077,
      "reward": 0.25910016894340515,
      "reward_std": 0.08024754002690315,
      "reward_std_all": 0.08288693614304066,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9670827090740204,
      "rewards/forensics_tempsamp_iou_reward": 0.25910016894340515,
      "step": 167
    },
    {
      "completion_length": 17.296875,
      "epoch": 0.4474034620505992,
      "grad_norm": 7.711524963378906,
      "kl": 0.32373046875,
      "learning_rate": 8.886666666666667e-07,
      "loss": 0.0129,
      "reward": 0.234516479074955,
      "reward_std": 0.1030810009688139,
      "reward_std_all": 0.10340221226215363,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6764455735683441,
      "rewards/forensics_tempsamp_iou_reward": 0.234516479074955,
      "step": 168
    },
    {
      "completion_length": 22.359375,
      "epoch": 0.45006657789613846,
      "grad_norm": 5.34246826171875,
      "kl": 0.2890625,
      "learning_rate": 8.88e-07,
      "loss": 0.0115,
      "reward": 0.3229261338710785,
      "reward_std": 0.09294774383306503,
      "reward_std_all": 0.10664921626448631,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9480381309986115,
      "rewards/forensics_tempsamp_iou_reward": 0.3229261338710785,
      "step": 169
    },
    {
      "completion_length": 15.234375,
      "epoch": 0.45272969374167776,
      "grad_norm": 6.850934028625488,
      "kl": 0.3583984375,
      "learning_rate": 8.873333333333333e-07,
      "loss": 0.0143,
      "reward": 0.16103361546993256,
      "reward_std": 0.0722191147506237,
      "reward_std_all": 0.07289769127964973,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5594594776630402,
      "rewards/forensics_tempsamp_iou_reward": 0.16103361546993256,
      "step": 170
    },
    {
      "completion_length": 13.640625,
      "epoch": 0.45539280958721706,
      "grad_norm": 5.588840961456299,
      "kl": 0.29296875,
      "learning_rate": 8.866666666666667e-07,
      "loss": 0.0117,
      "reward": 0.2579218000173569,
      "reward_std": 0.05994521826505661,
      "reward_std_all": 0.07443065941333771,
      "rewards-max/forensics_tempsamp_iou_reward": 0.832854151725769,
      "rewards/forensics_tempsamp_iou_reward": 0.2579218000173569,
      "step": 171
    },
    {
      "completion_length": 15.1875,
      "epoch": 0.4580559254327563,
      "grad_norm": 7.133452415466309,
      "kl": 0.349609375,
      "learning_rate": 8.86e-07,
      "loss": 0.014,
      "reward": 0.21830161660909653,
      "reward_std": 0.11682704836130142,
      "reward_std_all": 0.10598874092102051,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7529025971889496,
      "rewards/forensics_tempsamp_iou_reward": 0.21830161660909653,
      "step": 172
    },
    {
      "completion_length": 13.890625,
      "epoch": 0.4607190412782956,
      "grad_norm": 5.702676296234131,
      "kl": 0.29736328125,
      "learning_rate": 8.853333333333332e-07,
      "loss": 0.0119,
      "reward": 0.2227611020207405,
      "reward_std": 0.07374121807515621,
      "reward_std_all": 0.0743633471429348,
      "rewards-max/forensics_tempsamp_iou_reward": 0.77311572432518,
      "rewards/forensics_tempsamp_iou_reward": 0.2227611020207405,
      "step": 173
    },
    {
      "completion_length": 13.328125,
      "epoch": 0.4633821571238349,
      "grad_norm": 6.001574516296387,
      "kl": 0.3369140625,
      "learning_rate": 8.846666666666667e-07,
      "loss": 0.0135,
      "reward": 0.2731288820505142,
      "reward_std": 0.08031130954623222,
      "reward_std_all": 0.0882757268846035,
      "rewards-max/forensics_tempsamp_iou_reward": 0.902482271194458,
      "rewards/forensics_tempsamp_iou_reward": 0.2731288820505142,
      "step": 174
    },
    {
      "completion_length": 13.578125,
      "epoch": 0.46604527296937415,
      "grad_norm": 6.422615051269531,
      "kl": 0.306640625,
      "learning_rate": 8.839999999999999e-07,
      "loss": 0.0122,
      "reward": 0.2239302173256874,
      "reward_std": 0.07233496382832527,
      "reward_std_all": 0.06886404380202293,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7978464365005493,
      "rewards/forensics_tempsamp_iou_reward": 0.2239302173256874,
      "step": 175
    },
    {
      "completion_length": 13.359375,
      "epoch": 0.46870838881491345,
      "grad_norm": 7.2757415771484375,
      "kl": 0.17041015625,
      "learning_rate": 8.833333333333333e-07,
      "loss": 0.0068,
      "reward": 0.4491385668516159,
      "reward_std": 0.1072426550090313,
      "reward_std_all": 0.11249497532844543,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9738103449344635,
      "rewards/forensics_tempsamp_iou_reward": 0.4491385668516159,
      "step": 176
    },
    {
      "completion_length": 16.90625,
      "epoch": 0.47137150466045274,
      "grad_norm": 5.741480827331543,
      "kl": 0.3232421875,
      "learning_rate": 8.826666666666666e-07,
      "loss": 0.013,
      "reward": 0.15780166164040565,
      "reward_std": 0.07794825173914433,
      "reward_std_all": 0.07548090070486069,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5809319019317627,
      "rewards/forensics_tempsamp_iou_reward": 0.15780166164040565,
      "step": 177
    },
    {
      "completion_length": 14.125,
      "epoch": 0.474034620505992,
      "grad_norm": 5.674465179443359,
      "kl": 0.29345703125,
      "learning_rate": 8.82e-07,
      "loss": 0.0118,
      "reward": 0.21818261593580246,
      "reward_std": 0.07569944858551025,
      "reward_std_all": 0.0787174366414547,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7990235984325409,
      "rewards/forensics_tempsamp_iou_reward": 0.21818261593580246,
      "step": 178
    },
    {
      "completion_length": 13.640625,
      "epoch": 0.4766977363515313,
      "grad_norm": 5.4412078857421875,
      "kl": 0.23828125,
      "learning_rate": 8.813333333333332e-07,
      "loss": 0.0095,
      "reward": 0.32831060886383057,
      "reward_std": 0.060482315719127655,
      "reward_std_all": 0.07459389045834541,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9546418190002441,
      "rewards/forensics_tempsamp_iou_reward": 0.32831060886383057,
      "step": 179
    },
    {
      "completion_length": 15.5,
      "epoch": 0.4793608521970706,
      "grad_norm": 5.783870697021484,
      "kl": 0.25732421875,
      "learning_rate": 8.806666666666667e-07,
      "loss": 0.0103,
      "reward": 0.1796475276350975,
      "reward_std": 0.09163844957947731,
      "reward_std_all": 0.09215892106294632,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8623456954956055,
      "rewards/forensics_tempsamp_iou_reward": 0.1796475276350975,
      "step": 180
    },
    {
      "completion_length": 13.96875,
      "epoch": 0.48202396804260983,
      "grad_norm": 6.245997905731201,
      "kl": 0.18701171875,
      "learning_rate": 8.799999999999999e-07,
      "loss": 0.0075,
      "reward": 0.2883293032646179,
      "reward_std": 0.07512363791465759,
      "reward_std_all": 0.07668144255876541,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8425370454788208,
      "rewards/forensics_tempsamp_iou_reward": 0.2883293032646179,
      "step": 181
    },
    {
      "completion_length": 16.078125,
      "epoch": 0.48468708388814913,
      "grad_norm": 7.314426898956299,
      "kl": 0.349609375,
      "learning_rate": 8.793333333333333e-07,
      "loss": 0.014,
      "reward": 0.15365931391716003,
      "reward_std": 0.08979358151555061,
      "reward_std_all": 0.07981661148369312,
      "rewards-max/forensics_tempsamp_iou_reward": 0.713553249835968,
      "rewards/forensics_tempsamp_iou_reward": 0.15365931391716003,
      "step": 182
    },
    {
      "completion_length": 14.34375,
      "epoch": 0.4873501997336884,
      "grad_norm": 6.670960903167725,
      "kl": 0.236328125,
      "learning_rate": 8.786666666666666e-07,
      "loss": 0.0095,
      "reward": 0.25855424255132675,
      "reward_std": 0.11305621266365051,
      "reward_std_all": 0.11589882150292397,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8459064364433289,
      "rewards/forensics_tempsamp_iou_reward": 0.25855424255132675,
      "step": 183
    },
    {
      "completion_length": 16.703125,
      "epoch": 0.49001331557922767,
      "grad_norm": 5.998586177825928,
      "kl": 0.458984375,
      "learning_rate": 8.78e-07,
      "loss": 0.0183,
      "reward": 0.316829614341259,
      "reward_std": 0.0842905156314373,
      "reward_std_all": 0.0977017804980278,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8007316887378693,
      "rewards/forensics_tempsamp_iou_reward": 0.316829614341259,
      "step": 184
    },
    {
      "completion_length": 16.578125,
      "epoch": 0.49267643142476697,
      "grad_norm": 12.813743591308594,
      "kl": 0.63671875,
      "learning_rate": 8.773333333333332e-07,
      "loss": 0.0255,
      "reward": 0.21835431456565857,
      "reward_std": 0.08865295350551605,
      "reward_std_all": 0.09051407128572464,
      "rewards-max/forensics_tempsamp_iou_reward": 0.726782351732254,
      "rewards/forensics_tempsamp_iou_reward": 0.21835431456565857,
      "step": 185
    },
    {
      "completion_length": 16.875,
      "epoch": 0.49533954727030627,
      "grad_norm": 6.877189636230469,
      "kl": 0.388671875,
      "learning_rate": 8.766666666666667e-07,
      "loss": 0.0156,
      "reward": 0.2726806178689003,
      "reward_std": 0.08969824761152267,
      "reward_std_all": 0.09684697911143303,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8065410256385803,
      "rewards/forensics_tempsamp_iou_reward": 0.2726806178689003,
      "step": 186
    },
    {
      "completion_length": 13.75,
      "epoch": 0.4980026631158455,
      "grad_norm": 5.900713920593262,
      "kl": 0.25,
      "learning_rate": 8.76e-07,
      "loss": 0.01,
      "reward": 0.21516700088977814,
      "reward_std": 0.07314258441329002,
      "reward_std_all": 0.06556654162704945,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8255793452262878,
      "rewards/forensics_tempsamp_iou_reward": 0.21516700088977814,
      "step": 187
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.5006657789613849,
      "grad_norm": 4.980974197387695,
      "kl": 0.19384765625,
      "learning_rate": 8.753333333333332e-07,
      "loss": 0.0078,
      "reward": 0.23628738522529602,
      "reward_std": 0.10097991675138474,
      "reward_std_all": 0.11705559119582176,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8431428670883179,
      "rewards/forensics_tempsamp_iou_reward": 0.23628738522529602,
      "step": 188
    },
    {
      "completion_length": 19.0625,
      "epoch": 0.5033288948069241,
      "grad_norm": 4.749919891357422,
      "kl": 0.2216796875,
      "learning_rate": 8.746666666666667e-07,
      "loss": 0.0089,
      "reward": 0.17854968458414078,
      "reward_std": 0.0697649847716093,
      "reward_std_all": 0.0922616645693779,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7040354311466217,
      "rewards/forensics_tempsamp_iou_reward": 0.17854968458414078,
      "step": 189
    },
    {
      "completion_length": 15.71875,
      "epoch": 0.5059920106524634,
      "grad_norm": 7.533915996551514,
      "kl": 0.3740234375,
      "learning_rate": 8.739999999999999e-07,
      "loss": 0.0149,
      "reward": 0.3575435280799866,
      "reward_std": 0.17345690727233887,
      "reward_std_all": 0.17067328095436096,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8403316140174866,
      "rewards/forensics_tempsamp_iou_reward": 0.3575435280799866,
      "step": 190
    },
    {
      "completion_length": 23.109375,
      "epoch": 0.5086551264980027,
      "grad_norm": 5.1989874839782715,
      "kl": 0.259765625,
      "learning_rate": 8.733333333333333e-07,
      "loss": 0.0104,
      "reward": 0.13216662406921387,
      "reward_std": 0.06494127959012985,
      "reward_std_all": 0.0793427862226963,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5345029085874557,
      "rewards/forensics_tempsamp_iou_reward": 0.13216662406921387,
      "step": 191
    },
    {
      "completion_length": 23.34375,
      "epoch": 0.511318242343542,
      "grad_norm": 4.894744396209717,
      "kl": 0.412109375,
      "learning_rate": 8.726666666666666e-07,
      "loss": 0.0165,
      "reward": 0.3089675009250641,
      "reward_std": 0.09502125158905983,
      "reward_std_all": 0.11792311817407608,
      "rewards-max/forensics_tempsamp_iou_reward": 0.935375839471817,
      "rewards/forensics_tempsamp_iou_reward": 0.3089675009250641,
      "step": 192
    },
    {
      "completion_length": 19.015625,
      "epoch": 0.5139813581890812,
      "grad_norm": 4.8290114402771,
      "kl": 0.265625,
      "learning_rate": 8.72e-07,
      "loss": 0.0106,
      "reward": 0.2917100638151169,
      "reward_std": 0.12376051768660545,
      "reward_std_all": 0.1271645501255989,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9130734205245972,
      "rewards/forensics_tempsamp_iou_reward": 0.2917100638151169,
      "step": 193
    },
    {
      "completion_length": 17.4375,
      "epoch": 0.5166444740346205,
      "grad_norm": 4.930580139160156,
      "kl": 0.177734375,
      "learning_rate": 8.713333333333332e-07,
      "loss": 0.0071,
      "reward": 0.29636096954345703,
      "reward_std": 0.08944160863757133,
      "reward_std_all": 0.10916823521256447,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9009244740009308,
      "rewards/forensics_tempsamp_iou_reward": 0.29636096954345703,
      "step": 194
    },
    {
      "completion_length": 16.375,
      "epoch": 0.5193075898801598,
      "grad_norm": 6.621883869171143,
      "kl": 0.2158203125,
      "learning_rate": 8.706666666666667e-07,
      "loss": 0.0086,
      "reward": 0.16999804973602295,
      "reward_std": 0.0635894313454628,
      "reward_std_all": 0.058634644374251366,
      "rewards-max/forensics_tempsamp_iou_reward": 0.666904553771019,
      "rewards/forensics_tempsamp_iou_reward": 0.16999804973602295,
      "step": 195
    },
    {
      "completion_length": 22.640625,
      "epoch": 0.521970705725699,
      "grad_norm": 5.628118991851807,
      "kl": 0.22998046875,
      "learning_rate": 8.699999999999999e-07,
      "loss": 0.0092,
      "reward": 0.20967252552509308,
      "reward_std": 0.093228068202734,
      "reward_std_all": 0.0887647271156311,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6830564737319946,
      "rewards/forensics_tempsamp_iou_reward": 0.20967252552509308,
      "step": 196
    },
    {
      "completion_length": 17.109375,
      "epoch": 0.5246338215712384,
      "grad_norm": 6.22587776184082,
      "kl": 0.20703125,
      "learning_rate": 8.693333333333333e-07,
      "loss": 0.0083,
      "reward": 0.17233961820602417,
      "reward_std": 0.07129587978124619,
      "reward_std_all": 0.06263910606503487,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9036993086338043,
      "rewards/forensics_tempsamp_iou_reward": 0.17233961820602417,
      "step": 197
    },
    {
      "completion_length": 18.140625,
      "epoch": 0.5272969374167776,
      "grad_norm": 5.202826976776123,
      "kl": 0.17919921875,
      "learning_rate": 8.686666666666666e-07,
      "loss": 0.0072,
      "reward": 0.21566863358020782,
      "reward_std": 0.07132084853947163,
      "reward_std_all": 0.08197768032550812,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6918139755725861,
      "rewards/forensics_tempsamp_iou_reward": 0.21566863358020782,
      "step": 198
    },
    {
      "completion_length": 21.015625,
      "epoch": 0.5299600532623169,
      "grad_norm": 5.625739574432373,
      "kl": 0.1708984375,
      "learning_rate": 8.68e-07,
      "loss": 0.0068,
      "reward": 0.3145056366920471,
      "reward_std": 0.11426941677927971,
      "reward_std_all": 0.12474226206541061,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9065139293670654,
      "rewards/forensics_tempsamp_iou_reward": 0.3145056366920471,
      "step": 199
    },
    {
      "completion_length": 17.625,
      "epoch": 0.5326231691078562,
      "grad_norm": 5.27986478805542,
      "kl": 0.22412109375,
      "learning_rate": 8.673333333333332e-07,
      "loss": 0.009,
      "reward": 0.20549560338258743,
      "reward_std": 0.0769210197031498,
      "reward_std_all": 0.0762711614370346,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6665937006473541,
      "rewards/forensics_tempsamp_iou_reward": 0.20549560338258743,
      "step": 200
    },
    {
      "completion_length": 16.953125,
      "epoch": 0.5352862849533955,
      "grad_norm": 7.020174503326416,
      "kl": 0.1953125,
      "learning_rate": 8.666666666666667e-07,
      "loss": 0.0078,
      "reward": 0.24255141615867615,
      "reward_std": 0.1116170734167099,
      "reward_std_all": 0.11265721172094345,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7855545580387115,
      "rewards/forensics_tempsamp_iou_reward": 0.24255141615867615,
      "step": 201
    },
    {
      "completion_length": 24.84375,
      "epoch": 0.5379494007989347,
      "grad_norm": 5.294422626495361,
      "kl": 0.19287109375,
      "learning_rate": 8.659999999999999e-07,
      "loss": 0.0077,
      "reward": 0.27892911434173584,
      "reward_std": 0.10075248777866364,
      "reward_std_all": 0.11366710066795349,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8973950445652008,
      "rewards/forensics_tempsamp_iou_reward": 0.27892911434173584,
      "step": 202
    },
    {
      "completion_length": 15.296875,
      "epoch": 0.5406125166444741,
      "grad_norm": 5.67017126083374,
      "kl": 0.22998046875,
      "learning_rate": 8.653333333333333e-07,
      "loss": 0.0092,
      "reward": 0.20687996596097946,
      "reward_std": 0.05888273008167744,
      "reward_std_all": 0.06659705936908722,
      "rewards-max/forensics_tempsamp_iou_reward": 0.883495956659317,
      "rewards/forensics_tempsamp_iou_reward": 0.20687996596097946,
      "step": 203
    },
    {
      "completion_length": 16.328125,
      "epoch": 0.5432756324900133,
      "grad_norm": 6.025546073913574,
      "kl": 0.263671875,
      "learning_rate": 8.646666666666667e-07,
      "loss": 0.0105,
      "reward": 0.2357284277677536,
      "reward_std": 0.08913788199424744,
      "reward_std_all": 0.1073385514318943,
      "rewards-max/forensics_tempsamp_iou_reward": 0.82967808842659,
      "rewards/forensics_tempsamp_iou_reward": 0.2357284277677536,
      "step": 204
    },
    {
      "completion_length": 14.078125,
      "epoch": 0.5459387483355526,
      "grad_norm": 6.516989231109619,
      "kl": 0.208984375,
      "learning_rate": 8.639999999999999e-07,
      "loss": 0.0084,
      "reward": 0.3064315617084503,
      "reward_std": 0.10892371833324432,
      "reward_std_all": 0.11667510122060776,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7910343408584595,
      "rewards/forensics_tempsamp_iou_reward": 0.3064315617084503,
      "step": 205
    },
    {
      "completion_length": 22.953125,
      "epoch": 0.5486018641810919,
      "grad_norm": 4.331488132476807,
      "kl": 0.21826171875,
      "learning_rate": 8.633333333333333e-07,
      "loss": 0.0088,
      "reward": 0.2966995984315872,
      "reward_std": 0.1027785949409008,
      "reward_std_all": 0.09159224852919579,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9119084775447845,
      "rewards/forensics_tempsamp_iou_reward": 0.2966995984315872,
      "step": 206
    },
    {
      "completion_length": 13.8125,
      "epoch": 0.5512649800266312,
      "grad_norm": 5.534148693084717,
      "kl": 0.2900390625,
      "learning_rate": 8.626666666666666e-07,
      "loss": 0.0116,
      "reward": 0.2582717165350914,
      "reward_std": 0.05966934561729431,
      "reward_std_all": 0.09204369783401489,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7487786412239075,
      "rewards/forensics_tempsamp_iou_reward": 0.2582717165350914,
      "step": 207
    },
    {
      "completion_length": 14.15625,
      "epoch": 0.5539280958721704,
      "grad_norm": 6.656670570373535,
      "kl": 0.3056640625,
      "learning_rate": 8.62e-07,
      "loss": 0.0122,
      "reward": 0.27542663365602493,
      "reward_std": 0.0843794010579586,
      "reward_std_all": 0.08725646510720253,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7600732743740082,
      "rewards/forensics_tempsamp_iou_reward": 0.27542663365602493,
      "step": 208
    },
    {
      "completion_length": 17.75,
      "epoch": 0.5565912117177098,
      "grad_norm": 4.903630256652832,
      "kl": 0.1962890625,
      "learning_rate": 8.613333333333332e-07,
      "loss": 0.0078,
      "reward": 0.28020618855953217,
      "reward_std": 0.06693053245544434,
      "reward_std_all": 0.08290344476699829,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9131413996219635,
      "rewards/forensics_tempsamp_iou_reward": 0.28020618855953217,
      "step": 209
    },
    {
      "completion_length": 14.9375,
      "epoch": 0.559254327563249,
      "grad_norm": 5.073878765106201,
      "kl": 0.2861328125,
      "learning_rate": 8.606666666666667e-07,
      "loss": 0.0114,
      "reward": 0.21225713193416595,
      "reward_std": 0.102201197296381,
      "reward_std_all": 0.0936679095029831,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8824297189712524,
      "rewards/forensics_tempsamp_iou_reward": 0.21225713193416595,
      "step": 210
    },
    {
      "completion_length": 16.015625,
      "epoch": 0.5619174434087882,
      "grad_norm": 5.584758281707764,
      "kl": 0.1875,
      "learning_rate": 8.599999999999999e-07,
      "loss": 0.0075,
      "reward": 0.2850750908255577,
      "reward_std": 0.0686599463224411,
      "reward_std_all": 0.09491133317351341,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9124277830123901,
      "rewards/forensics_tempsamp_iou_reward": 0.2850750908255577,
      "step": 211
    },
    {
      "completion_length": 13.53125,
      "epoch": 0.5645805592543276,
      "grad_norm": 6.302996635437012,
      "kl": 0.18359375,
      "learning_rate": 8.593333333333333e-07,
      "loss": 0.0073,
      "reward": 0.28088315576314926,
      "reward_std": 0.08710857667028904,
      "reward_std_all": 0.08305247873067856,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7057139575481415,
      "rewards/forensics_tempsamp_iou_reward": 0.28088315576314926,
      "step": 212
    },
    {
      "completion_length": 14.046875,
      "epoch": 0.5672436750998668,
      "grad_norm": 4.3426079750061035,
      "kl": 0.220703125,
      "learning_rate": 8.586666666666666e-07,
      "loss": 0.0088,
      "reward": 0.2938823848962784,
      "reward_std": 0.07254445552825928,
      "reward_std_all": 0.08648204803466797,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8671292960643768,
      "rewards/forensics_tempsamp_iou_reward": 0.2938823848962784,
      "step": 213
    },
    {
      "completion_length": 15.5625,
      "epoch": 0.5699067909454061,
      "grad_norm": 6.221065521240234,
      "kl": 0.25634765625,
      "learning_rate": 8.58e-07,
      "loss": 0.0102,
      "reward": 0.28255678713321686,
      "reward_std": 0.10355524346232414,
      "reward_std_all": 0.10291678830981255,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9184029698371887,
      "rewards/forensics_tempsamp_iou_reward": 0.28255678713321686,
      "step": 214
    },
    {
      "completion_length": 13.9375,
      "epoch": 0.5725699067909454,
      "grad_norm": 5.610901832580566,
      "kl": 0.3134765625,
      "learning_rate": 8.573333333333332e-07,
      "loss": 0.0126,
      "reward": 0.18071717023849487,
      "reward_std": 0.0636277049779892,
      "reward_std_all": 0.07047808915376663,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9144178032875061,
      "rewards/forensics_tempsamp_iou_reward": 0.18071717023849487,
      "step": 215
    },
    {
      "completion_length": 22.515625,
      "epoch": 0.5752330226364847,
      "grad_norm": 6.058388710021973,
      "kl": 0.25048828125,
      "learning_rate": 8.566666666666667e-07,
      "loss": 0.01,
      "reward": 0.3289477825164795,
      "reward_std": 0.126266211271286,
      "reward_std_all": 0.11021728441119194,
      "rewards-max/forensics_tempsamp_iou_reward": 0.946635365486145,
      "rewards/forensics_tempsamp_iou_reward": 0.3289477825164795,
      "step": 216
    },
    {
      "completion_length": 15.78125,
      "epoch": 0.5778961384820239,
      "grad_norm": 6.586050033569336,
      "kl": 0.5166015625,
      "learning_rate": 8.559999999999999e-07,
      "loss": 0.0206,
      "reward": 0.21737436950206757,
      "reward_std": 0.06294161081314087,
      "reward_std_all": 0.06005456671118736,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6023444980382919,
      "rewards/forensics_tempsamp_iou_reward": 0.21737436950206757,
      "step": 217
    },
    {
      "completion_length": 14.578125,
      "epoch": 0.5805592543275633,
      "grad_norm": 5.375906944274902,
      "kl": 0.296875,
      "learning_rate": 8.553333333333333e-07,
      "loss": 0.0119,
      "reward": 0.228134885430336,
      "reward_std": 0.06556539237499237,
      "reward_std_all": 0.06217450834810734,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9143322110176086,
      "rewards/forensics_tempsamp_iou_reward": 0.228134885430336,
      "step": 218
    },
    {
      "completion_length": 15.75,
      "epoch": 0.5832223701731025,
      "grad_norm": 6.598174095153809,
      "kl": 0.3017578125,
      "learning_rate": 8.546666666666666e-07,
      "loss": 0.0121,
      "reward": 0.30885128676891327,
      "reward_std": 0.0726824402809143,
      "reward_std_all": 0.06846029125154018,
      "rewards-max/forensics_tempsamp_iou_reward": 0.940072625875473,
      "rewards/forensics_tempsamp_iou_reward": 0.30885128676891327,
      "step": 219
    },
    {
      "completion_length": 14.859375,
      "epoch": 0.5858854860186418,
      "grad_norm": 6.936506271362305,
      "kl": 0.22412109375,
      "learning_rate": 8.539999999999999e-07,
      "loss": 0.009,
      "reward": 0.284764900803566,
      "reward_std": 0.06482433341443539,
      "reward_std_all": 0.05908765830099583,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9218091368675232,
      "rewards/forensics_tempsamp_iou_reward": 0.284764900803566,
      "step": 220
    },
    {
      "completion_length": 17.359375,
      "epoch": 0.5885486018641811,
      "grad_norm": 4.828184604644775,
      "kl": 0.2265625,
      "learning_rate": 8.533333333333334e-07,
      "loss": 0.0091,
      "reward": 0.18704497069120407,
      "reward_std": 0.04237132705748081,
      "reward_std_all": 0.04217514768242836,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9089134037494659,
      "rewards/forensics_tempsamp_iou_reward": 0.18704497069120407,
      "step": 221
    },
    {
      "completion_length": 15.421875,
      "epoch": 0.5912117177097204,
      "grad_norm": 4.7501678466796875,
      "kl": 0.23779296875,
      "learning_rate": 8.526666666666666e-07,
      "loss": 0.0095,
      "reward": 0.41107794642448425,
      "reward_std": 0.07097817026078701,
      "reward_std_all": 0.09547404199838638,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9868629276752472,
      "rewards/forensics_tempsamp_iou_reward": 0.41107794642448425,
      "step": 222
    },
    {
      "completion_length": 21.03125,
      "epoch": 0.5938748335552596,
      "grad_norm": 4.515717506408691,
      "kl": 0.2431640625,
      "learning_rate": 8.52e-07,
      "loss": 0.0097,
      "reward": 0.2465137392282486,
      "reward_std": 0.07641318254172802,
      "reward_std_all": 0.07606201246380806,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9346648752689362,
      "rewards/forensics_tempsamp_iou_reward": 0.2465137392282486,
      "step": 223
    },
    {
      "completion_length": 16.1875,
      "epoch": 0.596537949400799,
      "grad_norm": 5.486954212188721,
      "kl": 0.23779296875,
      "learning_rate": 8.513333333333333e-07,
      "loss": 0.0095,
      "reward": 0.23266229778528214,
      "reward_std": 0.05193965509533882,
      "reward_std_all": 0.05346981808543205,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6265874356031418,
      "rewards/forensics_tempsamp_iou_reward": 0.23266229778528214,
      "step": 224
    },
    {
      "completion_length": 16.34375,
      "epoch": 0.5992010652463382,
      "grad_norm": 6.727500915527344,
      "kl": 0.310546875,
      "learning_rate": 8.506666666666667e-07,
      "loss": 0.0124,
      "reward": 0.2901010513305664,
      "reward_std": 0.09291478991508484,
      "reward_std_all": 0.0847269557416439,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9301542341709137,
      "rewards/forensics_tempsamp_iou_reward": 0.2901010513305664,
      "step": 225
    },
    {
      "completion_length": 20.15625,
      "epoch": 0.6018641810918774,
      "grad_norm": 5.766709804534912,
      "kl": 0.298828125,
      "learning_rate": 8.499999999999999e-07,
      "loss": 0.0119,
      "reward": 0.24528764188289642,
      "reward_std": 0.0747162252664566,
      "reward_std_all": 0.07703451626002789,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7172967046499252,
      "rewards/forensics_tempsamp_iou_reward": 0.24528764188289642,
      "step": 226
    },
    {
      "completion_length": 17.921875,
      "epoch": 0.6045272969374168,
      "grad_norm": 7.1932597160339355,
      "kl": 0.529296875,
      "learning_rate": 8.493333333333334e-07,
      "loss": 0.0211,
      "reward": 0.2228715419769287,
      "reward_std": 0.05944175459444523,
      "reward_std_all": 0.06433452665805817,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6497222185134888,
      "rewards/forensics_tempsamp_iou_reward": 0.2228715419769287,
      "step": 227
    },
    {
      "completion_length": 20.5,
      "epoch": 0.607190412782956,
      "grad_norm": 6.671161651611328,
      "kl": 0.25927734375,
      "learning_rate": 8.486666666666666e-07,
      "loss": 0.0104,
      "reward": 0.2442607432603836,
      "reward_std": 0.07178526930510998,
      "reward_std_all": 0.07671093009412289,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9842634201049805,
      "rewards/forensics_tempsamp_iou_reward": 0.2442607432603836,
      "step": 228
    },
    {
      "completion_length": 15.515625,
      "epoch": 0.6098535286284953,
      "grad_norm": 6.33613395690918,
      "kl": 0.4052734375,
      "learning_rate": 8.48e-07,
      "loss": 0.0162,
      "reward": 0.3425375074148178,
      "reward_std": 0.0899670422077179,
      "reward_std_all": 0.08542241901159286,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8876906335353851,
      "rewards/forensics_tempsamp_iou_reward": 0.3425375074148178,
      "step": 229
    },
    {
      "completion_length": 18.21875,
      "epoch": 0.6125166444740346,
      "grad_norm": 5.107021808624268,
      "kl": 0.4765625,
      "learning_rate": 8.473333333333333e-07,
      "loss": 0.019,
      "reward": 0.14456413686275482,
      "reward_std": 0.06107044033706188,
      "reward_std_all": 0.06376951187849045,
      "rewards-max/forensics_tempsamp_iou_reward": 0.36798490583896637,
      "rewards/forensics_tempsamp_iou_reward": 0.14456413686275482,
      "step": 230
    },
    {
      "completion_length": 16.453125,
      "epoch": 0.6151797603195739,
      "grad_norm": 5.9220781326293945,
      "kl": 0.5078125,
      "learning_rate": 8.466666666666667e-07,
      "loss": 0.0203,
      "reward": 0.24229218810796738,
      "reward_std": 0.0675327479839325,
      "reward_std_all": 0.09343018382787704,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7960526347160339,
      "rewards/forensics_tempsamp_iou_reward": 0.24229218810796738,
      "step": 231
    },
    {
      "completion_length": 17.453125,
      "epoch": 0.6178428761651131,
      "grad_norm": 5.150888442993164,
      "kl": 0.3984375,
      "learning_rate": 8.459999999999999e-07,
      "loss": 0.0159,
      "reward": 0.16515087336301804,
      "reward_std": 0.0668003186583519,
      "reward_std_all": 0.06358835846185684,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6468749940395355,
      "rewards/forensics_tempsamp_iou_reward": 0.16515087336301804,
      "step": 232
    },
    {
      "completion_length": 20.953125,
      "epoch": 0.6205059920106525,
      "grad_norm": 5.5357561111450195,
      "kl": 0.3359375,
      "learning_rate": 8.453333333333334e-07,
      "loss": 0.0135,
      "reward": 0.1842908188700676,
      "reward_std": 0.06263552606105804,
      "reward_std_all": 0.0740828737616539,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9367021322250366,
      "rewards/forensics_tempsamp_iou_reward": 0.1842908188700676,
      "step": 233
    },
    {
      "completion_length": 19.078125,
      "epoch": 0.6231691078561917,
      "grad_norm": 6.041022777557373,
      "kl": 0.361328125,
      "learning_rate": 8.446666666666666e-07,
      "loss": 0.0145,
      "reward": 0.26881321519613266,
      "reward_std": 0.10862521827220917,
      "reward_std_all": 0.11276963353157043,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7609953582286835,
      "rewards/forensics_tempsamp_iou_reward": 0.26881321519613266,
      "step": 234
    },
    {
      "completion_length": 18.765625,
      "epoch": 0.625832223701731,
      "grad_norm": 4.898859024047852,
      "kl": 0.3857421875,
      "learning_rate": 8.439999999999999e-07,
      "loss": 0.0155,
      "reward": 0.19057319313287735,
      "reward_std": 0.08702382817864418,
      "reward_std_all": 0.09251483902335167,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6383357048034668,
      "rewards/forensics_tempsamp_iou_reward": 0.19057319313287735,
      "step": 235
    },
    {
      "completion_length": 14.765625,
      "epoch": 0.6284953395472703,
      "grad_norm": 6.147519111633301,
      "kl": 0.3134765625,
      "learning_rate": 8.433333333333333e-07,
      "loss": 0.0125,
      "reward": 0.184956893324852,
      "reward_std": 0.10410400852560997,
      "reward_std_all": 0.10262122005224228,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6627661883831024,
      "rewards/forensics_tempsamp_iou_reward": 0.184956893324852,
      "step": 236
    },
    {
      "completion_length": 15.859375,
      "epoch": 0.6311584553928096,
      "grad_norm": 6.338924884796143,
      "kl": 0.5439453125,
      "learning_rate": 8.426666666666666e-07,
      "loss": 0.0217,
      "reward": 0.23907510936260223,
      "reward_std": 0.07854985073208809,
      "reward_std_all": 0.07982882484793663,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5974553227424622,
      "rewards/forensics_tempsamp_iou_reward": 0.23907510936260223,
      "step": 237
    },
    {
      "completion_length": 25.90625,
      "epoch": 0.6338215712383488,
      "grad_norm": 5.438832759857178,
      "kl": 0.451171875,
      "learning_rate": 8.419999999999999e-07,
      "loss": 0.0181,
      "reward": 0.15146180987358093,
      "reward_std": 0.07275760918855667,
      "reward_std_all": 0.0731690563261509,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4546099379658699,
      "rewards/forensics_tempsamp_iou_reward": 0.15146180987358093,
      "step": 238
    },
    {
      "completion_length": 16.234375,
      "epoch": 0.6364846870838882,
      "grad_norm": 6.592976093292236,
      "kl": 0.388671875,
      "learning_rate": 8.413333333333333e-07,
      "loss": 0.0156,
      "reward": 0.16853222995996475,
      "reward_std": 0.09540041536092758,
      "reward_std_all": 0.08797146379947662,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7366880178451538,
      "rewards/forensics_tempsamp_iou_reward": 0.16853222995996475,
      "step": 239
    },
    {
      "completion_length": 19.875,
      "epoch": 0.6391478029294274,
      "grad_norm": 6.471376419067383,
      "kl": 0.3544921875,
      "learning_rate": 8.406666666666667e-07,
      "loss": 0.0142,
      "reward": 0.22676845639944077,
      "reward_std": 0.11031627282500267,
      "reward_std_all": 0.10253492370247841,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7435488104820251,
      "rewards/forensics_tempsamp_iou_reward": 0.22676845639944077,
      "step": 240
    },
    {
      "completion_length": 20.890625,
      "epoch": 0.6418109187749668,
      "grad_norm": 4.584136009216309,
      "kl": 0.427734375,
      "learning_rate": 8.399999999999999e-07,
      "loss": 0.0171,
      "reward": 0.19232133775949478,
      "reward_std": 0.10454897582530975,
      "reward_std_all": 0.11726230382919312,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5390368700027466,
      "rewards/forensics_tempsamp_iou_reward": 0.19232133775949478,
      "step": 241
    },
    {
      "completion_length": 22.3125,
      "epoch": 0.644474034620506,
      "grad_norm": 5.699325084686279,
      "kl": 0.5595703125,
      "learning_rate": 8.393333333333334e-07,
      "loss": 0.0224,
      "reward": 0.23421868681907654,
      "reward_std": 0.12765461206436157,
      "reward_std_all": 0.12010926008224487,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9352112114429474,
      "rewards/forensics_tempsamp_iou_reward": 0.23421868681907654,
      "step": 242
    },
    {
      "completion_length": 16.953125,
      "epoch": 0.6471371504660453,
      "grad_norm": 7.534374713897705,
      "kl": 0.5068359375,
      "learning_rate": 8.386666666666666e-07,
      "loss": 0.0202,
      "reward": 0.24378079921007156,
      "reward_std": 0.08310146257281303,
      "reward_std_all": 0.08760624378919601,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8327812850475311,
      "rewards/forensics_tempsamp_iou_reward": 0.24378079921007156,
      "step": 243
    },
    {
      "completion_length": 19.84375,
      "epoch": 0.6498002663115846,
      "grad_norm": 6.817309379577637,
      "kl": 0.32421875,
      "learning_rate": 8.38e-07,
      "loss": 0.013,
      "reward": 0.15303073450922966,
      "reward_std": 0.0810882393270731,
      "reward_std_all": 0.08146847411990166,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5621936321258545,
      "rewards/forensics_tempsamp_iou_reward": 0.15303073450922966,
      "step": 244
    },
    {
      "completion_length": 19.609375,
      "epoch": 0.6524633821571239,
      "grad_norm": 6.246949195861816,
      "kl": 0.3095703125,
      "learning_rate": 8.373333333333333e-07,
      "loss": 0.0124,
      "reward": 0.17179231345653534,
      "reward_std": 0.05485768988728523,
      "reward_std_all": 0.07126571238040924,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6345575451850891,
      "rewards/forensics_tempsamp_iou_reward": 0.17179231345653534,
      "step": 245
    },
    {
      "completion_length": 17.296875,
      "epoch": 0.6551264980026631,
      "grad_norm": 4.913091659545898,
      "kl": 0.373046875,
      "learning_rate": 8.366666666666667e-07,
      "loss": 0.015,
      "reward": 0.17356783896684647,
      "reward_std": 0.10550040006637573,
      "reward_std_all": 0.09644215181469917,
      "rewards-max/forensics_tempsamp_iou_reward": 0.889902800321579,
      "rewards/forensics_tempsamp_iou_reward": 0.17356783896684647,
      "step": 246
    },
    {
      "completion_length": 18.59375,
      "epoch": 0.6577896138482024,
      "grad_norm": 5.3997907638549805,
      "kl": 0.435546875,
      "learning_rate": 8.359999999999999e-07,
      "loss": 0.0174,
      "reward": 0.23801323771476746,
      "reward_std": 0.1441967785358429,
      "reward_std_all": 0.12623220309615135,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7714545726776123,
      "rewards/forensics_tempsamp_iou_reward": 0.23801323771476746,
      "step": 247
    },
    {
      "completion_length": 16.890625,
      "epoch": 0.6604527296937417,
      "grad_norm": 5.630625247955322,
      "kl": 0.353515625,
      "learning_rate": 8.353333333333334e-07,
      "loss": 0.0142,
      "reward": 0.21593325585126877,
      "reward_std": 0.09190655127167702,
      "reward_std_all": 0.08540750667452812,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9069575667381287,
      "rewards/forensics_tempsamp_iou_reward": 0.21593325585126877,
      "step": 248
    },
    {
      "completion_length": 21.3125,
      "epoch": 0.6631158455392809,
      "grad_norm": 5.394142150878906,
      "kl": 0.482421875,
      "learning_rate": 8.346666666666666e-07,
      "loss": 0.0193,
      "reward": 0.15762213245034218,
      "reward_std": 0.08528949320316315,
      "reward_std_all": 0.08981345593929291,
      "rewards-max/forensics_tempsamp_iou_reward": 0.47348177433013916,
      "rewards/forensics_tempsamp_iou_reward": 0.15762213245034218,
      "step": 249
    },
    {
      "completion_length": 14.09375,
      "epoch": 0.6657789613848203,
      "grad_norm": 5.693565845489502,
      "kl": 0.435546875,
      "learning_rate": 8.34e-07,
      "loss": 0.0175,
      "reward": 0.23242084681987762,
      "reward_std": 0.08001421764492989,
      "reward_std_all": 0.07987945154309273,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8658808469772339,
      "rewards/forensics_tempsamp_iou_reward": 0.23242084681987762,
      "step": 250
    },
    {
      "completion_length": 23.546875,
      "epoch": 0.6684420772303595,
      "grad_norm": 5.372890472412109,
      "kl": 0.2705078125,
      "learning_rate": 8.333333333333333e-07,
      "loss": 0.0108,
      "reward": 0.19710763543844223,
      "reward_std": 0.07738799601793289,
      "reward_std_all": 0.07786525599658489,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5855653285980225,
      "rewards/forensics_tempsamp_iou_reward": 0.19710763543844223,
      "step": 251
    },
    {
      "completion_length": 15.140625,
      "epoch": 0.6711051930758988,
      "grad_norm": 5.461293697357178,
      "kl": 0.23583984375,
      "learning_rate": 8.326666666666666e-07,
      "loss": 0.0094,
      "reward": 0.1837805211544037,
      "reward_std": 0.059296172112226486,
      "reward_std_all": 0.060920340940356255,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6299461424350739,
      "rewards/forensics_tempsamp_iou_reward": 0.1837805211544037,
      "step": 252
    },
    {
      "completion_length": 16.0,
      "epoch": 0.6737683089214381,
      "grad_norm": 5.729190349578857,
      "kl": 0.291015625,
      "learning_rate": 8.319999999999999e-07,
      "loss": 0.0116,
      "reward": 0.2666804939508438,
      "reward_std": 0.12328039854764938,
      "reward_std_all": 0.11561546474695206,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9676190614700317,
      "rewards/forensics_tempsamp_iou_reward": 0.2666804939508438,
      "step": 253
    },
    {
      "completion_length": 18.453125,
      "epoch": 0.6764314247669774,
      "grad_norm": 5.560762405395508,
      "kl": 0.2998046875,
      "learning_rate": 8.313333333333333e-07,
      "loss": 0.012,
      "reward": 0.2863418757915497,
      "reward_std": 0.10539811849594116,
      "reward_std_all": 0.11680380254983902,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7635583579540253,
      "rewards/forensics_tempsamp_iou_reward": 0.2863418757915497,
      "step": 254
    },
    {
      "completion_length": 19.71875,
      "epoch": 0.6790945406125166,
      "grad_norm": 6.079843044281006,
      "kl": 0.42578125,
      "learning_rate": 8.306666666666666e-07,
      "loss": 0.017,
      "reward": 0.20194246619939804,
      "reward_std": 0.05701414495706558,
      "reward_std_all": 0.07557150721549988,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7438720762729645,
      "rewards/forensics_tempsamp_iou_reward": 0.20194246619939804,
      "step": 255
    },
    {
      "completion_length": 15.421875,
      "epoch": 0.681757656458056,
      "grad_norm": 5.66619348526001,
      "kl": 0.19091796875,
      "learning_rate": 8.299999999999999e-07,
      "loss": 0.0076,
      "reward": 0.3171147555112839,
      "reward_std": 0.08256105333566666,
      "reward_std_all": 0.0956258475780487,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8863577842712402,
      "rewards/forensics_tempsamp_iou_reward": 0.3171147555112839,
      "step": 256
    },
    {
      "completion_length": 15.453125,
      "epoch": 0.6844207723035952,
      "grad_norm": 5.1022868156433105,
      "kl": 0.416015625,
      "learning_rate": 8.293333333333333e-07,
      "loss": 0.0166,
      "reward": 0.2932962253689766,
      "reward_std": 0.12154075130820274,
      "reward_std_all": 0.12852854281663895,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8676894009113312,
      "rewards/forensics_tempsamp_iou_reward": 0.2932962253689766,
      "step": 257
    },
    {
      "completion_length": 17.8125,
      "epoch": 0.6870838881491345,
      "grad_norm": 4.3318095207214355,
      "kl": 0.318359375,
      "learning_rate": 8.286666666666666e-07,
      "loss": 0.0127,
      "reward": 0.2827015817165375,
      "reward_std": 0.09634725749492645,
      "reward_std_all": 0.10332279279828072,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7924815714359283,
      "rewards/forensics_tempsamp_iou_reward": 0.2827015817165375,
      "step": 258
    },
    {
      "completion_length": 13.890625,
      "epoch": 0.6897470039946738,
      "grad_norm": 4.565997123718262,
      "kl": 0.2451171875,
      "learning_rate": 8.28e-07,
      "loss": 0.0098,
      "reward": 0.23599692434072495,
      "reward_std": 0.08943525329232216,
      "reward_std_all": 0.08728991076350212,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7208333313465118,
      "rewards/forensics_tempsamp_iou_reward": 0.23599692434072495,
      "step": 259
    },
    {
      "completion_length": 24.75,
      "epoch": 0.6924101198402131,
      "grad_norm": 5.369793891906738,
      "kl": 0.26904296875,
      "learning_rate": 8.273333333333333e-07,
      "loss": 0.0108,
      "reward": 0.18553848564624786,
      "reward_std": 0.08066588267683983,
      "reward_std_all": 0.09812664240598679,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6605835556983948,
      "rewards/forensics_tempsamp_iou_reward": 0.18553848564624786,
      "step": 260
    },
    {
      "completion_length": 15.375,
      "epoch": 0.6950732356857523,
      "grad_norm": 5.212441921234131,
      "kl": 0.2646484375,
      "learning_rate": 8.266666666666667e-07,
      "loss": 0.0105,
      "reward": 0.27703340351581573,
      "reward_std": 0.08880712650716305,
      "reward_std_all": 0.09487670287489891,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7384301126003265,
      "rewards/forensics_tempsamp_iou_reward": 0.27703340351581573,
      "step": 261
    },
    {
      "completion_length": 17.9375,
      "epoch": 0.6977363515312917,
      "grad_norm": 5.137880802154541,
      "kl": 0.3486328125,
      "learning_rate": 8.259999999999999e-07,
      "loss": 0.0139,
      "reward": 0.23823552578687668,
      "reward_std": 0.11012744903564453,
      "reward_std_all": 0.1035943403840065,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8596567213535309,
      "rewards/forensics_tempsamp_iou_reward": 0.23823552578687668,
      "step": 262
    },
    {
      "completion_length": 18.625,
      "epoch": 0.7003994673768309,
      "grad_norm": 4.9309234619140625,
      "kl": 0.2939453125,
      "learning_rate": 8.253333333333334e-07,
      "loss": 0.0117,
      "reward": 0.2022620290517807,
      "reward_std": 0.061853282153606415,
      "reward_std_all": 0.06244021654129028,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8683008849620819,
      "rewards/forensics_tempsamp_iou_reward": 0.2022620290517807,
      "step": 263
    },
    {
      "completion_length": 19.5,
      "epoch": 0.7030625832223701,
      "grad_norm": 5.948638916015625,
      "kl": 0.345703125,
      "learning_rate": 8.246666666666666e-07,
      "loss": 0.0138,
      "reward": 0.16262966394424438,
      "reward_std": 0.09333517402410507,
      "reward_std_all": 0.09010576456785202,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8034535050392151,
      "rewards/forensics_tempsamp_iou_reward": 0.16262966394424438,
      "step": 264
    },
    {
      "completion_length": 27.5,
      "epoch": 0.7057256990679095,
      "grad_norm": 5.80911922454834,
      "kl": 0.3115234375,
      "learning_rate": 8.24e-07,
      "loss": 0.0125,
      "reward": 0.2335468828678131,
      "reward_std": 0.12202131748199463,
      "reward_std_all": 0.11238722503185272,
      "rewards-max/forensics_tempsamp_iou_reward": 0.831223726272583,
      "rewards/forensics_tempsamp_iou_reward": 0.2335468828678131,
      "step": 265
    },
    {
      "completion_length": 26.296875,
      "epoch": 0.7083888149134487,
      "grad_norm": 8.982056617736816,
      "kl": 0.3271484375,
      "learning_rate": 8.233333333333333e-07,
      "loss": 0.0131,
      "reward": 0.18618003278970718,
      "reward_std": 0.06467921659350395,
      "reward_std_all": 0.07472069188952446,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7542109191417694,
      "rewards/forensics_tempsamp_iou_reward": 0.18618003278970718,
      "step": 266
    },
    {
      "completion_length": 24.5,
      "epoch": 0.711051930758988,
      "grad_norm": 12.140727043151855,
      "kl": 0.6611328125,
      "learning_rate": 8.226666666666666e-07,
      "loss": 0.0264,
      "reward": 0.20433645695447922,
      "reward_std": 0.08813458681106567,
      "reward_std_all": 0.08741786330938339,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7308691293001175,
      "rewards/forensics_tempsamp_iou_reward": 0.20433645695447922,
      "step": 267
    },
    {
      "completion_length": 22.1875,
      "epoch": 0.7137150466045273,
      "grad_norm": 5.987174987792969,
      "kl": 0.314453125,
      "learning_rate": 8.219999999999999e-07,
      "loss": 0.0126,
      "reward": 0.3183131515979767,
      "reward_std": 0.1619175225496292,
      "reward_std_all": 0.15597036480903625,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9270541071891785,
      "rewards/forensics_tempsamp_iou_reward": 0.3183131515979767,
      "step": 268
    },
    {
      "completion_length": 29.03125,
      "epoch": 0.7163781624500666,
      "grad_norm": 5.977273941040039,
      "kl": 0.3486328125,
      "learning_rate": 8.213333333333333e-07,
      "loss": 0.0139,
      "reward": 0.21651839837431908,
      "reward_std": 0.10189526528120041,
      "reward_std_all": 0.09813802316784859,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5292493551969528,
      "rewards/forensics_tempsamp_iou_reward": 0.21651839837431908,
      "step": 269
    },
    {
      "completion_length": 21.28125,
      "epoch": 0.7190412782956058,
      "grad_norm": 6.6330718994140625,
      "kl": 0.384765625,
      "learning_rate": 8.206666666666666e-07,
      "loss": 0.0154,
      "reward": 0.336975559592247,
      "reward_std": 0.11604434251785278,
      "reward_std_all": 0.11736851185560226,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8950126469135284,
      "rewards/forensics_tempsamp_iou_reward": 0.336975559592247,
      "step": 270
    },
    {
      "completion_length": 25.78125,
      "epoch": 0.7217043941411452,
      "grad_norm": 7.067047119140625,
      "kl": 0.498046875,
      "learning_rate": 8.199999999999999e-07,
      "loss": 0.02,
      "reward": 0.28372377157211304,
      "reward_std": 0.11048695631325245,
      "reward_std_all": 0.1025288887321949,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7753537595272064,
      "rewards/forensics_tempsamp_iou_reward": 0.28372377157211304,
      "step": 271
    },
    {
      "completion_length": 23.0625,
      "epoch": 0.7243675099866844,
      "grad_norm": 5.735743999481201,
      "kl": 0.421875,
      "learning_rate": 8.193333333333333e-07,
      "loss": 0.0169,
      "reward": 0.19627348333597183,
      "reward_std": 0.08512846753001213,
      "reward_std_all": 0.0923803336918354,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8203654289245605,
      "rewards/forensics_tempsamp_iou_reward": 0.19627348333597183,
      "step": 272
    },
    {
      "completion_length": 21.21875,
      "epoch": 0.7270306258322237,
      "grad_norm": 6.3170061111450195,
      "kl": 0.3984375,
      "learning_rate": 8.186666666666666e-07,
      "loss": 0.0159,
      "reward": 0.23042401671409607,
      "reward_std": 0.07996688783168793,
      "reward_std_all": 0.08184469863772392,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8821894228458405,
      "rewards/forensics_tempsamp_iou_reward": 0.23042401671409607,
      "step": 273
    },
    {
      "completion_length": 19.125,
      "epoch": 0.729693741677763,
      "grad_norm": 5.058082103729248,
      "kl": 0.3994140625,
      "learning_rate": 8.179999999999999e-07,
      "loss": 0.016,
      "reward": 0.19149735569953918,
      "reward_std": 0.08691009506583214,
      "reward_std_all": 0.0835203155875206,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6976551115512848,
      "rewards/forensics_tempsamp_iou_reward": 0.19149735569953918,
      "step": 274
    },
    {
      "completion_length": 28.953125,
      "epoch": 0.7323568575233023,
      "grad_norm": 6.244419097900391,
      "kl": 0.4111328125,
      "learning_rate": 8.173333333333333e-07,
      "loss": 0.0164,
      "reward": 0.3148021250963211,
      "reward_std": 0.10631891712546349,
      "reward_std_all": 0.13444304466247559,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8705200552940369,
      "rewards/forensics_tempsamp_iou_reward": 0.3148021250963211,
      "step": 275
    },
    {
      "completion_length": 36.109375,
      "epoch": 0.7350199733688415,
      "grad_norm": 6.407504081726074,
      "kl": 0.67578125,
      "learning_rate": 8.166666666666666e-07,
      "loss": 0.0271,
      "reward": 0.2328912317752838,
      "reward_std": 0.09996416419744492,
      "reward_std_all": 0.09538634866476059,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8817743360996246,
      "rewards/forensics_tempsamp_iou_reward": 0.2328912317752838,
      "step": 276
    },
    {
      "completion_length": 28.890625,
      "epoch": 0.7376830892143809,
      "grad_norm": 5.114567279815674,
      "kl": 0.3076171875,
      "learning_rate": 8.159999999999999e-07,
      "loss": 0.0123,
      "reward": 0.25272198766469955,
      "reward_std": 0.13646995276212692,
      "reward_std_all": 0.1360865905880928,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7479651868343353,
      "rewards/forensics_tempsamp_iou_reward": 0.25272198766469955,
      "step": 277
    },
    {
      "completion_length": 16.109375,
      "epoch": 0.7403462050599201,
      "grad_norm": 5.952296733856201,
      "kl": 0.3935546875,
      "learning_rate": 8.153333333333334e-07,
      "loss": 0.0157,
      "reward": 0.24092070013284683,
      "reward_std": 0.07590636238455772,
      "reward_std_all": 0.09549444168806076,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9002061486244202,
      "rewards/forensics_tempsamp_iou_reward": 0.24092070013284683,
      "step": 278
    },
    {
      "completion_length": 22.828125,
      "epoch": 0.7430093209054593,
      "grad_norm": 4.920569896697998,
      "kl": 0.41796875,
      "learning_rate": 8.146666666666666e-07,
      "loss": 0.0167,
      "reward": 0.2423596978187561,
      "reward_std": 0.08498752862215042,
      "reward_std_all": 0.09510862827301025,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8432830572128296,
      "rewards/forensics_tempsamp_iou_reward": 0.2423596978187561,
      "step": 279
    },
    {
      "completion_length": 28.65625,
      "epoch": 0.7456724367509987,
      "grad_norm": 4.442055702209473,
      "kl": 0.4453125,
      "learning_rate": 8.14e-07,
      "loss": 0.0178,
      "reward": 0.19222534447908401,
      "reward_std": 0.0846869982779026,
      "reward_std_all": 0.09226885065436363,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7064569294452667,
      "rewards/forensics_tempsamp_iou_reward": 0.19222534447908401,
      "step": 280
    },
    {
      "completion_length": 19.09375,
      "epoch": 0.748335552596538,
      "grad_norm": 6.823477745056152,
      "kl": 0.4619140625,
      "learning_rate": 8.133333333333333e-07,
      "loss": 0.0185,
      "reward": 0.2584715485572815,
      "reward_std": 0.10023472085595131,
      "reward_std_all": 0.114248126745224,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8605392277240753,
      "rewards/forensics_tempsamp_iou_reward": 0.2584715485572815,
      "step": 281
    },
    {
      "completion_length": 23.25,
      "epoch": 0.7509986684420772,
      "grad_norm": 5.029567241668701,
      "kl": 0.345703125,
      "learning_rate": 8.126666666666666e-07,
      "loss": 0.0139,
      "reward": 0.24489327520132065,
      "reward_std": 0.07311710342764854,
      "reward_std_all": 0.07130157575011253,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8188992440700531,
      "rewards/forensics_tempsamp_iou_reward": 0.24489327520132065,
      "step": 282
    },
    {
      "completion_length": 17.296875,
      "epoch": 0.7536617842876165,
      "grad_norm": 6.828907012939453,
      "kl": 0.53125,
      "learning_rate": 8.12e-07,
      "loss": 0.0213,
      "reward": 0.24813736230134964,
      "reward_std": 0.11914046853780746,
      "reward_std_all": 0.11833576112985611,
      "rewards-max/forensics_tempsamp_iou_reward": 0.92030069231987,
      "rewards/forensics_tempsamp_iou_reward": 0.24813736230134964,
      "step": 283
    },
    {
      "completion_length": 18.375,
      "epoch": 0.7563249001331558,
      "grad_norm": 7.762462615966797,
      "kl": 0.4462890625,
      "learning_rate": 8.113333333333333e-07,
      "loss": 0.0178,
      "reward": 0.24796844273805618,
      "reward_std": 0.10424487665295601,
      "reward_std_all": 0.11150568351149559,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8425492942333221,
      "rewards/forensics_tempsamp_iou_reward": 0.24796844273805618,
      "step": 284
    },
    {
      "completion_length": 20.015625,
      "epoch": 0.758988015978695,
      "grad_norm": 6.581786632537842,
      "kl": 0.3447265625,
      "learning_rate": 8.106666666666666e-07,
      "loss": 0.0138,
      "reward": 0.18070320039987564,
      "reward_std": 0.07281313836574554,
      "reward_std_all": 0.07578133419156075,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7069067358970642,
      "rewards/forensics_tempsamp_iou_reward": 0.18070320039987564,
      "step": 285
    },
    {
      "completion_length": 15.15625,
      "epoch": 0.7616511318242344,
      "grad_norm": 5.96824836730957,
      "kl": 0.240234375,
      "learning_rate": 8.1e-07,
      "loss": 0.0096,
      "reward": 0.2232055589556694,
      "reward_std": 0.09919590130448341,
      "reward_std_all": 0.0961399283260107,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7909170687198639,
      "rewards/forensics_tempsamp_iou_reward": 0.2232055589556694,
      "step": 286
    },
    {
      "completion_length": 19.5625,
      "epoch": 0.7643142476697736,
      "grad_norm": 6.411684513092041,
      "kl": 0.35498046875,
      "learning_rate": 8.093333333333333e-07,
      "loss": 0.0142,
      "reward": 0.20947803556919098,
      "reward_std": 0.10001286119222641,
      "reward_std_all": 0.09679477289319038,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8474116325378418,
      "rewards/forensics_tempsamp_iou_reward": 0.20947803556919098,
      "step": 287
    },
    {
      "completion_length": 16.734375,
      "epoch": 0.7669773635153129,
      "grad_norm": 6.288114547729492,
      "kl": 0.3173828125,
      "learning_rate": 8.086666666666666e-07,
      "loss": 0.0127,
      "reward": 0.35127604752779007,
      "reward_std": 0.09933291375637054,
      "reward_std_all": 0.12089956551790237,
      "rewards-max/forensics_tempsamp_iou_reward": 0.916898250579834,
      "rewards/forensics_tempsamp_iou_reward": 0.35127604752779007,
      "step": 288
    },
    {
      "completion_length": 15.78125,
      "epoch": 0.7696404793608522,
      "grad_norm": 4.422727108001709,
      "kl": 0.40234375,
      "learning_rate": 8.08e-07,
      "loss": 0.0161,
      "reward": 0.2281946837902069,
      "reward_std": 0.10146572813391685,
      "reward_std_all": 0.10231265798211098,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8652393221855164,
      "rewards/forensics_tempsamp_iou_reward": 0.2281946837902069,
      "step": 289
    },
    {
      "completion_length": 14.15625,
      "epoch": 0.7723035952063915,
      "grad_norm": 6.21001672744751,
      "kl": 0.296875,
      "learning_rate": 8.073333333333333e-07,
      "loss": 0.0119,
      "reward": 0.26245006918907166,
      "reward_std": 0.08131571114063263,
      "reward_std_all": 0.10311927646398544,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6053189039230347,
      "rewards/forensics_tempsamp_iou_reward": 0.26245006918907166,
      "step": 290
    },
    {
      "completion_length": 17.953125,
      "epoch": 0.7749667110519307,
      "grad_norm": 5.347382545471191,
      "kl": 0.25048828125,
      "learning_rate": 8.066666666666666e-07,
      "loss": 0.01,
      "reward": 0.24393214285373688,
      "reward_std": 0.05261905491352081,
      "reward_std_all": 0.05745208263397217,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7526552379131317,
      "rewards/forensics_tempsamp_iou_reward": 0.24393214285373688,
      "step": 291
    },
    {
      "completion_length": 18.78125,
      "epoch": 0.7776298268974701,
      "grad_norm": 4.668868541717529,
      "kl": 0.2587890625,
      "learning_rate": 8.06e-07,
      "loss": 0.0103,
      "reward": 0.38591861724853516,
      "reward_std": 0.09879416227340698,
      "reward_std_all": 0.09723090007901192,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9559392631053925,
      "rewards/forensics_tempsamp_iou_reward": 0.38591861724853516,
      "step": 292
    },
    {
      "completion_length": 13.453125,
      "epoch": 0.7802929427430093,
      "grad_norm": 6.900239944458008,
      "kl": 0.23388671875,
      "learning_rate": 8.053333333333333e-07,
      "loss": 0.0093,
      "reward": 0.2655256688594818,
      "reward_std": 0.10203680396080017,
      "reward_std_all": 0.09931985661387444,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7975131571292877,
      "rewards/forensics_tempsamp_iou_reward": 0.2655256688594818,
      "step": 293
    },
    {
      "completion_length": 14.546875,
      "epoch": 0.7829560585885486,
      "grad_norm": 4.546805381774902,
      "kl": 0.2890625,
      "learning_rate": 8.046666666666666e-07,
      "loss": 0.0116,
      "reward": 0.2797667980194092,
      "reward_std": 0.10110880807042122,
      "reward_std_all": 0.1039479486644268,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8930982649326324,
      "rewards/forensics_tempsamp_iou_reward": 0.2797667980194092,
      "step": 294
    },
    {
      "completion_length": 15.09375,
      "epoch": 0.7856191744340879,
      "grad_norm": 5.007926940917969,
      "kl": 0.240234375,
      "learning_rate": 8.04e-07,
      "loss": 0.0096,
      "reward": 0.3129418417811394,
      "reward_std": 0.07524896413087845,
      "reward_std_all": 0.07796556875109673,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8200483322143555,
      "rewards/forensics_tempsamp_iou_reward": 0.3129418417811394,
      "step": 295
    },
    {
      "completion_length": 14.359375,
      "epoch": 0.7882822902796272,
      "grad_norm": 5.184530735015869,
      "kl": 0.3349609375,
      "learning_rate": 8.033333333333333e-07,
      "loss": 0.0134,
      "reward": 0.2036619484424591,
      "reward_std": 0.07769115455448627,
      "reward_std_all": 0.09593011438846588,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5487998276948929,
      "rewards/forensics_tempsamp_iou_reward": 0.2036619484424591,
      "step": 296
    },
    {
      "completion_length": 16.296875,
      "epoch": 0.7909454061251664,
      "grad_norm": 5.048938274383545,
      "kl": 0.23779296875,
      "learning_rate": 8.026666666666667e-07,
      "loss": 0.0095,
      "reward": 0.25481634587049484,
      "reward_std": 0.07197731360793114,
      "reward_std_all": 0.0720144547522068,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5794944167137146,
      "rewards/forensics_tempsamp_iou_reward": 0.25481634587049484,
      "step": 297
    },
    {
      "completion_length": 21.640625,
      "epoch": 0.7936085219707057,
      "grad_norm": 5.217719078063965,
      "kl": 0.2431640625,
      "learning_rate": 8.02e-07,
      "loss": 0.0097,
      "reward": 0.18694111704826355,
      "reward_std": 0.0530512984842062,
      "reward_std_all": 0.06482714042067528,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7905123233795166,
      "rewards/forensics_tempsamp_iou_reward": 0.18694111704826355,
      "step": 298
    },
    {
      "completion_length": 17.015625,
      "epoch": 0.796271637816245,
      "grad_norm": 5.938996315002441,
      "kl": 0.18505859375,
      "learning_rate": 8.013333333333333e-07,
      "loss": 0.0074,
      "reward": 0.19563817977905273,
      "reward_std": 0.09881363436579704,
      "reward_std_all": 0.09613197669386864,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8790486454963684,
      "rewards/forensics_tempsamp_iou_reward": 0.19563817977905273,
      "step": 299
    },
    {
      "completion_length": 13.734375,
      "epoch": 0.7989347536617842,
      "grad_norm": 5.199770450592041,
      "kl": 0.2822265625,
      "learning_rate": 8.006666666666666e-07,
      "loss": 0.0113,
      "reward": 0.2699928432703018,
      "reward_std": 0.07066335715353489,
      "reward_std_all": 0.07743894308805466,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7145553231239319,
      "rewards/forensics_tempsamp_iou_reward": 0.2699928432703018,
      "step": 300
    },
    {
      "completion_length": 14.390625,
      "epoch": 0.8015978695073236,
      "grad_norm": 7.806066513061523,
      "kl": 0.26611328125,
      "learning_rate": 8e-07,
      "loss": 0.0106,
      "reward": 0.27411190420389175,
      "reward_std": 0.08282243087887764,
      "reward_std_all": 0.09372072666883469,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9323322176933289,
      "rewards/forensics_tempsamp_iou_reward": 0.27411190420389175,
      "step": 301
    },
    {
      "completion_length": 15.09375,
      "epoch": 0.8042609853528628,
      "grad_norm": 6.421881675720215,
      "kl": 0.3193359375,
      "learning_rate": 7.993333333333333e-07,
      "loss": 0.0128,
      "reward": 0.2118602767586708,
      "reward_std": 0.0933472216129303,
      "reward_std_all": 0.09921249002218246,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6740125715732574,
      "rewards/forensics_tempsamp_iou_reward": 0.2118602767586708,
      "step": 302
    },
    {
      "completion_length": 15.0625,
      "epoch": 0.8069241011984021,
      "grad_norm": 6.055543422698975,
      "kl": 0.244140625,
      "learning_rate": 7.986666666666666e-07,
      "loss": 0.0098,
      "reward": 0.21737873554229736,
      "reward_std": 0.04919558949768543,
      "reward_std_all": 0.05803632736206055,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6620388925075531,
      "rewards/forensics_tempsamp_iou_reward": 0.21737873554229736,
      "step": 303
    },
    {
      "completion_length": 15.25,
      "epoch": 0.8095872170439414,
      "grad_norm": 6.133969783782959,
      "kl": 0.2919921875,
      "learning_rate": 7.98e-07,
      "loss": 0.0117,
      "reward": 0.24775145947933197,
      "reward_std": 0.12411396205425262,
      "reward_std_all": 0.12041625007987022,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8674004077911377,
      "rewards/forensics_tempsamp_iou_reward": 0.24775145947933197,
      "step": 304
    },
    {
      "completion_length": 18.140625,
      "epoch": 0.8122503328894807,
      "grad_norm": 5.138533115386963,
      "kl": 0.298828125,
      "learning_rate": 7.973333333333333e-07,
      "loss": 0.0119,
      "reward": 0.29175233840942383,
      "reward_std": 0.11329793930053711,
      "reward_std_all": 0.13707178086042404,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8594129085540771,
      "rewards/forensics_tempsamp_iou_reward": 0.29175233840942383,
      "step": 305
    },
    {
      "completion_length": 15.5,
      "epoch": 0.8149134487350199,
      "grad_norm": 5.533470153808594,
      "kl": 0.208984375,
      "learning_rate": 7.966666666666666e-07,
      "loss": 0.0084,
      "reward": 0.18967410176992416,
      "reward_std": 0.050200157798826694,
      "reward_std_all": 0.07314534112811089,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6302520930767059,
      "rewards/forensics_tempsamp_iou_reward": 0.18967410176992416,
      "step": 306
    },
    {
      "completion_length": 13.578125,
      "epoch": 0.8175765645805593,
      "grad_norm": 4.805436611175537,
      "kl": 0.20849609375,
      "learning_rate": 7.96e-07,
      "loss": 0.0083,
      "reward": 0.205734521150589,
      "reward_std": 0.08604270219802856,
      "reward_std_all": 0.08586304634809494,
      "rewards-max/forensics_tempsamp_iou_reward": 0.63932004570961,
      "rewards/forensics_tempsamp_iou_reward": 0.205734521150589,
      "step": 307
    },
    {
      "completion_length": 13.5,
      "epoch": 0.8202396804260985,
      "grad_norm": 6.0676350593566895,
      "kl": 0.2470703125,
      "learning_rate": 7.953333333333333e-07,
      "loss": 0.0099,
      "reward": 0.23716650903224945,
      "reward_std": 0.09191245585680008,
      "reward_std_all": 0.08294202014803886,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8334412574768066,
      "rewards/forensics_tempsamp_iou_reward": 0.23716650903224945,
      "step": 308
    },
    {
      "completion_length": 15.46875,
      "epoch": 0.8229027962716379,
      "grad_norm": 5.412919998168945,
      "kl": 0.259765625,
      "learning_rate": 7.946666666666666e-07,
      "loss": 0.0104,
      "reward": 0.2980074882507324,
      "reward_std": 0.0726701021194458,
      "reward_std_all": 0.07126930356025696,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8471498489379883,
      "rewards/forensics_tempsamp_iou_reward": 0.2980074882507324,
      "step": 309
    },
    {
      "completion_length": 13.4375,
      "epoch": 0.8255659121171771,
      "grad_norm": 6.068820476531982,
      "kl": 0.19189453125,
      "learning_rate": 7.94e-07,
      "loss": 0.0077,
      "reward": 0.39867886900901794,
      "reward_std": 0.1208425983786583,
      "reward_std_all": 0.12436040863394737,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8678694665431976,
      "rewards/forensics_tempsamp_iou_reward": 0.39867886900901794,
      "step": 310
    },
    {
      "completion_length": 16.421875,
      "epoch": 0.8282290279627164,
      "grad_norm": 7.386653900146484,
      "kl": 0.390625,
      "learning_rate": 7.933333333333333e-07,
      "loss": 0.0156,
      "reward": 0.27655450999736786,
      "reward_std": 0.09858395159244537,
      "reward_std_all": 0.09102562442421913,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9355503618717194,
      "rewards/forensics_tempsamp_iou_reward": 0.27655450999736786,
      "step": 311
    },
    {
      "completion_length": 14.515625,
      "epoch": 0.8308921438082557,
      "grad_norm": 4.456099033355713,
      "kl": 0.30419921875,
      "learning_rate": 7.926666666666666e-07,
      "loss": 0.0121,
      "reward": 0.1579420194029808,
      "reward_std": 0.06026948243379593,
      "reward_std_all": 0.07169973105192184,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5589285790920258,
      "rewards/forensics_tempsamp_iou_reward": 0.1579420194029808,
      "step": 312
    },
    {
      "completion_length": 16.015625,
      "epoch": 0.833555259653795,
      "grad_norm": 5.8487067222595215,
      "kl": 0.220703125,
      "learning_rate": 7.92e-07,
      "loss": 0.0088,
      "reward": 0.31812168657779694,
      "reward_std": 0.1218041367828846,
      "reward_std_all": 0.12104850262403488,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8590818345546722,
      "rewards/forensics_tempsamp_iou_reward": 0.31812168657779694,
      "step": 313
    },
    {
      "completion_length": 18.71875,
      "epoch": 0.8362183754993342,
      "grad_norm": 5.34653902053833,
      "kl": 0.1474609375,
      "learning_rate": 7.913333333333332e-07,
      "loss": 0.0059,
      "reward": 0.39397379755973816,
      "reward_std": 0.07011303678154945,
      "reward_std_all": 0.06925579160451889,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9633535742759705,
      "rewards/forensics_tempsamp_iou_reward": 0.39397379755973816,
      "step": 314
    },
    {
      "completion_length": 15.28125,
      "epoch": 0.8388814913448736,
      "grad_norm": 5.142402648925781,
      "kl": 0.3603515625,
      "learning_rate": 7.906666666666666e-07,
      "loss": 0.0144,
      "reward": 0.36225171387195587,
      "reward_std": 0.09080557152628899,
      "reward_std_all": 0.11243665590882301,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9060350358486176,
      "rewards/forensics_tempsamp_iou_reward": 0.36225171387195587,
      "step": 315
    },
    {
      "completion_length": 16.140625,
      "epoch": 0.8415446071904128,
      "grad_norm": 5.062387466430664,
      "kl": 0.275390625,
      "learning_rate": 7.9e-07,
      "loss": 0.011,
      "reward": 0.2520717456936836,
      "reward_std": 0.0886566024273634,
      "reward_std_all": 0.10544582456350327,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9187867641448975,
      "rewards/forensics_tempsamp_iou_reward": 0.2520717456936836,
      "step": 316
    },
    {
      "completion_length": 19.265625,
      "epoch": 0.844207723035952,
      "grad_norm": 5.255529403686523,
      "kl": 0.4111328125,
      "learning_rate": 7.893333333333333e-07,
      "loss": 0.0165,
      "reward": 0.27337032556533813,
      "reward_std": 0.10112595930695534,
      "reward_std_all": 0.09539016336202621,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9078980684280396,
      "rewards/forensics_tempsamp_iou_reward": 0.27337032556533813,
      "step": 317
    },
    {
      "completion_length": 14.234375,
      "epoch": 0.8468708388814914,
      "grad_norm": 4.677273273468018,
      "kl": 0.21435546875,
      "learning_rate": 7.886666666666666e-07,
      "loss": 0.0086,
      "reward": 0.22114703059196472,
      "reward_std": 0.054665425792336464,
      "reward_std_all": 0.04842718131840229,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8416733741760254,
      "rewards/forensics_tempsamp_iou_reward": 0.22114703059196472,
      "step": 318
    },
    {
      "completion_length": 14.21875,
      "epoch": 0.8495339547270306,
      "grad_norm": 6.021945953369141,
      "kl": 0.3173828125,
      "learning_rate": 7.88e-07,
      "loss": 0.0127,
      "reward": 0.3248838633298874,
      "reward_std": 0.06814510747790337,
      "reward_std_all": 0.07584729790687561,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9715428948402405,
      "rewards/forensics_tempsamp_iou_reward": 0.3248838633298874,
      "step": 319
    },
    {
      "completion_length": 14.65625,
      "epoch": 0.8521970705725699,
      "grad_norm": 5.406853675842285,
      "kl": 0.22509765625,
      "learning_rate": 7.873333333333333e-07,
      "loss": 0.009,
      "reward": 0.3538464456796646,
      "reward_std": 0.05042704753577709,
      "reward_std_all": 0.05498303286731243,
      "rewards-max/forensics_tempsamp_iou_reward": 0.897177129983902,
      "rewards/forensics_tempsamp_iou_reward": 0.3538464456796646,
      "step": 320
    },
    {
      "completion_length": 18.140625,
      "epoch": 0.8548601864181092,
      "grad_norm": 5.932587623596191,
      "kl": 0.31591796875,
      "learning_rate": 7.866666666666666e-07,
      "loss": 0.0127,
      "reward": 0.3062294200062752,
      "reward_std": 0.06070602498948574,
      "reward_std_all": 0.07135594636201859,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9865354299545288,
      "rewards/forensics_tempsamp_iou_reward": 0.3062294200062752,
      "step": 321
    },
    {
      "completion_length": 18.40625,
      "epoch": 0.8575233022636485,
      "grad_norm": 4.504756450653076,
      "kl": 0.25341796875,
      "learning_rate": 7.86e-07,
      "loss": 0.0101,
      "reward": 0.17769630998373032,
      "reward_std": 0.05454045906662941,
      "reward_std_all": 0.06090093031525612,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6119732558727264,
      "rewards/forensics_tempsamp_iou_reward": 0.17769630998373032,
      "step": 322
    },
    {
      "completion_length": 15.109375,
      "epoch": 0.8601864181091877,
      "grad_norm": 8.661443710327148,
      "kl": 0.7275390625,
      "learning_rate": 7.853333333333333e-07,
      "loss": 0.0291,
      "reward": 0.23100809752941132,
      "reward_std": 0.09375148639082909,
      "reward_std_all": 0.0839100182056427,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7960191369056702,
      "rewards/forensics_tempsamp_iou_reward": 0.23100809752941132,
      "step": 323
    },
    {
      "completion_length": 18.515625,
      "epoch": 0.8628495339547271,
      "grad_norm": 4.59807825088501,
      "kl": 0.32470703125,
      "learning_rate": 7.846666666666666e-07,
      "loss": 0.013,
      "reward": 0.26680028438568115,
      "reward_std": 0.07864688336849213,
      "reward_std_all": 0.08642112836241722,
      "rewards-max/forensics_tempsamp_iou_reward": 0.717435285449028,
      "rewards/forensics_tempsamp_iou_reward": 0.26680028438568115,
      "step": 324
    },
    {
      "completion_length": 18.765625,
      "epoch": 0.8655126498002663,
      "grad_norm": 15.568939208984375,
      "kl": 1.04345703125,
      "learning_rate": 7.84e-07,
      "loss": 0.0418,
      "reward": 0.19699469208717346,
      "reward_std": 0.04822338744997978,
      "reward_std_all": 0.057129811495542526,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5371506065130234,
      "rewards/forensics_tempsamp_iou_reward": 0.19699469208717346,
      "step": 325
    },
    {
      "completion_length": 20.125,
      "epoch": 0.8681757656458056,
      "grad_norm": 5.570540904998779,
      "kl": 0.5048828125,
      "learning_rate": 7.833333333333333e-07,
      "loss": 0.0201,
      "reward": 0.29307983815670013,
      "reward_std": 0.061821069568395615,
      "reward_std_all": 0.06163695827126503,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9758047759532928,
      "rewards/forensics_tempsamp_iou_reward": 0.29307983815670013,
      "step": 326
    },
    {
      "completion_length": 15.375,
      "epoch": 0.8708388814913449,
      "grad_norm": 5.631195068359375,
      "kl": 0.310546875,
      "learning_rate": 7.826666666666666e-07,
      "loss": 0.0124,
      "reward": 0.22978652268648148,
      "reward_std": 0.0743161179125309,
      "reward_std_all": 0.06984785571694374,
      "rewards-max/forensics_tempsamp_iou_reward": 0.820417046546936,
      "rewards/forensics_tempsamp_iou_reward": 0.22978652268648148,
      "step": 327
    },
    {
      "completion_length": 13.578125,
      "epoch": 0.8735019973368842,
      "grad_norm": 5.40777587890625,
      "kl": 0.18359375,
      "learning_rate": 7.82e-07,
      "loss": 0.0074,
      "reward": 0.34251342713832855,
      "reward_std": 0.07768124155700207,
      "reward_std_all": 0.07373279705643654,
      "rewards-max/forensics_tempsamp_iou_reward": 0.966509997844696,
      "rewards/forensics_tempsamp_iou_reward": 0.34251342713832855,
      "step": 328
    },
    {
      "completion_length": 19.890625,
      "epoch": 0.8761651131824234,
      "grad_norm": 4.503154277801514,
      "kl": 0.255859375,
      "learning_rate": 7.813333333333332e-07,
      "loss": 0.0102,
      "reward": 0.2973565012216568,
      "reward_std": 0.0770181268453598,
      "reward_std_all": 0.07952478528022766,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8729592263698578,
      "rewards/forensics_tempsamp_iou_reward": 0.2973565012216568,
      "step": 329
    },
    {
      "completion_length": 14.015625,
      "epoch": 0.8788282290279628,
      "grad_norm": 5.963900566101074,
      "kl": 0.2978515625,
      "learning_rate": 7.806666666666666e-07,
      "loss": 0.0119,
      "reward": 0.1506643444299698,
      "reward_std": 0.07380717992782593,
      "reward_std_all": 0.07295458391308784,
      "rewards-max/forensics_tempsamp_iou_reward": 0.475436806678772,
      "rewards/forensics_tempsamp_iou_reward": 0.1506643444299698,
      "step": 330
    },
    {
      "completion_length": 17.0625,
      "epoch": 0.881491344873502,
      "grad_norm": 4.850175857543945,
      "kl": 0.2919921875,
      "learning_rate": 7.799999999999999e-07,
      "loss": 0.0117,
      "reward": 0.24421419948339462,
      "reward_std": 0.07695640996098518,
      "reward_std_all": 0.0824102833867073,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8585937321186066,
      "rewards/forensics_tempsamp_iou_reward": 0.24421419948339462,
      "step": 331
    },
    {
      "completion_length": 22.265625,
      "epoch": 0.8841544607190412,
      "grad_norm": 6.152480602264404,
      "kl": 0.42041015625,
      "learning_rate": 7.793333333333333e-07,
      "loss": 0.0168,
      "reward": 0.20901837944984436,
      "reward_std": 0.07333713956177235,
      "reward_std_all": 0.0805085115134716,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6960151195526123,
      "rewards/forensics_tempsamp_iou_reward": 0.20901837944984436,
      "step": 332
    },
    {
      "completion_length": 16.375,
      "epoch": 0.8868175765645806,
      "grad_norm": 4.763383865356445,
      "kl": 0.4658203125,
      "learning_rate": 7.786666666666665e-07,
      "loss": 0.0187,
      "reward": 0.20596280694007874,
      "reward_std": 0.09428765997290611,
      "reward_std_all": 0.0868297852575779,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8971067070960999,
      "rewards/forensics_tempsamp_iou_reward": 0.20596280694007874,
      "step": 333
    },
    {
      "completion_length": 15.9375,
      "epoch": 0.8894806924101198,
      "grad_norm": 5.268939018249512,
      "kl": 0.3134765625,
      "learning_rate": 7.78e-07,
      "loss": 0.0125,
      "reward": 0.27664990723133087,
      "reward_std": 0.0884842500090599,
      "reward_std_all": 0.08830033987760544,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7678675353527069,
      "rewards/forensics_tempsamp_iou_reward": 0.27664990723133087,
      "step": 334
    },
    {
      "completion_length": 17.578125,
      "epoch": 0.8921438082556591,
      "grad_norm": 4.248449802398682,
      "kl": 0.3544921875,
      "learning_rate": 7.773333333333333e-07,
      "loss": 0.0142,
      "reward": 0.20321984216570854,
      "reward_std": 0.08267050050199032,
      "reward_std_all": 0.08614538237452507,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5831042230129242,
      "rewards/forensics_tempsamp_iou_reward": 0.20321984216570854,
      "step": 335
    },
    {
      "completion_length": 15.0625,
      "epoch": 0.8948069241011984,
      "grad_norm": 5.178606033325195,
      "kl": 0.27587890625,
      "learning_rate": 7.766666666666666e-07,
      "loss": 0.0111,
      "reward": 0.30449704825878143,
      "reward_std": 0.12579850107431412,
      "reward_std_all": 0.11705315113067627,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8639390766620636,
      "rewards/forensics_tempsamp_iou_reward": 0.30449704825878143,
      "step": 336
    },
    {
      "completion_length": 17.25,
      "epoch": 0.8974700399467377,
      "grad_norm": 4.934381008148193,
      "kl": 0.2783203125,
      "learning_rate": 7.76e-07,
      "loss": 0.0111,
      "reward": 0.3130265772342682,
      "reward_std": 0.10722233727574348,
      "reward_std_all": 0.12872298806905746,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8079063296318054,
      "rewards/forensics_tempsamp_iou_reward": 0.3130265772342682,
      "step": 337
    },
    {
      "completion_length": 17.21875,
      "epoch": 0.9001331557922769,
      "grad_norm": 15.134682655334473,
      "kl": 1.1396484375,
      "learning_rate": 7.753333333333333e-07,
      "loss": 0.0456,
      "reward": 0.23232603073120117,
      "reward_std": 0.09861419349908829,
      "reward_std_all": 0.09379534795880318,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7154481112957001,
      "rewards/forensics_tempsamp_iou_reward": 0.23232603073120117,
      "step": 338
    },
    {
      "completion_length": 16.8125,
      "epoch": 0.9027962716378163,
      "grad_norm": 4.417171955108643,
      "kl": 0.27685546875,
      "learning_rate": 7.746666666666666e-07,
      "loss": 0.0111,
      "reward": 0.3757460117340088,
      "reward_std": 0.14106284081935883,
      "reward_std_all": 0.13243717700242996,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7970173954963684,
      "rewards/forensics_tempsamp_iou_reward": 0.3757460117340088,
      "step": 339
    },
    {
      "completion_length": 16.078125,
      "epoch": 0.9054593874833555,
      "grad_norm": 4.018825531005859,
      "kl": 0.2958984375,
      "learning_rate": 7.74e-07,
      "loss": 0.0118,
      "reward": 0.3110183924436569,
      "reward_std": 0.07239322923123837,
      "reward_std_all": 0.0793159268796444,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9291084706783295,
      "rewards/forensics_tempsamp_iou_reward": 0.3110183924436569,
      "step": 340
    },
    {
      "completion_length": 15.890625,
      "epoch": 0.9081225033288948,
      "grad_norm": 5.152292728424072,
      "kl": 0.3232421875,
      "learning_rate": 7.733333333333333e-07,
      "loss": 0.0129,
      "reward": 0.28491102904081345,
      "reward_std": 0.12101971358060837,
      "reward_std_all": 0.12103008478879929,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9629870057106018,
      "rewards/forensics_tempsamp_iou_reward": 0.28491102904081345,
      "step": 341
    },
    {
      "completion_length": 17.328125,
      "epoch": 0.9107856191744341,
      "grad_norm": 6.1450090408325195,
      "kl": 0.25341796875,
      "learning_rate": 7.726666666666666e-07,
      "loss": 0.0101,
      "reward": 0.29382799565792084,
      "reward_std": 0.10601731017231941,
      "reward_std_all": 0.12205852940678596,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9507769048213959,
      "rewards/forensics_tempsamp_iou_reward": 0.29382799565792084,
      "step": 342
    },
    {
      "completion_length": 14.140625,
      "epoch": 0.9134487350199734,
      "grad_norm": 4.7045207023620605,
      "kl": 0.1640625,
      "learning_rate": 7.72e-07,
      "loss": 0.0066,
      "reward": 0.3218175694346428,
      "reward_std": 0.10332491993904114,
      "reward_std_all": 0.103194959461689,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8762007653713226,
      "rewards/forensics_tempsamp_iou_reward": 0.3218175694346428,
      "step": 343
    },
    {
      "completion_length": 16.6875,
      "epoch": 0.9161118508655126,
      "grad_norm": 5.155881404876709,
      "kl": 0.30029296875,
      "learning_rate": 7.713333333333333e-07,
      "loss": 0.012,
      "reward": 0.2786712944507599,
      "reward_std": 0.09304140880703926,
      "reward_std_all": 0.09256766363978386,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8754976391792297,
      "rewards/forensics_tempsamp_iou_reward": 0.2786712944507599,
      "step": 344
    },
    {
      "completion_length": 15.625,
      "epoch": 0.918774966711052,
      "grad_norm": 4.78438663482666,
      "kl": 0.27001953125,
      "learning_rate": 7.706666666666667e-07,
      "loss": 0.0108,
      "reward": 0.1589168682694435,
      "reward_std": 0.05037318170070648,
      "reward_std_all": 0.057433441281318665,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6203813552856445,
      "rewards/forensics_tempsamp_iou_reward": 0.1589168682694435,
      "step": 345
    },
    {
      "completion_length": 14.1875,
      "epoch": 0.9214380825565912,
      "grad_norm": 5.740540981292725,
      "kl": 0.3193359375,
      "learning_rate": 7.699999999999999e-07,
      "loss": 0.0128,
      "reward": 0.21121810376644135,
      "reward_std": 0.08699994161725044,
      "reward_std_all": 0.08846119791269302,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8445826172828674,
      "rewards/forensics_tempsamp_iou_reward": 0.21121810376644135,
      "step": 346
    },
    {
      "completion_length": 16.625,
      "epoch": 0.9241011984021305,
      "grad_norm": 7.94056510925293,
      "kl": 0.3701171875,
      "learning_rate": 7.693333333333333e-07,
      "loss": 0.0148,
      "reward": 0.27149924635887146,
      "reward_std": 0.06036272272467613,
      "reward_std_all": 0.061187900602817535,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8092073500156403,
      "rewards/forensics_tempsamp_iou_reward": 0.27149924635887146,
      "step": 347
    },
    {
      "completion_length": 13.53125,
      "epoch": 0.9267643142476698,
      "grad_norm": 4.435539245605469,
      "kl": 0.14892578125,
      "learning_rate": 7.686666666666666e-07,
      "loss": 0.006,
      "reward": 0.3762570917606354,
      "reward_std": 0.07479453831911087,
      "reward_std_all": 0.08515274152159691,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9351080358028412,
      "rewards/forensics_tempsamp_iou_reward": 0.3762570917606354,
      "step": 348
    },
    {
      "completion_length": 14.1875,
      "epoch": 0.929427430093209,
      "grad_norm": 4.236568450927734,
      "kl": 0.22607421875,
      "learning_rate": 7.68e-07,
      "loss": 0.009,
      "reward": 0.2949148118495941,
      "reward_std": 0.08651421964168549,
      "reward_std_all": 0.08162147551774979,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9146558046340942,
      "rewards/forensics_tempsamp_iou_reward": 0.2949148118495941,
      "step": 349
    },
    {
      "completion_length": 15.84375,
      "epoch": 0.9320905459387483,
      "grad_norm": 5.429799556732178,
      "kl": 0.28466796875,
      "learning_rate": 7.673333333333332e-07,
      "loss": 0.0114,
      "reward": 0.20702100545167923,
      "reward_std": 0.0780794657766819,
      "reward_std_all": 0.07644238695502281,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7352741062641144,
      "rewards/forensics_tempsamp_iou_reward": 0.20702100545167923,
      "step": 350
    },
    {
      "completion_length": 15.5,
      "epoch": 0.9347536617842876,
      "grad_norm": 4.376647472381592,
      "kl": 0.267578125,
      "learning_rate": 7.666666666666667e-07,
      "loss": 0.0107,
      "reward": 0.21040508896112442,
      "reward_std": 0.10057131201028824,
      "reward_std_all": 0.0946841724216938,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8364402055740356,
      "rewards/forensics_tempsamp_iou_reward": 0.21040508896112442,
      "step": 351
    },
    {
      "completion_length": 14.765625,
      "epoch": 0.9374167776298269,
      "grad_norm": 6.226120948791504,
      "kl": 0.183837890625,
      "learning_rate": 7.66e-07,
      "loss": 0.0073,
      "reward": 0.39989252388477325,
      "reward_std": 0.10340167582035065,
      "reward_std_all": 0.1140248104929924,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9385542273521423,
      "rewards/forensics_tempsamp_iou_reward": 0.39989252388477325,
      "step": 352
    },
    {
      "completion_length": 14.46875,
      "epoch": 0.9400798934753661,
      "grad_norm": 4.674248695373535,
      "kl": 0.2080078125,
      "learning_rate": 7.653333333333333e-07,
      "loss": 0.0083,
      "reward": 0.3275313824415207,
      "reward_std": 0.055261991918087006,
      "reward_std_all": 0.06510879844427109,
      "rewards-max/forensics_tempsamp_iou_reward": 0.817716658115387,
      "rewards/forensics_tempsamp_iou_reward": 0.3275313824415207,
      "step": 353
    },
    {
      "completion_length": 18.625,
      "epoch": 0.9427430093209055,
      "grad_norm": 6.514876842498779,
      "kl": 0.4267578125,
      "learning_rate": 7.646666666666667e-07,
      "loss": 0.017,
      "reward": 0.28905779123306274,
      "reward_std": 0.09399723634123802,
      "reward_std_all": 0.09522635489702225,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9033226370811462,
      "rewards/forensics_tempsamp_iou_reward": 0.28905779123306274,
      "step": 354
    },
    {
      "completion_length": 13.890625,
      "epoch": 0.9454061251664447,
      "grad_norm": 4.820710182189941,
      "kl": 0.24462890625,
      "learning_rate": 7.64e-07,
      "loss": 0.0098,
      "reward": 0.25654934346675873,
      "reward_std": 0.039512744173407555,
      "reward_std_all": 0.044506183825433254,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9021320641040802,
      "rewards/forensics_tempsamp_iou_reward": 0.25654934346675873,
      "step": 355
    },
    {
      "completion_length": 14.53125,
      "epoch": 0.948069241011984,
      "grad_norm": 5.038139343261719,
      "kl": 0.23876953125,
      "learning_rate": 7.633333333333333e-07,
      "loss": 0.0095,
      "reward": 0.20059018582105637,
      "reward_std": 0.050720496103167534,
      "reward_std_all": 0.0496541541069746,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6277167201042175,
      "rewards/forensics_tempsamp_iou_reward": 0.20059018582105637,
      "step": 356
    },
    {
      "completion_length": 15.296875,
      "epoch": 0.9507323568575233,
      "grad_norm": 5.714890003204346,
      "kl": 0.267578125,
      "learning_rate": 7.626666666666667e-07,
      "loss": 0.0107,
      "reward": 0.18229994177818298,
      "reward_std": 0.06800435483455658,
      "reward_std_all": 0.06059256009757519,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8003966510295868,
      "rewards/forensics_tempsamp_iou_reward": 0.18229994177818298,
      "step": 357
    },
    {
      "completion_length": 14.6875,
      "epoch": 0.9533954727030626,
      "grad_norm": 3.4851129055023193,
      "kl": 0.1806640625,
      "learning_rate": 7.62e-07,
      "loss": 0.0072,
      "reward": 0.30000290274620056,
      "reward_std": 0.06134882569313049,
      "reward_std_all": 0.07485849410295486,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9299548864364624,
      "rewards/forensics_tempsamp_iou_reward": 0.30000290274620056,
      "step": 358
    },
    {
      "completion_length": 15.125,
      "epoch": 0.9560585885486018,
      "grad_norm": 4.346646308898926,
      "kl": 0.2275390625,
      "learning_rate": 7.613333333333333e-07,
      "loss": 0.0091,
      "reward": 0.2761727645993233,
      "reward_std": 0.06619591265916824,
      "reward_std_all": 0.06072383187711239,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9603845179080963,
      "rewards/forensics_tempsamp_iou_reward": 0.2761727645993233,
      "step": 359
    },
    {
      "completion_length": 16.453125,
      "epoch": 0.9587217043941412,
      "grad_norm": 6.220542907714844,
      "kl": 0.37548828125,
      "learning_rate": 7.606666666666667e-07,
      "loss": 0.0151,
      "reward": 0.30916687846183777,
      "reward_std": 0.09520691819489002,
      "reward_std_all": 0.11091903410851955,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9576732516288757,
      "rewards/forensics_tempsamp_iou_reward": 0.30916687846183777,
      "step": 360
    },
    {
      "completion_length": 15.140625,
      "epoch": 0.9613848202396804,
      "grad_norm": 7.109319686889648,
      "kl": 0.26171875,
      "learning_rate": 7.599999999999999e-07,
      "loss": 0.0105,
      "reward": 0.3445216864347458,
      "reward_std": 0.07534944266080856,
      "reward_std_all": 0.07364245504140854,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9181082844734192,
      "rewards/forensics_tempsamp_iou_reward": 0.3445216864347458,
      "step": 361
    },
    {
      "completion_length": 15.6875,
      "epoch": 0.9640479360852197,
      "grad_norm": 5.470913887023926,
      "kl": 0.3076171875,
      "learning_rate": 7.593333333333333e-07,
      "loss": 0.0123,
      "reward": 0.32911691069602966,
      "reward_std": 0.10434313490986824,
      "reward_std_all": 0.10648159310221672,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9397128224372864,
      "rewards/forensics_tempsamp_iou_reward": 0.32911691069602966,
      "step": 362
    },
    {
      "completion_length": 16.984375,
      "epoch": 0.966711051930759,
      "grad_norm": 5.132920742034912,
      "kl": 0.330078125,
      "learning_rate": 7.586666666666666e-07,
      "loss": 0.0132,
      "reward": 0.22171539813280106,
      "reward_std": 0.06874233670532703,
      "reward_std_all": 0.06379437074065208,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9138946533203125,
      "rewards/forensics_tempsamp_iou_reward": 0.22171539813280106,
      "step": 363
    },
    {
      "completion_length": 18.328125,
      "epoch": 0.9693741677762983,
      "grad_norm": 5.302683353424072,
      "kl": 0.3427734375,
      "learning_rate": 7.58e-07,
      "loss": 0.0137,
      "reward": 0.23946450650691986,
      "reward_std": 0.07580206170678139,
      "reward_std_all": 0.08034716546535492,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8614583313465118,
      "rewards/forensics_tempsamp_iou_reward": 0.23946450650691986,
      "step": 364
    },
    {
      "completion_length": 22.59375,
      "epoch": 0.9720372836218375,
      "grad_norm": 5.781130790710449,
      "kl": 0.3271484375,
      "learning_rate": 7.573333333333332e-07,
      "loss": 0.0131,
      "reward": 0.2939585745334625,
      "reward_std": 0.08692921698093414,
      "reward_std_all": 0.08436058089137077,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9286842942237854,
      "rewards/forensics_tempsamp_iou_reward": 0.2939585745334625,
      "step": 365
    },
    {
      "completion_length": 15.65625,
      "epoch": 0.9747003994673769,
      "grad_norm": 4.551123142242432,
      "kl": 0.3583984375,
      "learning_rate": 7.566666666666667e-07,
      "loss": 0.0144,
      "reward": 0.2980172485113144,
      "reward_std": 0.07556614466011524,
      "reward_std_all": 0.08398952893912792,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8147860169410706,
      "rewards/forensics_tempsamp_iou_reward": 0.2980172485113144,
      "step": 366
    },
    {
      "completion_length": 22.28125,
      "epoch": 0.9773635153129161,
      "grad_norm": 4.347292900085449,
      "kl": 0.3671875,
      "learning_rate": 7.559999999999999e-07,
      "loss": 0.0147,
      "reward": 0.20507720857858658,
      "reward_std": 0.05894564837217331,
      "reward_std_all": 0.06053859740495682,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8864526152610779,
      "rewards/forensics_tempsamp_iou_reward": 0.20507720857858658,
      "step": 367
    },
    {
      "completion_length": 17.546875,
      "epoch": 0.9800266311584553,
      "grad_norm": 4.859766483306885,
      "kl": 0.568359375,
      "learning_rate": 7.553333333333333e-07,
      "loss": 0.0228,
      "reward": 0.23692423850297928,
      "reward_std": 0.1034441888332367,
      "reward_std_all": 0.0961671993136406,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8615956008434296,
      "rewards/forensics_tempsamp_iou_reward": 0.23692423850297928,
      "step": 368
    },
    {
      "completion_length": 26.171875,
      "epoch": 0.9826897470039947,
      "grad_norm": 4.820156574249268,
      "kl": 0.4365234375,
      "learning_rate": 7.546666666666666e-07,
      "loss": 0.0175,
      "reward": 0.31978311389684677,
      "reward_std": 0.06921537220478058,
      "reward_std_all": 0.08400402590632439,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9383646547794342,
      "rewards/forensics_tempsamp_iou_reward": 0.31978311389684677,
      "step": 369
    },
    {
      "completion_length": 17.453125,
      "epoch": 0.9853528628495339,
      "grad_norm": 5.426544666290283,
      "kl": 0.5068359375,
      "learning_rate": 7.54e-07,
      "loss": 0.0203,
      "reward": 0.2720285505056381,
      "reward_std": 0.0940374806523323,
      "reward_std_all": 0.09532920643687248,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9753668308258057,
      "rewards/forensics_tempsamp_iou_reward": 0.2720285505056381,
      "step": 370
    },
    {
      "completion_length": 30.375,
      "epoch": 0.9880159786950732,
      "grad_norm": 5.778920650482178,
      "kl": 0.388671875,
      "learning_rate": 7.533333333333332e-07,
      "loss": 0.0155,
      "reward": 0.1668183133006096,
      "reward_std": 0.058526722714304924,
      "reward_std_all": 0.05657104775309563,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7020424902439117,
      "rewards/forensics_tempsamp_iou_reward": 0.1668183133006096,
      "step": 371
    },
    {
      "completion_length": 17.953125,
      "epoch": 0.9906790945406125,
      "grad_norm": 6.2955474853515625,
      "kl": 0.3984375,
      "learning_rate": 7.526666666666667e-07,
      "loss": 0.016,
      "reward": 0.1614287979900837,
      "reward_std": 0.055133349262177944,
      "reward_std_all": 0.05985165946185589,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6551910638809204,
      "rewards/forensics_tempsamp_iou_reward": 0.1614287979900837,
      "step": 372
    },
    {
      "completion_length": 19.890625,
      "epoch": 0.9933422103861518,
      "grad_norm": 5.133004665374756,
      "kl": 0.4013671875,
      "learning_rate": 7.52e-07,
      "loss": 0.016,
      "reward": 0.3400300294160843,
      "reward_std": 0.0957484245300293,
      "reward_std_all": 0.11171986535191536,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9507751762866974,
      "rewards/forensics_tempsamp_iou_reward": 0.3400300294160843,
      "step": 373
    },
    {
      "completion_length": 21.515625,
      "epoch": 0.996005326231691,
      "grad_norm": 5.443723201751709,
      "kl": 0.3203125,
      "learning_rate": 7.513333333333333e-07,
      "loss": 0.0128,
      "reward": 0.34191375970840454,
      "reward_std": 0.11761291325092316,
      "reward_std_all": 0.14818166196346283,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7984616756439209,
      "rewards/forensics_tempsamp_iou_reward": 0.34191375970840454,
      "step": 374
    },
    {
      "completion_length": 23.84375,
      "epoch": 0.9986684420772304,
      "grad_norm": 5.65633487701416,
      "kl": 0.4404296875,
      "learning_rate": 7.506666666666667e-07,
      "loss": 0.0177,
      "reward": 0.22069154679775238,
      "reward_std": 0.13281233608722687,
      "reward_std_all": 0.12908882647752762,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6429133862257004,
      "rewards/forensics_tempsamp_iou_reward": 0.22069154679775238,
      "step": 375
    },
    {
      "completion_length": 13.0,
      "epoch": 1.0,
      "grad_norm": 4.093985557556152,
      "kl": 0.265625,
      "learning_rate": 7.5e-07,
      "loss": 0.007,
      "reward": 0.37959182262420654,
      "reward_std": 0.2010517567396164,
      "reward_std_all": 0.17090517282485962,
      "rewards-max/forensics_tempsamp_iou_reward": 0.37959182262420654,
      "rewards/forensics_tempsamp_iou_reward": 0.37959182262420654,
      "step": 376
    },
    {
      "completion_length": 16.734375,
      "epoch": 1.0026631158455392,
      "grad_norm": 5.85721492767334,
      "kl": 0.31640625,
      "learning_rate": 7.493333333333333e-07,
      "loss": 0.0127,
      "reward": 0.2829827591776848,
      "reward_std": 0.09001574665307999,
      "reward_std_all": 0.08801524713635445,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9264424741268158,
      "rewards/forensics_tempsamp_iou_reward": 0.2829827591776848,
      "step": 377
    },
    {
      "completion_length": 18.578125,
      "epoch": 1.0053262316910785,
      "grad_norm": 6.217059135437012,
      "kl": 0.4658203125,
      "learning_rate": 7.486666666666666e-07,
      "loss": 0.0187,
      "reward": 0.2672676742076874,
      "reward_std": 0.08176917396485806,
      "reward_std_all": 0.10196493193507195,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8816224932670593,
      "rewards/forensics_tempsamp_iou_reward": 0.2672676742076874,
      "step": 378
    },
    {
      "completion_length": 16.65625,
      "epoch": 1.007989347536618,
      "grad_norm": 5.683392524719238,
      "kl": 0.4248046875,
      "learning_rate": 7.48e-07,
      "loss": 0.017,
      "reward": 0.24151566624641418,
      "reward_std": 0.09170418232679367,
      "reward_std_all": 0.09686510637402534,
      "rewards-max/forensics_tempsamp_iou_reward": 0.873615562915802,
      "rewards/forensics_tempsamp_iou_reward": 0.24151566624641418,
      "step": 379
    },
    {
      "completion_length": 16.359375,
      "epoch": 1.0106524633821572,
      "grad_norm": 6.495439052581787,
      "kl": 0.48828125,
      "learning_rate": 7.473333333333332e-07,
      "loss": 0.0195,
      "reward": 0.26425062119960785,
      "reward_std": 0.10603040084242821,
      "reward_std_all": 0.11973368376493454,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9696934819221497,
      "rewards/forensics_tempsamp_iou_reward": 0.26425062119960785,
      "step": 380
    },
    {
      "completion_length": 18.703125,
      "epoch": 1.0133155792276964,
      "grad_norm": 6.080895900726318,
      "kl": 0.25927734375,
      "learning_rate": 7.466666666666667e-07,
      "loss": 0.0104,
      "reward": 0.3074478954076767,
      "reward_std": 0.12702349945902824,
      "reward_std_all": 0.1289670541882515,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9047132134437561,
      "rewards/forensics_tempsamp_iou_reward": 0.3074478954076767,
      "step": 381
    },
    {
      "completion_length": 20.25,
      "epoch": 1.0159786950732357,
      "grad_norm": 8.36552619934082,
      "kl": 0.33935546875,
      "learning_rate": 7.459999999999999e-07,
      "loss": 0.0136,
      "reward": 0.21123643219470978,
      "reward_std": 0.11429373919963837,
      "reward_std_all": 0.1140151359140873,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9071649610996246,
      "rewards/forensics_tempsamp_iou_reward": 0.21123643219470978,
      "step": 382
    },
    {
      "completion_length": 17.15625,
      "epoch": 1.018641810918775,
      "grad_norm": 5.578263759613037,
      "kl": 0.20166015625,
      "learning_rate": 7.453333333333333e-07,
      "loss": 0.0081,
      "reward": 0.22059722244739532,
      "reward_std": 0.08442715927958488,
      "reward_std_all": 0.08117666095495224,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7717332541942596,
      "rewards/forensics_tempsamp_iou_reward": 0.22059722244739532,
      "step": 383
    },
    {
      "completion_length": 13.59375,
      "epoch": 1.0213049267643142,
      "grad_norm": 5.335586071014404,
      "kl": 0.29296875,
      "learning_rate": 7.446666666666666e-07,
      "loss": 0.0118,
      "reward": 0.38401074707508087,
      "reward_std": 0.0935121700167656,
      "reward_std_all": 0.10767260193824768,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8876382112503052,
      "rewards/forensics_tempsamp_iou_reward": 0.38401074707508087,
      "step": 384
    },
    {
      "completion_length": 17.546875,
      "epoch": 1.0239680426098536,
      "grad_norm": 4.836752891540527,
      "kl": 0.2177734375,
      "learning_rate": 7.44e-07,
      "loss": 0.0087,
      "reward": 0.3253975957632065,
      "reward_std": 0.07919539511203766,
      "reward_std_all": 0.1028660200536251,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7988548576831818,
      "rewards/forensics_tempsamp_iou_reward": 0.3253975957632065,
      "step": 385
    },
    {
      "completion_length": 14.171875,
      "epoch": 1.0266311584553929,
      "grad_norm": 5.088316917419434,
      "kl": 0.1669921875,
      "learning_rate": 7.433333333333332e-07,
      "loss": 0.0067,
      "reward": 0.34848877787590027,
      "reward_std": 0.12212406471371651,
      "reward_std_all": 0.11630499735474586,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9400644302368164,
      "rewards/forensics_tempsamp_iou_reward": 0.34848877787590027,
      "step": 386
    },
    {
      "completion_length": 14.546875,
      "epoch": 1.0292942743009321,
      "grad_norm": 4.764191150665283,
      "kl": 0.30859375,
      "learning_rate": 7.426666666666667e-07,
      "loss": 0.0123,
      "reward": 0.19875085353851318,
      "reward_std": 0.06737029552459717,
      "reward_std_all": 0.07076878473162651,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5997813045978546,
      "rewards/forensics_tempsamp_iou_reward": 0.19875085353851318,
      "step": 387
    },
    {
      "completion_length": 19.0625,
      "epoch": 1.0319573901464714,
      "grad_norm": 6.08649206161499,
      "kl": 0.19482421875,
      "learning_rate": 7.42e-07,
      "loss": 0.0078,
      "reward": 0.3653087168931961,
      "reward_std": 0.0928693637251854,
      "reward_std_all": 0.08927071467041969,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9606101810932159,
      "rewards/forensics_tempsamp_iou_reward": 0.3653087168931961,
      "step": 388
    },
    {
      "completion_length": 15.515625,
      "epoch": 1.0346205059920106,
      "grad_norm": 9.543729782104492,
      "kl": 0.578125,
      "learning_rate": 7.413333333333333e-07,
      "loss": 0.0231,
      "reward": 0.150979183614254,
      "reward_std": 0.060265474021434784,
      "reward_std_all": 0.06489387899637222,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6166666597127914,
      "rewards/forensics_tempsamp_iou_reward": 0.150979183614254,
      "step": 389
    },
    {
      "completion_length": 16.171875,
      "epoch": 1.0372836218375499,
      "grad_norm": 5.49436092376709,
      "kl": 0.26708984375,
      "learning_rate": 7.406666666666667e-07,
      "loss": 0.0107,
      "reward": 0.34053686261177063,
      "reward_std": 0.08317187428474426,
      "reward_std_all": 0.08794981613755226,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9383468627929688,
      "rewards/forensics_tempsamp_iou_reward": 0.34053686261177063,
      "step": 390
    },
    {
      "completion_length": 15.40625,
      "epoch": 1.0399467376830893,
      "grad_norm": 5.187718391418457,
      "kl": 0.16259765625,
      "learning_rate": 7.4e-07,
      "loss": 0.0065,
      "reward": 0.2977806329727173,
      "reward_std": 0.08076201751828194,
      "reward_std_all": 0.08447354286909103,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9608535766601562,
      "rewards/forensics_tempsamp_iou_reward": 0.2977806329727173,
      "step": 391
    },
    {
      "completion_length": 17.171875,
      "epoch": 1.0426098535286286,
      "grad_norm": 4.612180709838867,
      "kl": 0.349609375,
      "learning_rate": 7.393333333333333e-07,
      "loss": 0.014,
      "reward": 0.3762803450226784,
      "reward_std": 0.05393759720027447,
      "reward_std_all": 0.08727575466036797,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6399387121200562,
      "rewards/forensics_tempsamp_iou_reward": 0.3762803450226784,
      "step": 392
    },
    {
      "completion_length": 22.0625,
      "epoch": 1.0452729693741678,
      "grad_norm": 6.541721343994141,
      "kl": 0.291015625,
      "learning_rate": 7.386666666666666e-07,
      "loss": 0.0117,
      "reward": 0.30786873400211334,
      "reward_std": 0.0896839126944542,
      "reward_std_all": 0.08846055343747139,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9593628346920013,
      "rewards/forensics_tempsamp_iou_reward": 0.30786873400211334,
      "step": 393
    },
    {
      "completion_length": 15.90625,
      "epoch": 1.047936085219707,
      "grad_norm": 6.444670677185059,
      "kl": 0.3935546875,
      "learning_rate": 7.38e-07,
      "loss": 0.0157,
      "reward": 0.15269935131072998,
      "reward_std": 0.06751989759504795,
      "reward_std_all": 0.0748445950448513,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6127957105636597,
      "rewards/forensics_tempsamp_iou_reward": 0.15269935131072998,
      "step": 394
    },
    {
      "completion_length": 17.578125,
      "epoch": 1.0505992010652463,
      "grad_norm": 4.948428153991699,
      "kl": 0.1513671875,
      "learning_rate": 7.373333333333332e-07,
      "loss": 0.0061,
      "reward": 0.2897196486592293,
      "reward_std": 0.07817007042467594,
      "reward_std_all": 0.07159074395895004,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9187822639942169,
      "rewards/forensics_tempsamp_iou_reward": 0.2897196486592293,
      "step": 395
    },
    {
      "completion_length": 15.953125,
      "epoch": 1.0532623169107855,
      "grad_norm": 4.508371353149414,
      "kl": 0.2822265625,
      "learning_rate": 7.366666666666667e-07,
      "loss": 0.0113,
      "reward": 0.3033928722143173,
      "reward_std": 0.06497206538915634,
      "reward_std_all": 0.06617635674774647,
      "rewards-max/forensics_tempsamp_iou_reward": 0.964993417263031,
      "rewards/forensics_tempsamp_iou_reward": 0.3033928722143173,
      "step": 396
    },
    {
      "completion_length": 15.703125,
      "epoch": 1.055925432756325,
      "grad_norm": 5.762344837188721,
      "kl": 0.2783203125,
      "learning_rate": 7.359999999999999e-07,
      "loss": 0.0112,
      "reward": 0.22973227500915527,
      "reward_std": 0.07677709311246872,
      "reward_std_all": 0.07493634521961212,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9449641108512878,
      "rewards/forensics_tempsamp_iou_reward": 0.22973227500915527,
      "step": 397
    },
    {
      "completion_length": 15.21875,
      "epoch": 1.0585885486018642,
      "grad_norm": 6.13419246673584,
      "kl": 0.3642578125,
      "learning_rate": 7.353333333333333e-07,
      "loss": 0.0146,
      "reward": 0.281336709856987,
      "reward_std": 0.0683990903198719,
      "reward_std_all": 0.08738110959529877,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9511622190475464,
      "rewards/forensics_tempsamp_iou_reward": 0.281336709856987,
      "step": 398
    },
    {
      "completion_length": 14.6875,
      "epoch": 1.0612516644474035,
      "grad_norm": 4.8957624435424805,
      "kl": 0.275390625,
      "learning_rate": 7.346666666666666e-07,
      "loss": 0.011,
      "reward": 0.43884454667568207,
      "reward_std": 0.08679267391562462,
      "reward_std_all": 0.10152958706021309,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9740259647369385,
      "rewards/forensics_tempsamp_iou_reward": 0.43884454667568207,
      "step": 399
    },
    {
      "completion_length": 14.84375,
      "epoch": 1.0639147802929427,
      "grad_norm": 5.007294654846191,
      "kl": 0.375,
      "learning_rate": 7.34e-07,
      "loss": 0.015,
      "reward": 0.2732096314430237,
      "reward_std": 0.0783889852464199,
      "reward_std_all": 0.08682272210717201,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9065982699394226,
      "rewards/forensics_tempsamp_iou_reward": 0.2732096314430237,
      "step": 400
    },
    {
      "completion_length": 18.0625,
      "epoch": 1.066577896138482,
      "grad_norm": 5.321002960205078,
      "kl": 0.470703125,
      "learning_rate": 7.333333333333332e-07,
      "loss": 0.0189,
      "reward": 0.2827562317252159,
      "reward_std": 0.08782846108078957,
      "reward_std_all": 0.09881948679685593,
      "rewards-max/forensics_tempsamp_iou_reward": 0.926373690366745,
      "rewards/forensics_tempsamp_iou_reward": 0.2827562317252159,
      "step": 401
    },
    {
      "completion_length": 20.40625,
      "epoch": 1.0692410119840212,
      "grad_norm": 6.30482292175293,
      "kl": 0.265625,
      "learning_rate": 7.326666666666667e-07,
      "loss": 0.0106,
      "reward": 0.32623737305402756,
      "reward_std": 0.0606513824313879,
      "reward_std_all": 0.06109960004687309,
      "rewards-max/forensics_tempsamp_iou_reward": 0.980902761220932,
      "rewards/forensics_tempsamp_iou_reward": 0.32623737305402756,
      "step": 402
    },
    {
      "completion_length": 18.75,
      "epoch": 1.0719041278295607,
      "grad_norm": 4.890859603881836,
      "kl": 0.19189453125,
      "learning_rate": 7.319999999999999e-07,
      "loss": 0.0077,
      "reward": 0.35076405107975006,
      "reward_std": 0.09240445867180824,
      "reward_std_all": 0.09937592223286629,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9508998692035675,
      "rewards/forensics_tempsamp_iou_reward": 0.35076405107975006,
      "step": 403
    },
    {
      "completion_length": 14.484375,
      "epoch": 1.0745672436751,
      "grad_norm": 7.238615989685059,
      "kl": 0.6455078125,
      "learning_rate": 7.313333333333333e-07,
      "loss": 0.0258,
      "reward": 0.2024814561009407,
      "reward_std": 0.07130370661616325,
      "reward_std_all": 0.06644907779991627,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9555964767932892,
      "rewards/forensics_tempsamp_iou_reward": 0.2024814561009407,
      "step": 404
    },
    {
      "completion_length": 17.859375,
      "epoch": 1.0772303595206392,
      "grad_norm": 5.4003190994262695,
      "kl": 0.33056640625,
      "learning_rate": 7.306666666666666e-07,
      "loss": 0.0132,
      "reward": 0.22423213720321655,
      "reward_std": 0.06929400563240051,
      "reward_std_all": 0.08247461542487144,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8099910318851471,
      "rewards/forensics_tempsamp_iou_reward": 0.22423213720321655,
      "step": 405
    },
    {
      "completion_length": 18.484375,
      "epoch": 1.0798934753661784,
      "grad_norm": 5.533071041107178,
      "kl": 0.5361328125,
      "learning_rate": 7.3e-07,
      "loss": 0.0215,
      "reward": 0.2641567960381508,
      "reward_std": 0.11416347697377205,
      "reward_std_all": 0.10453672707080841,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9612180888652802,
      "rewards/forensics_tempsamp_iou_reward": 0.2641567960381508,
      "step": 406
    },
    {
      "completion_length": 17.984375,
      "epoch": 1.0825565912117177,
      "grad_norm": 6.13942813873291,
      "kl": 0.2939453125,
      "learning_rate": 7.293333333333332e-07,
      "loss": 0.0118,
      "reward": 0.2664334923028946,
      "reward_std": 0.059384338557720184,
      "reward_std_all": 0.06621795520186424,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9098145067691803,
      "rewards/forensics_tempsamp_iou_reward": 0.2664334923028946,
      "step": 407
    },
    {
      "completion_length": 18.25,
      "epoch": 1.085219707057257,
      "grad_norm": 7.008934497833252,
      "kl": 0.4169921875,
      "learning_rate": 7.286666666666666e-07,
      "loss": 0.0167,
      "reward": 0.26707319915294647,
      "reward_std": 0.08711029216647148,
      "reward_std_all": 0.09343261644244194,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9154195189476013,
      "rewards/forensics_tempsamp_iou_reward": 0.26707319915294647,
      "step": 408
    },
    {
      "completion_length": 20.078125,
      "epoch": 1.0878828229027964,
      "grad_norm": 6.74454402923584,
      "kl": 0.4794921875,
      "learning_rate": 7.28e-07,
      "loss": 0.0192,
      "reward": 0.2889031022787094,
      "reward_std": 0.09110469743609428,
      "reward_std_all": 0.1014992818236351,
      "rewards-max/forensics_tempsamp_iou_reward": 0.920388251543045,
      "rewards/forensics_tempsamp_iou_reward": 0.2889031022787094,
      "step": 409
    },
    {
      "completion_length": 21.828125,
      "epoch": 1.0905459387483356,
      "grad_norm": 6.1425018310546875,
      "kl": 0.3125,
      "learning_rate": 7.273333333333333e-07,
      "loss": 0.0125,
      "reward": 0.2944145053625107,
      "reward_std": 0.09102082997560501,
      "reward_std_all": 0.1242728903889656,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8446861505508423,
      "rewards/forensics_tempsamp_iou_reward": 0.2944145053625107,
      "step": 410
    },
    {
      "completion_length": 18.40625,
      "epoch": 1.0932090545938749,
      "grad_norm": 6.577932834625244,
      "kl": 0.31005859375,
      "learning_rate": 7.266666666666667e-07,
      "loss": 0.0124,
      "reward": 0.23206007480621338,
      "reward_std": 0.07956728152930737,
      "reward_std_all": 0.08542505279183388,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8739583194255829,
      "rewards/forensics_tempsamp_iou_reward": 0.23206007480621338,
      "step": 411
    },
    {
      "completion_length": 16.796875,
      "epoch": 1.095872170439414,
      "grad_norm": 5.898149490356445,
      "kl": 0.47265625,
      "learning_rate": 7.259999999999999e-07,
      "loss": 0.0189,
      "reward": 0.22074207663536072,
      "reward_std": 0.09985409677028656,
      "reward_std_all": 0.09802129119634628,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7376511543989182,
      "rewards/forensics_tempsamp_iou_reward": 0.22074207663536072,
      "step": 412
    },
    {
      "completion_length": 19.953125,
      "epoch": 1.0985352862849533,
      "grad_norm": 7.6911468505859375,
      "kl": 0.3125,
      "learning_rate": 7.253333333333334e-07,
      "loss": 0.0125,
      "reward": 0.18203971534967422,
      "reward_std": 0.061056237667798996,
      "reward_std_all": 0.07413649186491966,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6594771295785904,
      "rewards/forensics_tempsamp_iou_reward": 0.18203971534967422,
      "step": 413
    },
    {
      "completion_length": 19.859375,
      "epoch": 1.1011984021304926,
      "grad_norm": 4.758620738983154,
      "kl": 0.44921875,
      "learning_rate": 7.246666666666666e-07,
      "loss": 0.0179,
      "reward": 0.17448514327406883,
      "reward_std": 0.06822706013917923,
      "reward_std_all": 0.07320082560181618,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6358261257410049,
      "rewards/forensics_tempsamp_iou_reward": 0.17448514327406883,
      "step": 414
    },
    {
      "completion_length": 24.9375,
      "epoch": 1.103861517976032,
      "grad_norm": 7.347010135650635,
      "kl": 0.2705078125,
      "learning_rate": 7.24e-07,
      "loss": 0.0108,
      "reward": 0.16312908381223679,
      "reward_std": 0.080218356102705,
      "reward_std_all": 0.08932684734463692,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6776967346668243,
      "rewards/forensics_tempsamp_iou_reward": 0.16312908381223679,
      "step": 415
    },
    {
      "completion_length": 20.390625,
      "epoch": 1.1065246338215713,
      "grad_norm": 6.631772518157959,
      "kl": 0.5693359375,
      "learning_rate": 7.233333333333333e-07,
      "loss": 0.0227,
      "reward": 0.17101866751909256,
      "reward_std": 0.08334573358297348,
      "reward_std_all": 0.09184156730771065,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5207357853651047,
      "rewards/forensics_tempsamp_iou_reward": 0.17101866751909256,
      "step": 416
    },
    {
      "completion_length": 19.96875,
      "epoch": 1.1091877496671105,
      "grad_norm": 5.159544467926025,
      "kl": 0.392578125,
      "learning_rate": 7.226666666666667e-07,
      "loss": 0.0157,
      "reward": 0.3111521005630493,
      "reward_std": 0.11605551093816757,
      "reward_std_all": 0.11268216744065285,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7762371301651001,
      "rewards/forensics_tempsamp_iou_reward": 0.3111521005630493,
      "step": 417
    },
    {
      "completion_length": 17.71875,
      "epoch": 1.1118508655126498,
      "grad_norm": 5.1551594734191895,
      "kl": 0.4609375,
      "learning_rate": 7.219999999999999e-07,
      "loss": 0.0184,
      "reward": 0.24157294631004333,
      "reward_std": 0.06982464902102947,
      "reward_std_all": 0.08523645251989365,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7377655804157257,
      "rewards/forensics_tempsamp_iou_reward": 0.24157294631004333,
      "step": 418
    },
    {
      "completion_length": 20.296875,
      "epoch": 1.114513981358189,
      "grad_norm": 6.08780574798584,
      "kl": 0.427734375,
      "learning_rate": 7.213333333333334e-07,
      "loss": 0.0171,
      "reward": 0.31909582018852234,
      "reward_std": 0.08265724778175354,
      "reward_std_all": 0.0952620804309845,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8653846085071564,
      "rewards/forensics_tempsamp_iou_reward": 0.31909582018852234,
      "step": 419
    },
    {
      "completion_length": 18.625,
      "epoch": 1.1171770972037283,
      "grad_norm": 8.414843559265137,
      "kl": 0.568359375,
      "learning_rate": 7.206666666666666e-07,
      "loss": 0.0227,
      "reward": 0.24666999280452728,
      "reward_std": 0.09726276248693466,
      "reward_std_all": 0.09524867311120033,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8324289917945862,
      "rewards/forensics_tempsamp_iou_reward": 0.24666999280452728,
      "step": 420
    },
    {
      "completion_length": 24.953125,
      "epoch": 1.1198402130492677,
      "grad_norm": 5.771307468414307,
      "kl": 0.3984375,
      "learning_rate": 7.2e-07,
      "loss": 0.0159,
      "reward": 0.19322269409894943,
      "reward_std": 0.07529480382800102,
      "reward_std_all": 0.08173980563879013,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6317417621612549,
      "rewards/forensics_tempsamp_iou_reward": 0.19322269409894943,
      "step": 421
    },
    {
      "completion_length": 22.078125,
      "epoch": 1.122503328894807,
      "grad_norm": 6.849403381347656,
      "kl": 0.376953125,
      "learning_rate": 7.193333333333333e-07,
      "loss": 0.0151,
      "reward": 0.26113712787628174,
      "reward_std": 0.11335304006934166,
      "reward_std_all": 0.11935120820999146,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7432489395141602,
      "rewards/forensics_tempsamp_iou_reward": 0.26113712787628174,
      "step": 422
    },
    {
      "completion_length": 21.109375,
      "epoch": 1.1251664447403462,
      "grad_norm": 7.982503414154053,
      "kl": 0.3056640625,
      "learning_rate": 7.186666666666667e-07,
      "loss": 0.0122,
      "reward": 0.14161689579486847,
      "reward_std": 0.08990873768925667,
      "reward_std_all": 0.09177261590957642,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5638316422700882,
      "rewards/forensics_tempsamp_iou_reward": 0.14161689579486847,
      "step": 423
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.1278295605858855,
      "grad_norm": 8.289421081542969,
      "kl": 0.5146484375,
      "learning_rate": 7.179999999999999e-07,
      "loss": 0.0206,
      "reward": 0.2128107249736786,
      "reward_std": 0.07348750159144402,
      "reward_std_all": 0.100868821144104,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7451310753822327,
      "rewards/forensics_tempsamp_iou_reward": 0.2128107249736786,
      "step": 424
    },
    {
      "completion_length": 23.546875,
      "epoch": 1.1304926764314247,
      "grad_norm": 6.30644416809082,
      "kl": 0.58984375,
      "learning_rate": 7.173333333333333e-07,
      "loss": 0.0236,
      "reward": 0.25013941526412964,
      "reward_std": 0.10498220846056938,
      "reward_std_all": 0.11780118197202682,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7504249811172485,
      "rewards/forensics_tempsamp_iou_reward": 0.25013941526412964,
      "step": 425
    },
    {
      "completion_length": 21.46875,
      "epoch": 1.133155792276964,
      "grad_norm": 5.596959114074707,
      "kl": 0.271484375,
      "learning_rate": 7.166666666666667e-07,
      "loss": 0.0109,
      "reward": 0.2730278968811035,
      "reward_std": 0.11620871722698212,
      "reward_std_all": 0.11770684272050858,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8215202391147614,
      "rewards/forensics_tempsamp_iou_reward": 0.2730278968811035,
      "step": 426
    },
    {
      "completion_length": 21.90625,
      "epoch": 1.1358189081225034,
      "grad_norm": 5.1786956787109375,
      "kl": 0.330078125,
      "learning_rate": 7.159999999999999e-07,
      "loss": 0.0132,
      "reward": 0.25320984423160553,
      "reward_std": 0.12959468737244606,
      "reward_std_all": 0.12348209321498871,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9496037662029266,
      "rewards/forensics_tempsamp_iou_reward": 0.25320984423160553,
      "step": 427
    },
    {
      "completion_length": 21.21875,
      "epoch": 1.1384820239680427,
      "grad_norm": 7.264177322387695,
      "kl": 0.4072265625,
      "learning_rate": 7.153333333333334e-07,
      "loss": 0.0163,
      "reward": 0.3022685647010803,
      "reward_std": 0.11661017313599586,
      "reward_std_all": 0.11265075579285622,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9521654844284058,
      "rewards/forensics_tempsamp_iou_reward": 0.3022685647010803,
      "step": 428
    },
    {
      "completion_length": 22.078125,
      "epoch": 1.141145139813582,
      "grad_norm": 7.373708248138428,
      "kl": 0.48046875,
      "learning_rate": 7.146666666666666e-07,
      "loss": 0.0192,
      "reward": 0.20836608111858368,
      "reward_std": 0.08766015246510506,
      "reward_std_all": 0.09088011458516121,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5778895020484924,
      "rewards/forensics_tempsamp_iou_reward": 0.20836608111858368,
      "step": 429
    },
    {
      "completion_length": 21.8125,
      "epoch": 1.1438082556591211,
      "grad_norm": 5.040229320526123,
      "kl": 0.4248046875,
      "learning_rate": 7.14e-07,
      "loss": 0.017,
      "reward": 0.202855184674263,
      "reward_std": 0.10881329327821732,
      "reward_std_all": 0.10690121352672577,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7160764336585999,
      "rewards/forensics_tempsamp_iou_reward": 0.202855184674263,
      "step": 430
    },
    {
      "completion_length": 17.5625,
      "epoch": 1.1464713715046604,
      "grad_norm": 7.136110782623291,
      "kl": 0.439453125,
      "learning_rate": 7.133333333333333e-07,
      "loss": 0.0176,
      "reward": 0.21965894103050232,
      "reward_std": 0.08209259249269962,
      "reward_std_all": 0.08637656085193157,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8000578880310059,
      "rewards/forensics_tempsamp_iou_reward": 0.21965894103050232,
      "step": 431
    },
    {
      "completion_length": 22.96875,
      "epoch": 1.1491344873501999,
      "grad_norm": 5.813126087188721,
      "kl": 0.390625,
      "learning_rate": 7.126666666666667e-07,
      "loss": 0.0156,
      "reward": 0.30156826972961426,
      "reward_std": 0.09885082393884659,
      "reward_std_all": 0.10454495251178741,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9203879833221436,
      "rewards/forensics_tempsamp_iou_reward": 0.30156826972961426,
      "step": 432
    },
    {
      "completion_length": 21.375,
      "epoch": 1.151797603195739,
      "grad_norm": 5.520924091339111,
      "kl": 0.431640625,
      "learning_rate": 7.119999999999999e-07,
      "loss": 0.0172,
      "reward": 0.17930394411087036,
      "reward_std": 0.0746020395308733,
      "reward_std_all": 0.07018859311938286,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8561019897460938,
      "rewards/forensics_tempsamp_iou_reward": 0.17930394411087036,
      "step": 433
    },
    {
      "completion_length": 25.28125,
      "epoch": 1.1544607190412783,
      "grad_norm": 5.65551233291626,
      "kl": 0.44140625,
      "learning_rate": 7.113333333333334e-07,
      "loss": 0.0177,
      "reward": 0.22088930010795593,
      "reward_std": 0.0925767794251442,
      "reward_std_all": 0.10450439155101776,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7311964333057404,
      "rewards/forensics_tempsamp_iou_reward": 0.22088930010795593,
      "step": 434
    },
    {
      "completion_length": 19.515625,
      "epoch": 1.1571238348868176,
      "grad_norm": 6.697516441345215,
      "kl": 0.4775390625,
      "learning_rate": 7.106666666666666e-07,
      "loss": 0.0191,
      "reward": 0.27185066044330597,
      "reward_std": 0.10389591008424759,
      "reward_std_all": 0.09959854185581207,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6136258095502853,
      "rewards/forensics_tempsamp_iou_reward": 0.27185066044330597,
      "step": 435
    },
    {
      "completion_length": 17.671875,
      "epoch": 1.1597869507323568,
      "grad_norm": 6.834056854248047,
      "kl": 0.3544921875,
      "learning_rate": 7.1e-07,
      "loss": 0.0142,
      "reward": 0.28581005334854126,
      "reward_std": 0.08302512019872665,
      "reward_std_all": 0.08213324658572674,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8454404175281525,
      "rewards/forensics_tempsamp_iou_reward": 0.28581005334854126,
      "step": 436
    },
    {
      "completion_length": 24.421875,
      "epoch": 1.162450066577896,
      "grad_norm": 6.971820831298828,
      "kl": 0.53125,
      "learning_rate": 7.093333333333333e-07,
      "loss": 0.0213,
      "reward": 0.29380153119564056,
      "reward_std": 0.15656428784132004,
      "reward_std_all": 0.15715204924345016,
      "rewards-max/forensics_tempsamp_iou_reward": 0.820197731256485,
      "rewards/forensics_tempsamp_iou_reward": 0.29380153119564056,
      "step": 437
    },
    {
      "completion_length": 19.109375,
      "epoch": 1.1651131824234353,
      "grad_norm": 7.877826690673828,
      "kl": 0.478515625,
      "learning_rate": 7.086666666666667e-07,
      "loss": 0.0191,
      "reward": 0.27585354447364807,
      "reward_std": 0.0920591689646244,
      "reward_std_all": 0.08887503668665886,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7474499046802521,
      "rewards/forensics_tempsamp_iou_reward": 0.27585354447364807,
      "step": 438
    },
    {
      "completion_length": 23.125,
      "epoch": 1.1677762982689748,
      "grad_norm": 7.392991542816162,
      "kl": 0.541015625,
      "learning_rate": 7.079999999999999e-07,
      "loss": 0.0217,
      "reward": 0.1611448898911476,
      "reward_std": 0.05425548180937767,
      "reward_std_all": 0.08111074566841125,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5619900971651077,
      "rewards/forensics_tempsamp_iou_reward": 0.1611448898911476,
      "step": 439
    },
    {
      "completion_length": 21.515625,
      "epoch": 1.170439414114514,
      "grad_norm": 7.071908950805664,
      "kl": 0.33837890625,
      "learning_rate": 7.073333333333333e-07,
      "loss": 0.0135,
      "reward": 0.366722047328949,
      "reward_std": 0.10723821818828583,
      "reward_std_all": 0.12555362284183502,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9543014466762543,
      "rewards/forensics_tempsamp_iou_reward": 0.366722047328949,
      "step": 440
    },
    {
      "completion_length": 19.515625,
      "epoch": 1.1731025299600533,
      "grad_norm": 5.8077778816223145,
      "kl": 0.548828125,
      "learning_rate": 7.066666666666666e-07,
      "loss": 0.022,
      "reward": 0.31062455475330353,
      "reward_std": 0.1092379242181778,
      "reward_std_all": 0.11000531539320946,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9074727594852448,
      "rewards/forensics_tempsamp_iou_reward": 0.31062455475330353,
      "step": 441
    },
    {
      "completion_length": 20.5625,
      "epoch": 1.1757656458055925,
      "grad_norm": 7.3544087409973145,
      "kl": 0.384765625,
      "learning_rate": 7.059999999999999e-07,
      "loss": 0.0154,
      "reward": 0.22106829285621643,
      "reward_std": 0.05850617401301861,
      "reward_std_all": 0.07736653462052345,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6358090341091156,
      "rewards/forensics_tempsamp_iou_reward": 0.22106829285621643,
      "step": 442
    },
    {
      "completion_length": 18.328125,
      "epoch": 1.1784287616511318,
      "grad_norm": 4.979027271270752,
      "kl": 0.37890625,
      "learning_rate": 7.053333333333333e-07,
      "loss": 0.0152,
      "reward": 0.1756921112537384,
      "reward_std": 0.06809777766466141,
      "reward_std_all": 0.07105658017098904,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6184850633144379,
      "rewards/forensics_tempsamp_iou_reward": 0.1756921112537384,
      "step": 443
    },
    {
      "completion_length": 15.890625,
      "epoch": 1.1810918774966712,
      "grad_norm": 5.844629287719727,
      "kl": 0.28125,
      "learning_rate": 7.046666666666666e-07,
      "loss": 0.0113,
      "reward": 0.28844671696424484,
      "reward_std": 0.08633052557706833,
      "reward_std_all": 0.08950604498386383,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9555555582046509,
      "rewards/forensics_tempsamp_iou_reward": 0.28844671696424484,
      "step": 444
    },
    {
      "completion_length": 13.390625,
      "epoch": 1.1837549933422105,
      "grad_norm": 6.445103168487549,
      "kl": 0.25732421875,
      "learning_rate": 7.04e-07,
      "loss": 0.0103,
      "reward": 0.25773923844099045,
      "reward_std": 0.080429308116436,
      "reward_std_all": 0.08410021290183067,
      "rewards-max/forensics_tempsamp_iou_reward": 0.894909679889679,
      "rewards/forensics_tempsamp_iou_reward": 0.25773923844099045,
      "step": 445
    },
    {
      "completion_length": 16.3125,
      "epoch": 1.1864181091877497,
      "grad_norm": 5.131852626800537,
      "kl": 0.24072265625,
      "learning_rate": 7.033333333333333e-07,
      "loss": 0.0096,
      "reward": 0.3195564076304436,
      "reward_std": 0.08171862736344337,
      "reward_std_all": 0.07955081015825272,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7892156839370728,
      "rewards/forensics_tempsamp_iou_reward": 0.3195564076304436,
      "step": 446
    },
    {
      "completion_length": 18.015625,
      "epoch": 1.189081225033289,
      "grad_norm": 5.971600532531738,
      "kl": 0.4755859375,
      "learning_rate": 7.026666666666667e-07,
      "loss": 0.019,
      "reward": 0.16479765623807907,
      "reward_std": 0.0720738135278225,
      "reward_std_all": 0.07146235182881355,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6822097450494766,
      "rewards/forensics_tempsamp_iou_reward": 0.16479765623807907,
      "step": 447
    },
    {
      "completion_length": 16.3125,
      "epoch": 1.1917443408788282,
      "grad_norm": 6.954166889190674,
      "kl": 0.3095703125,
      "learning_rate": 7.019999999999999e-07,
      "loss": 0.0124,
      "reward": 0.21390989422798157,
      "reward_std": 0.0792424064129591,
      "reward_std_all": 0.07832948490977287,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7168269753456116,
      "rewards/forensics_tempsamp_iou_reward": 0.21390989422798157,
      "step": 448
    },
    {
      "completion_length": 15.578125,
      "epoch": 1.1944074567243674,
      "grad_norm": 4.7751145362854,
      "kl": 0.560546875,
      "learning_rate": 7.013333333333334e-07,
      "loss": 0.0225,
      "reward": 0.27821896970272064,
      "reward_std": 0.08594910055398941,
      "reward_std_all": 0.08867138251662254,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8332760334014893,
      "rewards/forensics_tempsamp_iou_reward": 0.27821896970272064,
      "step": 449
    },
    {
      "completion_length": 14.28125,
      "epoch": 1.1970705725699067,
      "grad_norm": 6.826998233795166,
      "kl": 0.5224609375,
      "learning_rate": 7.006666666666666e-07,
      "loss": 0.0209,
      "reward": 0.31598280370235443,
      "reward_std": 0.12502698600292206,
      "reward_std_all": 0.12451799213886261,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9509479701519012,
      "rewards/forensics_tempsamp_iou_reward": 0.31598280370235443,
      "step": 450
    },
    {
      "completion_length": 16.21875,
      "epoch": 1.1997336884154461,
      "grad_norm": 7.93365478515625,
      "kl": 0.62890625,
      "learning_rate": 7e-07,
      "loss": 0.0252,
      "reward": 0.2344684600830078,
      "reward_std": 0.1122911088168621,
      "reward_std_all": 0.12230058014392853,
      "rewards-max/forensics_tempsamp_iou_reward": 0.774293452501297,
      "rewards/forensics_tempsamp_iou_reward": 0.2344684600830078,
      "step": 451
    },
    {
      "completion_length": 17.671875,
      "epoch": 1.2023968042609854,
      "grad_norm": 5.5738301277160645,
      "kl": 0.39013671875,
      "learning_rate": 6.993333333333333e-07,
      "loss": 0.0156,
      "reward": 0.26992465555667877,
      "reward_std": 0.10288843512535095,
      "reward_std_all": 0.10206742957234383,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7202019989490509,
      "rewards/forensics_tempsamp_iou_reward": 0.26992465555667877,
      "step": 452
    },
    {
      "completion_length": 15.15625,
      "epoch": 1.2050599201065246,
      "grad_norm": 6.214698314666748,
      "kl": 0.50390625,
      "learning_rate": 6.986666666666667e-07,
      "loss": 0.0202,
      "reward": 0.3075903058052063,
      "reward_std": 0.07896247506141663,
      "reward_std_all": 0.09478383511304855,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9481810629367828,
      "rewards/forensics_tempsamp_iou_reward": 0.3075903058052063,
      "step": 453
    },
    {
      "completion_length": 15.359375,
      "epoch": 1.2077230359520639,
      "grad_norm": 7.109299182891846,
      "kl": 0.29541015625,
      "learning_rate": 6.979999999999999e-07,
      "loss": 0.0118,
      "reward": 0.27293066680431366,
      "reward_std": 0.10362343490123749,
      "reward_std_all": 0.10093627870082855,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7933897078037262,
      "rewards/forensics_tempsamp_iou_reward": 0.27293066680431366,
      "step": 454
    },
    {
      "completion_length": 15.546875,
      "epoch": 1.2103861517976031,
      "grad_norm": 6.3956427574157715,
      "kl": 0.44921875,
      "learning_rate": 6.973333333333333e-07,
      "loss": 0.018,
      "reward": 0.1990642249584198,
      "reward_std": 0.05959024652838707,
      "reward_std_all": 0.08237450569868088,
      "rewards-max/forensics_tempsamp_iou_reward": 0.813567578792572,
      "rewards/forensics_tempsamp_iou_reward": 0.1990642249584198,
      "step": 455
    },
    {
      "completion_length": 20.6875,
      "epoch": 1.2130492676431426,
      "grad_norm": 5.170581817626953,
      "kl": 0.3291015625,
      "learning_rate": 6.966666666666666e-07,
      "loss": 0.0132,
      "reward": 0.1857682392001152,
      "reward_std": 0.08677975833415985,
      "reward_std_all": 0.08309493586421013,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5952380895614624,
      "rewards/forensics_tempsamp_iou_reward": 0.1857682392001152,
      "step": 456
    },
    {
      "completion_length": 18.65625,
      "epoch": 1.2157123834886818,
      "grad_norm": 5.977161884307861,
      "kl": 0.5234375,
      "learning_rate": 6.959999999999999e-07,
      "loss": 0.0209,
      "reward": 0.18783877789974213,
      "reward_std": 0.08740159496665001,
      "reward_std_all": 0.08833998814225197,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5964601784944534,
      "rewards/forensics_tempsamp_iou_reward": 0.18783877789974213,
      "step": 457
    },
    {
      "completion_length": 18.703125,
      "epoch": 1.218375499334221,
      "grad_norm": 6.488882541656494,
      "kl": 0.2578125,
      "learning_rate": 6.953333333333333e-07,
      "loss": 0.0104,
      "reward": 0.2062312513589859,
      "reward_std": 0.09292097017168999,
      "reward_std_all": 0.09177758544683456,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8832218647003174,
      "rewards/forensics_tempsamp_iou_reward": 0.2062312513589859,
      "step": 458
    },
    {
      "completion_length": 16.59375,
      "epoch": 1.2210386151797603,
      "grad_norm": 6.952360153198242,
      "kl": 0.396484375,
      "learning_rate": 6.946666666666666e-07,
      "loss": 0.0159,
      "reward": 0.20613553375005722,
      "reward_std": 0.08777660876512527,
      "reward_std_all": 0.08292581886053085,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5595362484455109,
      "rewards/forensics_tempsamp_iou_reward": 0.20613553375005722,
      "step": 459
    },
    {
      "completion_length": 15.203125,
      "epoch": 1.2237017310252996,
      "grad_norm": 8.342304229736328,
      "kl": 0.25048828125,
      "learning_rate": 6.939999999999999e-07,
      "loss": 0.01,
      "reward": 0.27785497158765793,
      "reward_std": 0.10876452550292015,
      "reward_std_all": 0.11377362906932831,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8439429104328156,
      "rewards/forensics_tempsamp_iou_reward": 0.27785497158765793,
      "step": 460
    },
    {
      "completion_length": 17.21875,
      "epoch": 1.2263648468708388,
      "grad_norm": 6.878493309020996,
      "kl": 0.515625,
      "learning_rate": 6.933333333333333e-07,
      "loss": 0.0206,
      "reward": 0.25992581248283386,
      "reward_std": 0.07267502322793007,
      "reward_std_all": 0.10772080719470978,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9145833253860474,
      "rewards/forensics_tempsamp_iou_reward": 0.25992581248283386,
      "step": 461
    },
    {
      "completion_length": 13.375,
      "epoch": 1.229027962716378,
      "grad_norm": 6.4706926345825195,
      "kl": 0.296875,
      "learning_rate": 6.926666666666666e-07,
      "loss": 0.0119,
      "reward": 0.13737184554338455,
      "reward_std": 0.050831910222768784,
      "reward_std_all": 0.054406462237238884,
      "rewards-max/forensics_tempsamp_iou_reward": 0.60779669880867,
      "rewards/forensics_tempsamp_iou_reward": 0.13737184554338455,
      "step": 462
    },
    {
      "completion_length": 19.59375,
      "epoch": 1.2316910785619175,
      "grad_norm": 6.738338470458984,
      "kl": 0.412109375,
      "learning_rate": 6.919999999999999e-07,
      "loss": 0.0164,
      "reward": 0.32949063181877136,
      "reward_std": 0.13280809298157692,
      "reward_std_all": 0.15631545335054398,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8053593933582306,
      "rewards/forensics_tempsamp_iou_reward": 0.32949063181877136,
      "step": 463
    },
    {
      "completion_length": 15.0625,
      "epoch": 1.2343541944074568,
      "grad_norm": 5.154170513153076,
      "kl": 0.2880859375,
      "learning_rate": 6.913333333333334e-07,
      "loss": 0.0115,
      "reward": 0.17641931772232056,
      "reward_std": 0.06363973021507263,
      "reward_std_all": 0.08665214851498604,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6259495615959167,
      "rewards/forensics_tempsamp_iou_reward": 0.17641931772232056,
      "step": 464
    },
    {
      "completion_length": 13.15625,
      "epoch": 1.237017310252996,
      "grad_norm": 7.076888084411621,
      "kl": 0.201171875,
      "learning_rate": 6.906666666666666e-07,
      "loss": 0.0081,
      "reward": 0.30389632284641266,
      "reward_std": 0.08357268199324608,
      "reward_std_all": 0.08654941618442535,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8799792528152466,
      "rewards/forensics_tempsamp_iou_reward": 0.30389632284641266,
      "step": 465
    },
    {
      "completion_length": 25.8125,
      "epoch": 1.2396804260985352,
      "grad_norm": 5.473649024963379,
      "kl": 0.42578125,
      "learning_rate": 6.9e-07,
      "loss": 0.017,
      "reward": 0.2326282486319542,
      "reward_std": 0.12403534352779388,
      "reward_std_all": 0.11982159875333309,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6458529382944107,
      "rewards/forensics_tempsamp_iou_reward": 0.2326282486319542,
      "step": 466
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.2423435419440745,
      "grad_norm": 7.424006938934326,
      "kl": 0.486328125,
      "learning_rate": 6.893333333333333e-07,
      "loss": 0.0194,
      "reward": 0.19101699441671371,
      "reward_std": 0.08983291685581207,
      "reward_std_all": 0.08605680242180824,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6536507904529572,
      "rewards/forensics_tempsamp_iou_reward": 0.19101699441671371,
      "step": 467
    },
    {
      "completion_length": 15.953125,
      "epoch": 1.245006657789614,
      "grad_norm": 5.472952842712402,
      "kl": 0.224609375,
      "learning_rate": 6.886666666666667e-07,
      "loss": 0.009,
      "reward": 0.26760464906692505,
      "reward_std": 0.1153276190161705,
      "reward_std_all": 0.10241818428039551,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8836187720298767,
      "rewards/forensics_tempsamp_iou_reward": 0.26760464906692505,
      "step": 468
    },
    {
      "completion_length": 16.34375,
      "epoch": 1.2476697736351532,
      "grad_norm": 5.535576820373535,
      "kl": 0.4404296875,
      "learning_rate": 6.879999999999999e-07,
      "loss": 0.0177,
      "reward": 0.220784991979599,
      "reward_std": 0.08006089180707932,
      "reward_std_all": 0.07465878501534462,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8463149666786194,
      "rewards/forensics_tempsamp_iou_reward": 0.220784991979599,
      "step": 469
    },
    {
      "completion_length": 16.4375,
      "epoch": 1.2503328894806924,
      "grad_norm": 5.452118396759033,
      "kl": 0.44921875,
      "learning_rate": 6.873333333333334e-07,
      "loss": 0.018,
      "reward": 0.29119832813739777,
      "reward_std": 0.07359790056943893,
      "reward_std_all": 0.07477663829922676,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9859219789505005,
      "rewards/forensics_tempsamp_iou_reward": 0.29119832813739777,
      "step": 470
    },
    {
      "completion_length": 17.515625,
      "epoch": 1.2529960053262317,
      "grad_norm": 5.164065361022949,
      "kl": 0.39404296875,
      "learning_rate": 6.866666666666666e-07,
      "loss": 0.0158,
      "reward": 0.3196866661310196,
      "reward_std": 0.06807951629161835,
      "reward_std_all": 0.08534237369894981,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9419248998165131,
      "rewards/forensics_tempsamp_iou_reward": 0.3196866661310196,
      "step": 471
    },
    {
      "completion_length": 20.5625,
      "epoch": 1.255659121171771,
      "grad_norm": 6.180549621582031,
      "kl": 0.396484375,
      "learning_rate": 6.86e-07,
      "loss": 0.0159,
      "reward": 0.30064885318279266,
      "reward_std": 0.0983964279294014,
      "reward_std_all": 0.09403841011226177,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9443211853504181,
      "rewards/forensics_tempsamp_iou_reward": 0.30064885318279266,
      "step": 472
    },
    {
      "completion_length": 18.234375,
      "epoch": 1.2583222370173104,
      "grad_norm": 4.908322334289551,
      "kl": 0.2744140625,
      "learning_rate": 6.853333333333333e-07,
      "loss": 0.011,
      "reward": 0.24093789607286453,
      "reward_std": 0.09438136219978333,
      "reward_std_all": 0.09056784957647324,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9216944575309753,
      "rewards/forensics_tempsamp_iou_reward": 0.24093789607286453,
      "step": 473
    },
    {
      "completion_length": 17.5625,
      "epoch": 1.2609853528628494,
      "grad_norm": 7.263335704803467,
      "kl": 0.640625,
      "learning_rate": 6.846666666666666e-07,
      "loss": 0.0256,
      "reward": 0.2196395844221115,
      "reward_std": 0.07207897864282131,
      "reward_std_all": 0.0680575743317604,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9247281551361084,
      "rewards/forensics_tempsamp_iou_reward": 0.2196395844221115,
      "step": 474
    },
    {
      "completion_length": 16.96875,
      "epoch": 1.2636484687083889,
      "grad_norm": 6.970869541168213,
      "kl": 0.56982421875,
      "learning_rate": 6.84e-07,
      "loss": 0.0228,
      "reward": 0.2657005786895752,
      "reward_std": 0.09816013649106026,
      "reward_std_all": 0.0943864956498146,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9488503038883209,
      "rewards/forensics_tempsamp_iou_reward": 0.2657005786895752,
      "step": 475
    },
    {
      "completion_length": 14.6875,
      "epoch": 1.2663115845539281,
      "grad_norm": 4.684751033782959,
      "kl": 0.25732421875,
      "learning_rate": 6.833333333333333e-07,
      "loss": 0.0103,
      "reward": 0.19136547297239304,
      "reward_std": 0.058859724551439285,
      "reward_std_all": 0.05770501773804426,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6888888776302338,
      "rewards/forensics_tempsamp_iou_reward": 0.19136547297239304,
      "step": 476
    },
    {
      "completion_length": 19.03125,
      "epoch": 1.2689747003994674,
      "grad_norm": 5.65306282043457,
      "kl": 0.24560546875,
      "learning_rate": 6.826666666666666e-07,
      "loss": 0.0098,
      "reward": 0.22359725087881088,
      "reward_std": 0.08853470534086227,
      "reward_std_all": 0.08134619519114494,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7284925282001495,
      "rewards/forensics_tempsamp_iou_reward": 0.22359725087881088,
      "step": 477
    },
    {
      "completion_length": 17.671875,
      "epoch": 1.2716378162450066,
      "grad_norm": 5.379079818725586,
      "kl": 0.533203125,
      "learning_rate": 6.82e-07,
      "loss": 0.0212,
      "reward": 0.25664281100034714,
      "reward_std": 0.08561865240335464,
      "reward_std_all": 0.08478989824652672,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9669097661972046,
      "rewards/forensics_tempsamp_iou_reward": 0.25664281100034714,
      "step": 478
    },
    {
      "completion_length": 18.0625,
      "epoch": 1.2743009320905458,
      "grad_norm": 5.643711566925049,
      "kl": 0.6259765625,
      "learning_rate": 6.813333333333333e-07,
      "loss": 0.025,
      "reward": 0.23872940242290497,
      "reward_std": 0.09370672702789307,
      "reward_std_all": 0.09165336564183235,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8531966209411621,
      "rewards/forensics_tempsamp_iou_reward": 0.23872940242290497,
      "step": 479
    },
    {
      "completion_length": 23.578125,
      "epoch": 1.2769640479360853,
      "grad_norm": 4.356174468994141,
      "kl": 0.337890625,
      "learning_rate": 6.806666666666666e-07,
      "loss": 0.0135,
      "reward": 0.35337240993976593,
      "reward_std": 0.14353149756789207,
      "reward_std_all": 0.12546059116721153,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9238944947719574,
      "rewards/forensics_tempsamp_iou_reward": 0.35337240993976593,
      "step": 480
    },
    {
      "completion_length": 22.40625,
      "epoch": 1.2796271637816246,
      "grad_norm": 5.967645168304443,
      "kl": 0.345703125,
      "learning_rate": 6.800000000000001e-07,
      "loss": 0.0138,
      "reward": 0.1886606588959694,
      "reward_std": 0.0680063720792532,
      "reward_std_all": 0.06912779435515404,
      "rewards-max/forensics_tempsamp_iou_reward": 0.680706799030304,
      "rewards/forensics_tempsamp_iou_reward": 0.1886606588959694,
      "step": 481
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.2822902796271638,
      "grad_norm": 7.1854729652404785,
      "kl": 0.55859375,
      "learning_rate": 6.793333333333333e-07,
      "loss": 0.0224,
      "reward": 0.2151215299963951,
      "reward_std": 0.11073913797736168,
      "reward_std_all": 0.10979267582297325,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6405660510063171,
      "rewards/forensics_tempsamp_iou_reward": 0.2151215299963951,
      "step": 482
    },
    {
      "completion_length": 19.6875,
      "epoch": 1.284953395472703,
      "grad_norm": 5.173044681549072,
      "kl": 0.46484375,
      "learning_rate": 6.786666666666667e-07,
      "loss": 0.0186,
      "reward": 0.20430488884449005,
      "reward_std": 0.0624725129455328,
      "reward_std_all": 0.0824400782585144,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8844175040721893,
      "rewards/forensics_tempsamp_iou_reward": 0.20430488884449005,
      "step": 483
    },
    {
      "completion_length": 17.5,
      "epoch": 1.2876165113182423,
      "grad_norm": 5.2609405517578125,
      "kl": 0.52783203125,
      "learning_rate": 6.78e-07,
      "loss": 0.0211,
      "reward": 0.3535127192735672,
      "reward_std": 0.08737220242619514,
      "reward_std_all": 0.11399056762456894,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8482142686843872,
      "rewards/forensics_tempsamp_iou_reward": 0.3535127192735672,
      "step": 484
    },
    {
      "completion_length": 25.203125,
      "epoch": 1.2902796271637818,
      "grad_norm": 5.274218559265137,
      "kl": 0.4775390625,
      "learning_rate": 6.773333333333334e-07,
      "loss": 0.0192,
      "reward": 0.1693267971277237,
      "reward_std": 0.11395034193992615,
      "reward_std_all": 0.12447234988212585,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8543965816497803,
      "rewards/forensics_tempsamp_iou_reward": 0.1693267971277237,
      "step": 485
    },
    {
      "completion_length": 24.359375,
      "epoch": 1.2929427430093208,
      "grad_norm": 5.1508564949035645,
      "kl": 0.26806640625,
      "learning_rate": 6.766666666666666e-07,
      "loss": 0.0107,
      "reward": 0.24792737513780594,
      "reward_std": 0.04908360727131367,
      "reward_std_all": 0.04599047265946865,
      "rewards-max/forensics_tempsamp_iou_reward": 0.952359676361084,
      "rewards/forensics_tempsamp_iou_reward": 0.24792737513780594,
      "step": 486
    },
    {
      "completion_length": 19.984375,
      "epoch": 1.2956058588548602,
      "grad_norm": 6.902978897094727,
      "kl": 0.47265625,
      "learning_rate": 6.76e-07,
      "loss": 0.019,
      "reward": 0.18561211228370667,
      "reward_std": 0.05565165542066097,
      "reward_std_all": 0.07821016013622284,
      "rewards-max/forensics_tempsamp_iou_reward": 0.555085077881813,
      "rewards/forensics_tempsamp_iou_reward": 0.18561211228370667,
      "step": 487
    },
    {
      "completion_length": 24.078125,
      "epoch": 1.2982689747003995,
      "grad_norm": 6.755097389221191,
      "kl": 0.705078125,
      "learning_rate": 6.753333333333333e-07,
      "loss": 0.0283,
      "reward": 0.24935632199048996,
      "reward_std": 0.11093607544898987,
      "reward_std_all": 0.12753523513674736,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9392410516738892,
      "rewards/forensics_tempsamp_iou_reward": 0.24935632199048996,
      "step": 488
    },
    {
      "completion_length": 20.4375,
      "epoch": 1.3009320905459387,
      "grad_norm": 4.786919116973877,
      "kl": 0.560546875,
      "learning_rate": 6.746666666666666e-07,
      "loss": 0.0224,
      "reward": 0.20178872346878052,
      "reward_std": 0.08332698047161102,
      "reward_std_all": 0.0922096036374569,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7456761002540588,
      "rewards/forensics_tempsamp_iou_reward": 0.20178872346878052,
      "step": 489
    },
    {
      "completion_length": 21.578125,
      "epoch": 1.303595206391478,
      "grad_norm": 5.702413558959961,
      "kl": 0.4443359375,
      "learning_rate": 6.74e-07,
      "loss": 0.0177,
      "reward": 0.21764517575502396,
      "reward_std": 0.08303960412740707,
      "reward_std_all": 0.08324713632464409,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8169805109500885,
      "rewards/forensics_tempsamp_iou_reward": 0.21764517575502396,
      "step": 490
    },
    {
      "completion_length": 17.0,
      "epoch": 1.3062583222370172,
      "grad_norm": 6.002793312072754,
      "kl": 0.5498046875,
      "learning_rate": 6.733333333333333e-07,
      "loss": 0.022,
      "reward": 0.22520878911018372,
      "reward_std": 0.11325733736157417,
      "reward_std_all": 0.11493607610464096,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9008758068084717,
      "rewards/forensics_tempsamp_iou_reward": 0.22520878911018372,
      "step": 491
    },
    {
      "completion_length": 18.453125,
      "epoch": 1.3089214380825567,
      "grad_norm": 6.347090244293213,
      "kl": 0.8037109375,
      "learning_rate": 6.726666666666666e-07,
      "loss": 0.0321,
      "reward": 0.1967754364013672,
      "reward_std": 0.10615004226565361,
      "reward_std_all": 0.1023293249309063,
      "rewards-max/forensics_tempsamp_iou_reward": 0.883071094751358,
      "rewards/forensics_tempsamp_iou_reward": 0.1967754364013672,
      "step": 492
    },
    {
      "completion_length": 18.921875,
      "epoch": 1.311584553928096,
      "grad_norm": 7.055381774902344,
      "kl": 0.353515625,
      "learning_rate": 6.72e-07,
      "loss": 0.0141,
      "reward": 0.194037064909935,
      "reward_std": 0.08332270383834839,
      "reward_std_all": 0.09276725351810455,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6777137219905853,
      "rewards/forensics_tempsamp_iou_reward": 0.194037064909935,
      "step": 493
    },
    {
      "completion_length": 20.171875,
      "epoch": 1.3142476697736352,
      "grad_norm": 16.25654411315918,
      "kl": 0.611328125,
      "learning_rate": 6.713333333333333e-07,
      "loss": 0.0245,
      "reward": 0.2137734293937683,
      "reward_std": 0.09316081553697586,
      "reward_std_all": 0.09206695854663849,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5892791599035263,
      "rewards/forensics_tempsamp_iou_reward": 0.2137734293937683,
      "step": 494
    },
    {
      "completion_length": 18.109375,
      "epoch": 1.3169107856191744,
      "grad_norm": 7.078381538391113,
      "kl": 0.46484375,
      "learning_rate": 6.706666666666666e-07,
      "loss": 0.0186,
      "reward": 0.193692646920681,
      "reward_std": 0.10646744817495346,
      "reward_std_all": 0.09500968083739281,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8156165182590485,
      "rewards/forensics_tempsamp_iou_reward": 0.193692646920681,
      "step": 495
    },
    {
      "completion_length": 20.203125,
      "epoch": 1.3195739014647137,
      "grad_norm": 5.963439464569092,
      "kl": 0.59375,
      "learning_rate": 6.7e-07,
      "loss": 0.0237,
      "reward": 0.16712362319231033,
      "reward_std": 0.0795624740421772,
      "reward_std_all": 0.10040086135268211,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6723565608263016,
      "rewards/forensics_tempsamp_iou_reward": 0.16712362319231033,
      "step": 496
    },
    {
      "completion_length": 17.515625,
      "epoch": 1.3222370173102531,
      "grad_norm": 4.786433219909668,
      "kl": 0.4697265625,
      "learning_rate": 6.693333333333333e-07,
      "loss": 0.0188,
      "reward": 0.259251169860363,
      "reward_std": 0.07729493081569672,
      "reward_std_all": 0.08820775896310806,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8867136240005493,
      "rewards/forensics_tempsamp_iou_reward": 0.259251169860363,
      "step": 497
    },
    {
      "completion_length": 20.328125,
      "epoch": 1.3249001331557924,
      "grad_norm": 5.592378616333008,
      "kl": 0.5302734375,
      "learning_rate": 6.686666666666666e-07,
      "loss": 0.0212,
      "reward": 0.24314457178115845,
      "reward_std": 0.11155128479003906,
      "reward_std_all": 0.10312367230653763,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9602816998958588,
      "rewards/forensics_tempsamp_iou_reward": 0.24314457178115845,
      "step": 498
    },
    {
      "completion_length": 22.609375,
      "epoch": 1.3275632490013316,
      "grad_norm": 5.189847946166992,
      "kl": 0.3349609375,
      "learning_rate": 6.68e-07,
      "loss": 0.0134,
      "reward": 0.2811911776661873,
      "reward_std": 0.08496584743261337,
      "reward_std_all": 0.09077567234635353,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9625099301338196,
      "rewards/forensics_tempsamp_iou_reward": 0.2811911776661873,
      "step": 499
    },
    {
      "completion_length": 19.5,
      "epoch": 1.3302263648468708,
      "grad_norm": 7.787737846374512,
      "kl": 0.658203125,
      "learning_rate": 6.673333333333334e-07,
      "loss": 0.0262,
      "reward": 0.279164656996727,
      "reward_std": 0.07293536700308323,
      "reward_std_all": 0.09239024668931961,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8925148546695709,
      "rewards/forensics_tempsamp_iou_reward": 0.279164656996727,
      "step": 500
    },
    {
      "completion_length": 21.703125,
      "epoch": 1.33288948069241,
      "grad_norm": 6.100408554077148,
      "kl": 0.2783203125,
      "learning_rate": 6.666666666666666e-07,
      "loss": 0.0111,
      "reward": 0.24345964193344116,
      "reward_std": 0.07782384008169174,
      "reward_std_all": 0.08209022134542465,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8063168823719025,
      "rewards/forensics_tempsamp_iou_reward": 0.24345964193344116,
      "step": 501
    },
    {
      "completion_length": 17.453125,
      "epoch": 1.3355525965379493,
      "grad_norm": 5.401427745819092,
      "kl": 0.38916015625,
      "learning_rate": 6.66e-07,
      "loss": 0.0155,
      "reward": 0.310065358877182,
      "reward_std": 0.11557039245963097,
      "reward_std_all": 0.11424039304256439,
      "rewards-max/forensics_tempsamp_iou_reward": 0.873701423406601,
      "rewards/forensics_tempsamp_iou_reward": 0.310065358877182,
      "step": 502
    },
    {
      "completion_length": 19.125,
      "epoch": 1.3382157123834886,
      "grad_norm": 7.420879364013672,
      "kl": 0.580078125,
      "learning_rate": 6.653333333333333e-07,
      "loss": 0.0232,
      "reward": 0.33233898878097534,
      "reward_std": 0.12134146690368652,
      "reward_std_all": 0.1295616365969181,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8776615858078003,
      "rewards/forensics_tempsamp_iou_reward": 0.33233898878097534,
      "step": 503
    },
    {
      "completion_length": 13.75,
      "epoch": 1.340878828229028,
      "grad_norm": 5.886183738708496,
      "kl": 0.3642578125,
      "learning_rate": 6.646666666666666e-07,
      "loss": 0.0146,
      "reward": 0.24647586047649384,
      "reward_std": 0.07814867049455643,
      "reward_std_all": 0.09275886416435242,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9267506897449493,
      "rewards/forensics_tempsamp_iou_reward": 0.24647586047649384,
      "step": 504
    },
    {
      "completion_length": 15.390625,
      "epoch": 1.3435419440745673,
      "grad_norm": 7.623484134674072,
      "kl": 0.3515625,
      "learning_rate": 6.64e-07,
      "loss": 0.0141,
      "reward": 0.15110071748495102,
      "reward_std": 0.09014435857534409,
      "reward_std_all": 0.08867191523313522,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5615248382091522,
      "rewards/forensics_tempsamp_iou_reward": 0.15110071748495102,
      "step": 505
    },
    {
      "completion_length": 20.921875,
      "epoch": 1.3462050599201065,
      "grad_norm": 5.017059326171875,
      "kl": 0.2421875,
      "learning_rate": 6.633333333333333e-07,
      "loss": 0.0097,
      "reward": 0.3111255168914795,
      "reward_std": 0.0994194746017456,
      "reward_std_all": 0.12289538234472275,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8837653994560242,
      "rewards/forensics_tempsamp_iou_reward": 0.3111255168914795,
      "step": 506
    },
    {
      "completion_length": 20.265625,
      "epoch": 1.3488681757656458,
      "grad_norm": 5.696448802947998,
      "kl": 0.255859375,
      "learning_rate": 6.626666666666666e-07,
      "loss": 0.0102,
      "reward": 0.18357903510332108,
      "reward_std": 0.07644686847925186,
      "reward_std_all": 0.08114631101489067,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8326663672924042,
      "rewards/forensics_tempsamp_iou_reward": 0.18357903510332108,
      "step": 507
    },
    {
      "completion_length": 15.171875,
      "epoch": 1.351531291611185,
      "grad_norm": 4.698728084564209,
      "kl": 0.35498046875,
      "learning_rate": 6.62e-07,
      "loss": 0.0142,
      "reward": 0.1658986210823059,
      "reward_std": 0.04241778329014778,
      "reward_std_all": 0.039560794830322266,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7002086043357849,
      "rewards/forensics_tempsamp_iou_reward": 0.1658986210823059,
      "step": 508
    },
    {
      "completion_length": 19.078125,
      "epoch": 1.3541944074567245,
      "grad_norm": 6.256978988647461,
      "kl": 0.4306640625,
      "learning_rate": 6.613333333333333e-07,
      "loss": 0.0172,
      "reward": 0.2740078866481781,
      "reward_std": 0.10600528120994568,
      "reward_std_all": 0.10490641742944717,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8991618752479553,
      "rewards/forensics_tempsamp_iou_reward": 0.2740078866481781,
      "step": 509
    },
    {
      "completion_length": 18.359375,
      "epoch": 1.3568575233022637,
      "grad_norm": 5.85725736618042,
      "kl": 0.4765625,
      "learning_rate": 6.606666666666666e-07,
      "loss": 0.0191,
      "reward": 0.15020178258419037,
      "reward_std": 0.07178844511508942,
      "reward_std_all": 0.07019487768411636,
      "rewards-max/forensics_tempsamp_iou_reward": 0.690851628780365,
      "rewards/forensics_tempsamp_iou_reward": 0.15020178258419037,
      "step": 510
    },
    {
      "completion_length": 25.75,
      "epoch": 1.359520639147803,
      "grad_norm": 6.156455993652344,
      "kl": 0.296875,
      "learning_rate": 6.6e-07,
      "loss": 0.0119,
      "reward": 0.28658391535282135,
      "reward_std": 0.11013500764966011,
      "reward_std_all": 0.1098414734005928,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8165627121925354,
      "rewards/forensics_tempsamp_iou_reward": 0.28658391535282135,
      "step": 511
    },
    {
      "completion_length": 33.125,
      "epoch": 1.3621837549933422,
      "grad_norm": 4.749648094177246,
      "kl": 0.5126953125,
      "learning_rate": 6.593333333333333e-07,
      "loss": 0.0205,
      "reward": 0.14673305302858353,
      "reward_std": 0.06259455904364586,
      "reward_std_all": 0.0813978761434555,
      "rewards-max/forensics_tempsamp_iou_reward": 0.3721012622117996,
      "rewards/forensics_tempsamp_iou_reward": 0.14673305302858353,
      "step": 512
    },
    {
      "completion_length": 27.453125,
      "epoch": 1.3648468708388815,
      "grad_norm": 5.206135272979736,
      "kl": 0.447265625,
      "learning_rate": 6.586666666666666e-07,
      "loss": 0.0179,
      "reward": 0.1636016145348549,
      "reward_std": 0.05817430280148983,
      "reward_std_all": 0.07020674645900726,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6273365914821625,
      "rewards/forensics_tempsamp_iou_reward": 0.1636016145348549,
      "step": 513
    },
    {
      "completion_length": 18.78125,
      "epoch": 1.3675099866844207,
      "grad_norm": 6.125884532928467,
      "kl": 0.21728515625,
      "learning_rate": 6.58e-07,
      "loss": 0.0087,
      "reward": 0.3291543275117874,
      "reward_std": 0.09783929958939552,
      "reward_std_all": 0.10371240228414536,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8861525654792786,
      "rewards/forensics_tempsamp_iou_reward": 0.3291543275117874,
      "step": 514
    },
    {
      "completion_length": 21.1875,
      "epoch": 1.37017310252996,
      "grad_norm": 5.353962421417236,
      "kl": 0.24462890625,
      "learning_rate": 6.573333333333333e-07,
      "loss": 0.0098,
      "reward": 0.3196937069296837,
      "reward_std": 0.08561785146594048,
      "reward_std_all": 0.08668551221489906,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9646074771881104,
      "rewards/forensics_tempsamp_iou_reward": 0.3196937069296837,
      "step": 515
    },
    {
      "completion_length": 17.390625,
      "epoch": 1.3728362183754994,
      "grad_norm": 5.298189163208008,
      "kl": 0.2548828125,
      "learning_rate": 6.566666666666666e-07,
      "loss": 0.0102,
      "reward": 0.289703868329525,
      "reward_std": 0.07139403745532036,
      "reward_std_all": 0.08648817986249924,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6884417831897736,
      "rewards/forensics_tempsamp_iou_reward": 0.289703868329525,
      "step": 516
    },
    {
      "completion_length": 16.0625,
      "epoch": 1.3754993342210386,
      "grad_norm": 6.157247066497803,
      "kl": 0.26123046875,
      "learning_rate": 6.56e-07,
      "loss": 0.0104,
      "reward": 0.4615631401538849,
      "reward_std": 0.1546260565519333,
      "reward_std_all": 0.14023562893271446,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9680847525596619,
      "rewards/forensics_tempsamp_iou_reward": 0.4615631401538849,
      "step": 517
    },
    {
      "completion_length": 19.328125,
      "epoch": 1.378162450066578,
      "grad_norm": 5.780140399932861,
      "kl": 0.556640625,
      "learning_rate": 6.553333333333333e-07,
      "loss": 0.0224,
      "reward": 0.24629535526037216,
      "reward_std": 0.08997569605708122,
      "reward_std_all": 0.09221499040722847,
      "rewards-max/forensics_tempsamp_iou_reward": 0.876514196395874,
      "rewards/forensics_tempsamp_iou_reward": 0.24629535526037216,
      "step": 518
    },
    {
      "completion_length": 19.734375,
      "epoch": 1.3808255659121171,
      "grad_norm": 6.858870983123779,
      "kl": 0.390625,
      "learning_rate": 6.546666666666665e-07,
      "loss": 0.0156,
      "reward": 0.24143756926059723,
      "reward_std": 0.05281079187989235,
      "reward_std_all": 0.06295040994882584,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7303791791200638,
      "rewards/forensics_tempsamp_iou_reward": 0.24143756926059723,
      "step": 519
    },
    {
      "completion_length": 22.359375,
      "epoch": 1.3834886817576564,
      "grad_norm": 4.761932373046875,
      "kl": 0.51904296875,
      "learning_rate": 6.54e-07,
      "loss": 0.0208,
      "reward": 0.2378655970096588,
      "reward_std": 0.1304846704006195,
      "reward_std_all": 0.1294749528169632,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9742820560932159,
      "rewards/forensics_tempsamp_iou_reward": 0.2378655970096588,
      "step": 520
    },
    {
      "completion_length": 17.4375,
      "epoch": 1.3861517976031958,
      "grad_norm": 8.188858985900879,
      "kl": 0.31201171875,
      "learning_rate": 6.533333333333333e-07,
      "loss": 0.0125,
      "reward": 0.26143424957990646,
      "reward_std": 0.05548262596130371,
      "reward_std_all": 0.09795943647623062,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9103616774082184,
      "rewards/forensics_tempsamp_iou_reward": 0.26143424957990646,
      "step": 521
    },
    {
      "completion_length": 21.9375,
      "epoch": 1.388814913448735,
      "grad_norm": 6.1045002937316895,
      "kl": 0.6044921875,
      "learning_rate": 6.526666666666666e-07,
      "loss": 0.0242,
      "reward": 0.18518120795488358,
      "reward_std": 0.07442908734083176,
      "reward_std_all": 0.06928183138370514,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7007333189249039,
      "rewards/forensics_tempsamp_iou_reward": 0.18518120795488358,
      "step": 522
    },
    {
      "completion_length": 27.015625,
      "epoch": 1.3914780292942743,
      "grad_norm": 5.362558364868164,
      "kl": 0.375,
      "learning_rate": 6.52e-07,
      "loss": 0.015,
      "reward": 0.25969310849905014,
      "reward_std": 0.09343784675002098,
      "reward_std_all": 0.09776977077126503,
      "rewards-max/forensics_tempsamp_iou_reward": 0.892985612154007,
      "rewards/forensics_tempsamp_iou_reward": 0.25969310849905014,
      "step": 523
    },
    {
      "completion_length": 21.234375,
      "epoch": 1.3941411451398136,
      "grad_norm": 6.168169975280762,
      "kl": 0.3916015625,
      "learning_rate": 6.513333333333333e-07,
      "loss": 0.0157,
      "reward": 0.3070301115512848,
      "reward_std": 0.08615870587527752,
      "reward_std_all": 0.08170942217111588,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9826104938983917,
      "rewards/forensics_tempsamp_iou_reward": 0.3070301115512848,
      "step": 524
    },
    {
      "completion_length": 23.140625,
      "epoch": 1.3968042609853528,
      "grad_norm": 6.342609405517578,
      "kl": 0.39892578125,
      "learning_rate": 6.506666666666666e-07,
      "loss": 0.016,
      "reward": 0.3568579852581024,
      "reward_std": 0.12816347181797028,
      "reward_std_all": 0.14772824943065643,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9179826080799103,
      "rewards/forensics_tempsamp_iou_reward": 0.3568579852581024,
      "step": 525
    },
    {
      "completion_length": 28.515625,
      "epoch": 1.399467376830892,
      "grad_norm": 4.893665790557861,
      "kl": 0.62109375,
      "learning_rate": 6.5e-07,
      "loss": 0.0248,
      "reward": 0.3983703702688217,
      "reward_std": 0.10630897805094719,
      "reward_std_all": 0.10136960074305534,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9976851940155029,
      "rewards/forensics_tempsamp_iou_reward": 0.3983703702688217,
      "step": 526
    },
    {
      "completion_length": 20.84375,
      "epoch": 1.4021304926764313,
      "grad_norm": 6.446109294891357,
      "kl": 0.2001953125,
      "learning_rate": 6.493333333333333e-07,
      "loss": 0.008,
      "reward": 0.35181663930416107,
      "reward_std": 0.02852474804967642,
      "reward_std_all": 0.030914909206330776,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9416520893573761,
      "rewards/forensics_tempsamp_iou_reward": 0.35181663930416107,
      "step": 527
    },
    {
      "completion_length": 31.9375,
      "epoch": 1.4047936085219708,
      "grad_norm": 4.761476039886475,
      "kl": 0.2421875,
      "learning_rate": 6.486666666666666e-07,
      "loss": 0.0097,
      "reward": 0.36800652742385864,
      "reward_std": 0.1084027960896492,
      "reward_std_all": 0.11228788644075394,
      "rewards-max/forensics_tempsamp_iou_reward": 0.942340612411499,
      "rewards/forensics_tempsamp_iou_reward": 0.36800652742385864,
      "step": 528
    },
    {
      "completion_length": 28.484375,
      "epoch": 1.40745672436751,
      "grad_norm": 7.357893466949463,
      "kl": 0.5166015625,
      "learning_rate": 6.48e-07,
      "loss": 0.0207,
      "reward": 0.286545068025589,
      "reward_std": 0.058877596631646156,
      "reward_std_all": 0.07824821397662163,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9812739491462708,
      "rewards/forensics_tempsamp_iou_reward": 0.286545068025589,
      "step": 529
    },
    {
      "completion_length": 17.25,
      "epoch": 1.4101198402130493,
      "grad_norm": 10.647322654724121,
      "kl": 0.7265625,
      "learning_rate": 6.473333333333333e-07,
      "loss": 0.0291,
      "reward": 0.3594081699848175,
      "reward_std": 0.10529842227697372,
      "reward_std_all": 0.1116204522550106,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9819791615009308,
      "rewards/forensics_tempsamp_iou_reward": 0.3594081699848175,
      "step": 530
    },
    {
      "completion_length": 19.0,
      "epoch": 1.4127829560585885,
      "grad_norm": 5.223423480987549,
      "kl": 0.40087890625,
      "learning_rate": 6.466666666666666e-07,
      "loss": 0.0161,
      "reward": 0.3683735430240631,
      "reward_std": 0.06496674381196499,
      "reward_std_all": 0.0790880024433136,
      "rewards-max/forensics_tempsamp_iou_reward": 0.961969792842865,
      "rewards/forensics_tempsamp_iou_reward": 0.3683735430240631,
      "step": 531
    },
    {
      "completion_length": 24.578125,
      "epoch": 1.4154460719041277,
      "grad_norm": 6.882837772369385,
      "kl": 0.421875,
      "learning_rate": 6.46e-07,
      "loss": 0.0169,
      "reward": 0.2454683557152748,
      "reward_std": 0.060928359627723694,
      "reward_std_all": 0.06372667476534843,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8238179981708527,
      "rewards/forensics_tempsamp_iou_reward": 0.2454683557152748,
      "step": 532
    },
    {
      "completion_length": 18.046875,
      "epoch": 1.4181091877496672,
      "grad_norm": 5.461915493011475,
      "kl": 0.56396484375,
      "learning_rate": 6.453333333333333e-07,
      "loss": 0.0225,
      "reward": 0.44767501950263977,
      "reward_std": 0.11796274036169052,
      "reward_std_all": 0.10945058614015579,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9879913926124573,
      "rewards/forensics_tempsamp_iou_reward": 0.44767501950263977,
      "step": 533
    },
    {
      "completion_length": 18.46875,
      "epoch": 1.4207723035952065,
      "grad_norm": 5.902198791503906,
      "kl": 0.3505859375,
      "learning_rate": 6.446666666666666e-07,
      "loss": 0.014,
      "reward": 0.2896662950515747,
      "reward_std": 0.12909718230366707,
      "reward_std_all": 0.11615778505802155,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9779162704944611,
      "rewards/forensics_tempsamp_iou_reward": 0.2896662950515747,
      "step": 534
    },
    {
      "completion_length": 16.421875,
      "epoch": 1.4234354194407457,
      "grad_norm": 5.750749111175537,
      "kl": 0.26416015625,
      "learning_rate": 6.44e-07,
      "loss": 0.0105,
      "reward": 0.20298050343990326,
      "reward_std": 0.0731613477692008,
      "reward_std_all": 0.0722789941355586,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6156515032052994,
      "rewards/forensics_tempsamp_iou_reward": 0.20298050343990326,
      "step": 535
    },
    {
      "completion_length": 20.59375,
      "epoch": 1.426098535286285,
      "grad_norm": 5.830048561096191,
      "kl": 0.24169921875,
      "learning_rate": 6.433333333333332e-07,
      "loss": 0.0097,
      "reward": 0.389102965593338,
      "reward_std": 0.046617924235761166,
      "reward_std_all": 0.06727512739598751,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9131971001625061,
      "rewards/forensics_tempsamp_iou_reward": 0.389102965593338,
      "step": 536
    },
    {
      "completion_length": 30.875,
      "epoch": 1.4287616511318242,
      "grad_norm": 6.215404033660889,
      "kl": 0.5107421875,
      "learning_rate": 6.426666666666667e-07,
      "loss": 0.0204,
      "reward": 0.2711132988333702,
      "reward_std": 0.07476985082030296,
      "reward_std_all": 0.07550949603319168,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9884541034698486,
      "rewards/forensics_tempsamp_iou_reward": 0.2711132988333702,
      "step": 537
    },
    {
      "completion_length": 22.578125,
      "epoch": 1.4314247669773636,
      "grad_norm": 6.541515827178955,
      "kl": 0.419921875,
      "learning_rate": 6.42e-07,
      "loss": 0.0168,
      "reward": 0.22450105845928192,
      "reward_std": 0.08748237043619156,
      "reward_std_all": 0.08116490021348,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9052780568599701,
      "rewards/forensics_tempsamp_iou_reward": 0.22450105845928192,
      "step": 538
    },
    {
      "completion_length": 29.8125,
      "epoch": 1.4340878828229027,
      "grad_norm": 6.497073650360107,
      "kl": 0.5341796875,
      "learning_rate": 6.413333333333333e-07,
      "loss": 0.0214,
      "reward": 0.19199447333812714,
      "reward_std": 0.06493248976767063,
      "reward_std_all": 0.07441268116235733,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6529704332351685,
      "rewards/forensics_tempsamp_iou_reward": 0.19199447333812714,
      "step": 539
    },
    {
      "completion_length": 23.96875,
      "epoch": 1.4367509986684421,
      "grad_norm": 7.783785820007324,
      "kl": 0.671875,
      "learning_rate": 6.406666666666667e-07,
      "loss": 0.0269,
      "reward": 0.29454459249973297,
      "reward_std": 0.07263725064694881,
      "reward_std_all": 0.06937301531434059,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9857936501502991,
      "rewards/forensics_tempsamp_iou_reward": 0.29454459249973297,
      "step": 540
    },
    {
      "completion_length": 33.15625,
      "epoch": 1.4394141145139814,
      "grad_norm": 5.570154666900635,
      "kl": 0.5,
      "learning_rate": 6.4e-07,
      "loss": 0.02,
      "reward": 0.3138798177242279,
      "reward_std": 0.12094668671488762,
      "reward_std_all": 0.13156630471348763,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8852513432502747,
      "rewards/forensics_tempsamp_iou_reward": 0.3138798177242279,
      "step": 541
    },
    {
      "completion_length": 19.109375,
      "epoch": 1.4420772303595206,
      "grad_norm": 7.311975002288818,
      "kl": 0.5625,
      "learning_rate": 6.393333333333333e-07,
      "loss": 0.0225,
      "reward": 0.25435470044612885,
      "reward_std": 0.07046302407979965,
      "reward_std_all": 0.07565318793058395,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9659090936183929,
      "rewards/forensics_tempsamp_iou_reward": 0.25435470044612885,
      "step": 542
    },
    {
      "completion_length": 40.734375,
      "epoch": 1.4447403462050599,
      "grad_norm": 27.409454345703125,
      "kl": 1.728515625,
      "learning_rate": 6.386666666666667e-07,
      "loss": 0.0694,
      "reward": 0.31986677646636963,
      "reward_std": 0.125005804002285,
      "reward_std_all": 0.11768218129873276,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9124472439289093,
      "rewards/forensics_tempsamp_iou_reward": 0.31986677646636963,
      "step": 543
    },
    {
      "completion_length": 19.796875,
      "epoch": 1.447403462050599,
      "grad_norm": 5.693724632263184,
      "kl": 0.396484375,
      "learning_rate": 6.38e-07,
      "loss": 0.0159,
      "reward": 0.26784100383520126,
      "reward_std": 0.06608465313911438,
      "reward_std_all": 0.08293364197015762,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8958751857280731,
      "rewards/forensics_tempsamp_iou_reward": 0.26784100383520126,
      "step": 544
    },
    {
      "completion_length": 31.671875,
      "epoch": 1.4500665778961386,
      "grad_norm": 5.719252586364746,
      "kl": 0.53515625,
      "learning_rate": 6.373333333333333e-07,
      "loss": 0.0214,
      "reward": 0.2611541897058487,
      "reward_std": 0.10152711346745491,
      "reward_std_all": 0.09038807824254036,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9242202639579773,
      "rewards/forensics_tempsamp_iou_reward": 0.2611541897058487,
      "step": 545
    },
    {
      "completion_length": 28.375,
      "epoch": 1.4527296937416778,
      "grad_norm": 7.108187675476074,
      "kl": 0.44140625,
      "learning_rate": 6.366666666666667e-07,
      "loss": 0.0177,
      "reward": 0.229653462767601,
      "reward_std": 0.09138578549027443,
      "reward_std_all": 0.09085159003734589,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7021027207374573,
      "rewards/forensics_tempsamp_iou_reward": 0.229653462767601,
      "step": 546
    },
    {
      "completion_length": 26.5625,
      "epoch": 1.455392809587217,
      "grad_norm": 7.677452087402344,
      "kl": 0.560546875,
      "learning_rate": 6.36e-07,
      "loss": 0.0224,
      "reward": 0.2771235406398773,
      "reward_std": 0.10589242726564407,
      "reward_std_all": 0.11771027371287346,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7302867472171783,
      "rewards/forensics_tempsamp_iou_reward": 0.2771235406398773,
      "step": 547
    },
    {
      "completion_length": 29.734375,
      "epoch": 1.4580559254327563,
      "grad_norm": 5.645936489105225,
      "kl": 0.5556640625,
      "learning_rate": 6.353333333333333e-07,
      "loss": 0.0223,
      "reward": 0.23306165635585785,
      "reward_std": 0.06119796261191368,
      "reward_std_all": 0.06073470413684845,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7318081110715866,
      "rewards/forensics_tempsamp_iou_reward": 0.23306165635585785,
      "step": 548
    },
    {
      "completion_length": 36.921875,
      "epoch": 1.4607190412782955,
      "grad_norm": 5.727203845977783,
      "kl": 0.302734375,
      "learning_rate": 6.346666666666666e-07,
      "loss": 0.0121,
      "reward": 0.29712578654289246,
      "reward_std": 0.0970709640532732,
      "reward_std_all": 0.09888429567217827,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7513200640678406,
      "rewards/forensics_tempsamp_iou_reward": 0.29712578654289246,
      "step": 549
    },
    {
      "completion_length": 40.171875,
      "epoch": 1.463382157123835,
      "grad_norm": 6.6802659034729,
      "kl": 0.365234375,
      "learning_rate": 6.34e-07,
      "loss": 0.0146,
      "reward": 0.21199381351470947,
      "reward_std": 0.13789447396993637,
      "reward_std_all": 0.13104842603206635,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8754940629005432,
      "rewards/forensics_tempsamp_iou_reward": 0.21199381351470947,
      "step": 550
    },
    {
      "completion_length": 29.5625,
      "epoch": 1.466045272969374,
      "grad_norm": 6.111229419708252,
      "kl": 0.4052734375,
      "learning_rate": 6.333333333333332e-07,
      "loss": 0.0162,
      "reward": 0.2052636593580246,
      "reward_std": 0.059358106926083565,
      "reward_std_all": 0.07071784697473049,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8903090357780457,
      "rewards/forensics_tempsamp_iou_reward": 0.2052636593580246,
      "step": 551
    },
    {
      "completion_length": 37.546875,
      "epoch": 1.4687083888149135,
      "grad_norm": 4.875775337219238,
      "kl": 0.84765625,
      "learning_rate": 6.326666666666667e-07,
      "loss": 0.0338,
      "reward": 0.24247153103351593,
      "reward_std": 0.12696093320846558,
      "reward_std_all": 0.12457024306058884,
      "rewards-max/forensics_tempsamp_iou_reward": 0.797802209854126,
      "rewards/forensics_tempsamp_iou_reward": 0.24247153103351593,
      "step": 552
    },
    {
      "completion_length": 36.921875,
      "epoch": 1.4713715046604527,
      "grad_norm": 5.35861349105835,
      "kl": 0.55078125,
      "learning_rate": 6.319999999999999e-07,
      "loss": 0.022,
      "reward": 0.2179807350039482,
      "reward_std": 0.12202463671565056,
      "reward_std_all": 0.13447082787752151,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7675173580646515,
      "rewards/forensics_tempsamp_iou_reward": 0.2179807350039482,
      "step": 553
    },
    {
      "completion_length": 35.3125,
      "epoch": 1.474034620505992,
      "grad_norm": 5.589230537414551,
      "kl": 0.3896484375,
      "learning_rate": 6.313333333333333e-07,
      "loss": 0.0156,
      "reward": 0.3263617157936096,
      "reward_std": 0.12500189244747162,
      "reward_std_all": 0.13665150851011276,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9096863269805908,
      "rewards/forensics_tempsamp_iou_reward": 0.3263617157936096,
      "step": 554
    },
    {
      "completion_length": 32.84375,
      "epoch": 1.4766977363515312,
      "grad_norm": 6.21517276763916,
      "kl": 0.3212890625,
      "learning_rate": 6.306666666666666e-07,
      "loss": 0.0129,
      "reward": 0.2614961862564087,
      "reward_std": 0.09650824218988419,
      "reward_std_all": 0.09526018798351288,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8379361033439636,
      "rewards/forensics_tempsamp_iou_reward": 0.2614961862564087,
      "step": 555
    },
    {
      "completion_length": 21.984375,
      "epoch": 1.4793608521970705,
      "grad_norm": 7.640132427215576,
      "kl": 0.4033203125,
      "learning_rate": 6.3e-07,
      "loss": 0.0161,
      "reward": 0.19520314037799835,
      "reward_std": 0.04498663730919361,
      "reward_std_all": 0.055924300104379654,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4772179573774338,
      "rewards/forensics_tempsamp_iou_reward": 0.19520314037799835,
      "step": 556
    },
    {
      "completion_length": 23.890625,
      "epoch": 1.48202396804261,
      "grad_norm": 5.447958469390869,
      "kl": 0.4140625,
      "learning_rate": 6.293333333333333e-07,
      "loss": 0.0166,
      "reward": 0.20641763508319855,
      "reward_std": 0.060693634673953056,
      "reward_std_all": 0.07035116851329803,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6652933359146118,
      "rewards/forensics_tempsamp_iou_reward": 0.20641763508319855,
      "step": 557
    },
    {
      "completion_length": 34.375,
      "epoch": 1.4846870838881492,
      "grad_norm": 5.556643486022949,
      "kl": 0.541015625,
      "learning_rate": 6.286666666666667e-07,
      "loss": 0.0216,
      "reward": 0.27457092702388763,
      "reward_std": 0.061519309878349304,
      "reward_std_all": 0.07304059341549873,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8907552659511566,
      "rewards/forensics_tempsamp_iou_reward": 0.27457092702388763,
      "step": 558
    },
    {
      "completion_length": 18.015625,
      "epoch": 1.4873501997336884,
      "grad_norm": 8.32532024383545,
      "kl": 0.4013671875,
      "learning_rate": 6.28e-07,
      "loss": 0.0161,
      "reward": 0.37380750477313995,
      "reward_std": 0.08814905025064945,
      "reward_std_all": 0.10010036826133728,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9230700433254242,
      "rewards/forensics_tempsamp_iou_reward": 0.37380750477313995,
      "step": 559
    },
    {
      "completion_length": 23.609375,
      "epoch": 1.4900133155792277,
      "grad_norm": 8.295445442199707,
      "kl": 0.3154296875,
      "learning_rate": 6.273333333333333e-07,
      "loss": 0.0126,
      "reward": 0.3146865516901016,
      "reward_std": 0.08723578229546547,
      "reward_std_all": 0.10885220393538475,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7374669313430786,
      "rewards/forensics_tempsamp_iou_reward": 0.3146865516901016,
      "step": 560
    },
    {
      "completion_length": 23.3125,
      "epoch": 1.492676431424767,
      "grad_norm": 6.3807902336120605,
      "kl": 0.396484375,
      "learning_rate": 6.266666666666667e-07,
      "loss": 0.0158,
      "reward": 0.2523014768958092,
      "reward_std": 0.0835559107363224,
      "reward_std_all": 0.08464137837290764,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6871758103370667,
      "rewards/forensics_tempsamp_iou_reward": 0.2523014768958092,
      "step": 561
    },
    {
      "completion_length": 28.125,
      "epoch": 1.4953395472703064,
      "grad_norm": 4.851255416870117,
      "kl": 0.4423828125,
      "learning_rate": 6.26e-07,
      "loss": 0.0177,
      "reward": 0.3469970375299454,
      "reward_std": 0.14985385164618492,
      "reward_std_all": 0.14973067492246628,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9522438049316406,
      "rewards/forensics_tempsamp_iou_reward": 0.3469970375299454,
      "step": 562
    },
    {
      "completion_length": 34.359375,
      "epoch": 1.4980026631158454,
      "grad_norm": 5.976434707641602,
      "kl": 0.453125,
      "learning_rate": 6.253333333333333e-07,
      "loss": 0.0181,
      "reward": 0.23681126534938812,
      "reward_std": 0.09778909757733345,
      "reward_std_all": 0.09076816216111183,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6914714425802231,
      "rewards/forensics_tempsamp_iou_reward": 0.23681126534938812,
      "step": 563
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.5006657789613849,
      "grad_norm": 7.236340045928955,
      "kl": 0.3251953125,
      "learning_rate": 6.246666666666667e-07,
      "loss": 0.013,
      "reward": 0.3007732182741165,
      "reward_std": 0.09888322278857231,
      "reward_std_all": 0.10876092314720154,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9320958852767944,
      "rewards/forensics_tempsamp_iou_reward": 0.3007732182741165,
      "step": 564
    },
    {
      "completion_length": 21.265625,
      "epoch": 1.503328894806924,
      "grad_norm": 5.724248886108398,
      "kl": 0.4072265625,
      "learning_rate": 6.24e-07,
      "loss": 0.0163,
      "reward": 0.28393978625535965,
      "reward_std": 0.09679250046610832,
      "reward_std_all": 0.08770167827606201,
      "rewards-max/forensics_tempsamp_iou_reward": 0.869907408952713,
      "rewards/forensics_tempsamp_iou_reward": 0.28393978625535965,
      "step": 565
    },
    {
      "completion_length": 17.3125,
      "epoch": 1.5059920106524634,
      "grad_norm": 6.11173677444458,
      "kl": 0.19677734375,
      "learning_rate": 6.233333333333332e-07,
      "loss": 0.0079,
      "reward": 0.40674540400505066,
      "reward_std": 0.07532690465450287,
      "reward_std_all": 0.07997161895036697,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9501020014286041,
      "rewards/forensics_tempsamp_iou_reward": 0.40674540400505066,
      "step": 566
    },
    {
      "completion_length": 16.59375,
      "epoch": 1.5086551264980028,
      "grad_norm": 4.596730709075928,
      "kl": 0.25,
      "learning_rate": 6.226666666666667e-07,
      "loss": 0.01,
      "reward": 0.28450874984264374,
      "reward_std": 0.08464634418487549,
      "reward_std_all": 0.07917356863617897,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9549457728862762,
      "rewards/forensics_tempsamp_iou_reward": 0.28450874984264374,
      "step": 567
    },
    {
      "completion_length": 18.03125,
      "epoch": 1.5113182423435418,
      "grad_norm": 6.197844982147217,
      "kl": 0.35546875,
      "learning_rate": 6.219999999999999e-07,
      "loss": 0.0142,
      "reward": 0.2823279872536659,
      "reward_std": 0.04892656113952398,
      "reward_std_all": 0.04736766219139099,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9075876474380493,
      "rewards/forensics_tempsamp_iou_reward": 0.2823279872536659,
      "step": 568
    },
    {
      "completion_length": 21.8125,
      "epoch": 1.5139813581890813,
      "grad_norm": 5.283036708831787,
      "kl": 0.2685546875,
      "learning_rate": 6.213333333333333e-07,
      "loss": 0.0107,
      "reward": 0.3176160901784897,
      "reward_std": 0.09152441471815109,
      "reward_std_all": 0.10316509380936623,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6909647434949875,
      "rewards/forensics_tempsamp_iou_reward": 0.3176160901784897,
      "step": 569
    },
    {
      "completion_length": 18.0625,
      "epoch": 1.5166444740346205,
      "grad_norm": 4.1150312423706055,
      "kl": 0.2421875,
      "learning_rate": 6.206666666666666e-07,
      "loss": 0.0097,
      "reward": 0.20636522769927979,
      "reward_std": 0.04699096828699112,
      "reward_std_all": 0.044303130358457565,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9390190541744232,
      "rewards/forensics_tempsamp_iou_reward": 0.20636522769927979,
      "step": 570
    },
    {
      "completion_length": 18.796875,
      "epoch": 1.5193075898801598,
      "grad_norm": 7.527156352996826,
      "kl": 0.23974609375,
      "learning_rate": 6.2e-07,
      "loss": 0.0096,
      "reward": 0.42315375804901123,
      "reward_std": 0.08568967133760452,
      "reward_std_all": 0.07910624146461487,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9737269580364227,
      "rewards/forensics_tempsamp_iou_reward": 0.42315375804901123,
      "step": 571
    },
    {
      "completion_length": 27.90625,
      "epoch": 1.521970705725699,
      "grad_norm": 5.629658222198486,
      "kl": 0.2919921875,
      "learning_rate": 6.193333333333332e-07,
      "loss": 0.0117,
      "reward": 0.30267134308815,
      "reward_std": 0.05390850268304348,
      "reward_std_all": 0.05279974825680256,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9803546667098999,
      "rewards/forensics_tempsamp_iou_reward": 0.30267134308815,
      "step": 572
    },
    {
      "completion_length": 14.828125,
      "epoch": 1.5246338215712383,
      "grad_norm": 6.157765865325928,
      "kl": 0.3095703125,
      "learning_rate": 6.186666666666667e-07,
      "loss": 0.0124,
      "reward": 0.3806801736354828,
      "reward_std": 0.07097302749752998,
      "reward_std_all": 0.06599104963243008,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9481812715530396,
      "rewards/forensics_tempsamp_iou_reward": 0.3806801736354828,
      "step": 573
    },
    {
      "completion_length": 25.0,
      "epoch": 1.5272969374167777,
      "grad_norm": 4.477895736694336,
      "kl": 0.4326171875,
      "learning_rate": 6.18e-07,
      "loss": 0.0173,
      "reward": 0.2756252884864807,
      "reward_std": 0.047298112884163857,
      "reward_std_all": 0.056742072105407715,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9625561237335205,
      "rewards/forensics_tempsamp_iou_reward": 0.2756252884864807,
      "step": 574
    },
    {
      "completion_length": 14.46875,
      "epoch": 1.5299600532623168,
      "grad_norm": 5.276923656463623,
      "kl": 0.400390625,
      "learning_rate": 6.173333333333333e-07,
      "loss": 0.016,
      "reward": 0.28724701702594757,
      "reward_std": 0.06765957176685333,
      "reward_std_all": 0.06892969459295273,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9611985683441162,
      "rewards/forensics_tempsamp_iou_reward": 0.28724701702594757,
      "step": 575
    },
    {
      "completion_length": 18.796875,
      "epoch": 1.5326231691078562,
      "grad_norm": 4.894660949707031,
      "kl": 0.31494140625,
      "learning_rate": 6.166666666666667e-07,
      "loss": 0.0126,
      "reward": 0.26517848670482635,
      "reward_std": 0.06955817341804504,
      "reward_std_all": 0.06390263512730598,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8371212184429169,
      "rewards/forensics_tempsamp_iou_reward": 0.26517848670482635,
      "step": 576
    },
    {
      "completion_length": 20.171875,
      "epoch": 1.5352862849533955,
      "grad_norm": 4.520611763000488,
      "kl": 0.529296875,
      "learning_rate": 6.16e-07,
      "loss": 0.0212,
      "reward": 0.268153116106987,
      "reward_std": 0.10513699427247047,
      "reward_std_all": 0.10550708323717117,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9527454674243927,
      "rewards/forensics_tempsamp_iou_reward": 0.268153116106987,
      "step": 577
    },
    {
      "completion_length": 22.125,
      "epoch": 1.5379494007989347,
      "grad_norm": 5.103950023651123,
      "kl": 0.2802734375,
      "learning_rate": 6.153333333333333e-07,
      "loss": 0.0112,
      "reward": 0.3082584738731384,
      "reward_std": 0.07168063335120678,
      "reward_std_all": 0.08271697722375393,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7081797569990158,
      "rewards/forensics_tempsamp_iou_reward": 0.3082584738731384,
      "step": 578
    },
    {
      "completion_length": 17.84375,
      "epoch": 1.5406125166444742,
      "grad_norm": 6.634123802185059,
      "kl": 0.41015625,
      "learning_rate": 6.146666666666667e-07,
      "loss": 0.0164,
      "reward": 0.29556551575660706,
      "reward_std": 0.1289292797446251,
      "reward_std_all": 0.12867309153079987,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9201824069023132,
      "rewards/forensics_tempsamp_iou_reward": 0.29556551575660706,
      "step": 579
    },
    {
      "completion_length": 19.359375,
      "epoch": 1.5432756324900132,
      "grad_norm": 5.344453811645508,
      "kl": 0.2998046875,
      "learning_rate": 6.14e-07,
      "loss": 0.012,
      "reward": 0.31267447769641876,
      "reward_std": 0.0715542621910572,
      "reward_std_all": 0.07666430063545704,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9614756107330322,
      "rewards/forensics_tempsamp_iou_reward": 0.31267447769641876,
      "step": 580
    },
    {
      "completion_length": 24.3125,
      "epoch": 1.5459387483355527,
      "grad_norm": 6.2672438621521,
      "kl": 0.3427734375,
      "learning_rate": 6.133333333333332e-07,
      "loss": 0.0137,
      "reward": 0.2504994124174118,
      "reward_std": 0.09751485660672188,
      "reward_std_all": 0.08888139948248863,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9820981025695801,
      "rewards/forensics_tempsamp_iou_reward": 0.2504994124174118,
      "step": 581
    },
    {
      "completion_length": 21.375,
      "epoch": 1.548601864181092,
      "grad_norm": 6.140757083892822,
      "kl": 0.38671875,
      "learning_rate": 6.126666666666667e-07,
      "loss": 0.0155,
      "reward": 0.29738326370716095,
      "reward_std": 0.07557184062898159,
      "reward_std_all": 0.09908924251794815,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9595713913440704,
      "rewards/forensics_tempsamp_iou_reward": 0.29738326370716095,
      "step": 582
    },
    {
      "completion_length": 20.765625,
      "epoch": 1.5512649800266312,
      "grad_norm": 4.883846759796143,
      "kl": 0.265625,
      "learning_rate": 6.119999999999999e-07,
      "loss": 0.0106,
      "reward": 0.2685750052332878,
      "reward_std": 0.06536102294921875,
      "reward_std_all": 0.0686266366392374,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8167827725410461,
      "rewards/forensics_tempsamp_iou_reward": 0.2685750052332878,
      "step": 583
    },
    {
      "completion_length": 20.125,
      "epoch": 1.5539280958721704,
      "grad_norm": 5.886969089508057,
      "kl": 0.5107421875,
      "learning_rate": 6.113333333333333e-07,
      "loss": 0.0204,
      "reward": 0.1819959208369255,
      "reward_std": 0.10287497937679291,
      "reward_std_all": 0.10367370769381523,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9480880200862885,
      "rewards/forensics_tempsamp_iou_reward": 0.1819959208369255,
      "step": 584
    },
    {
      "completion_length": 19.25,
      "epoch": 1.5565912117177096,
      "grad_norm": 6.1760454177856445,
      "kl": 0.298828125,
      "learning_rate": 6.106666666666666e-07,
      "loss": 0.012,
      "reward": 0.25125014781951904,
      "reward_std": 0.080825075507164,
      "reward_std_all": 0.09111344441771507,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7590277791023254,
      "rewards/forensics_tempsamp_iou_reward": 0.25125014781951904,
      "step": 585
    },
    {
      "completion_length": 21.84375,
      "epoch": 1.559254327563249,
      "grad_norm": 5.850139617919922,
      "kl": 0.49609375,
      "learning_rate": 6.1e-07,
      "loss": 0.0199,
      "reward": 0.2332974448800087,
      "reward_std": 0.05997936800122261,
      "reward_std_all": 0.07921948283910751,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7474353313446045,
      "rewards/forensics_tempsamp_iou_reward": 0.2332974448800087,
      "step": 586
    },
    {
      "completion_length": 32.5625,
      "epoch": 1.5619174434087881,
      "grad_norm": 5.125926494598389,
      "kl": 0.646484375,
      "learning_rate": 6.093333333333332e-07,
      "loss": 0.0259,
      "reward": 0.17046590149402618,
      "reward_std": 0.11223970353603363,
      "reward_std_all": 0.12334337830543518,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8211471736431122,
      "rewards/forensics_tempsamp_iou_reward": 0.17046590149402618,
      "step": 587
    },
    {
      "completion_length": 33.390625,
      "epoch": 1.5645805592543276,
      "grad_norm": 4.675698757171631,
      "kl": 0.2490234375,
      "learning_rate": 6.086666666666667e-07,
      "loss": 0.01,
      "reward": 0.18386100977659225,
      "reward_std": 0.08837344124913216,
      "reward_std_all": 0.08993365988135338,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6606060564517975,
      "rewards/forensics_tempsamp_iou_reward": 0.18386100977659225,
      "step": 588
    },
    {
      "completion_length": 33.421875,
      "epoch": 1.5672436750998668,
      "grad_norm": 6.028862953186035,
      "kl": 0.390625,
      "learning_rate": 6.079999999999999e-07,
      "loss": 0.0156,
      "reward": 0.22236299514770508,
      "reward_std": 0.10831889882683754,
      "reward_std_all": 0.10974232479929924,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7174242436885834,
      "rewards/forensics_tempsamp_iou_reward": 0.22236299514770508,
      "step": 589
    },
    {
      "completion_length": 30.4375,
      "epoch": 1.569906790945406,
      "grad_norm": 5.998435020446777,
      "kl": 0.33447265625,
      "learning_rate": 6.073333333333333e-07,
      "loss": 0.0134,
      "reward": 0.17697755992412567,
      "reward_std": 0.11494747176766396,
      "reward_std_all": 0.1172599196434021,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6511688232421875,
      "rewards/forensics_tempsamp_iou_reward": 0.17697755992412567,
      "step": 590
    },
    {
      "completion_length": 31.765625,
      "epoch": 1.5725699067909455,
      "grad_norm": 5.19828462600708,
      "kl": 0.4228515625,
      "learning_rate": 6.066666666666666e-07,
      "loss": 0.0169,
      "reward": 0.2138528972864151,
      "reward_std": 0.10592356324195862,
      "reward_std_all": 0.11275576800107956,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8086773753166199,
      "rewards/forensics_tempsamp_iou_reward": 0.2138528972864151,
      "step": 591
    },
    {
      "completion_length": 23.25,
      "epoch": 1.5752330226364846,
      "grad_norm": 4.655128479003906,
      "kl": 0.3857421875,
      "learning_rate": 6.06e-07,
      "loss": 0.0154,
      "reward": 0.3076898157596588,
      "reward_std": 0.10216279327869415,
      "reward_std_all": 0.1153913363814354,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9326926171779633,
      "rewards/forensics_tempsamp_iou_reward": 0.3076898157596588,
      "step": 592
    },
    {
      "completion_length": 27.8125,
      "epoch": 1.577896138482024,
      "grad_norm": 7.400330543518066,
      "kl": 0.4814453125,
      "learning_rate": 6.053333333333332e-07,
      "loss": 0.0192,
      "reward": 0.20600631088018417,
      "reward_std": 0.09495233744382858,
      "reward_std_all": 0.09872612729668617,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7567418813705444,
      "rewards/forensics_tempsamp_iou_reward": 0.20600631088018417,
      "step": 593
    },
    {
      "completion_length": 35.03125,
      "epoch": 1.5805592543275633,
      "grad_norm": 5.309972286224365,
      "kl": 0.4912109375,
      "learning_rate": 6.046666666666667e-07,
      "loss": 0.0196,
      "reward": 0.19755294173955917,
      "reward_std": 0.1066787876188755,
      "reward_std_all": 0.09711626172065735,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6665686070919037,
      "rewards/forensics_tempsamp_iou_reward": 0.19755294173955917,
      "step": 594
    },
    {
      "completion_length": 29.796875,
      "epoch": 1.5832223701731025,
      "grad_norm": 6.573810577392578,
      "kl": 0.3974609375,
      "learning_rate": 6.04e-07,
      "loss": 0.0159,
      "reward": 0.2160937264561653,
      "reward_std": 0.0730817411094904,
      "reward_std_all": 0.08433298766613007,
      "rewards-max/forensics_tempsamp_iou_reward": 0.616970494389534,
      "rewards/forensics_tempsamp_iou_reward": 0.2160937264561653,
      "step": 595
    },
    {
      "completion_length": 25.390625,
      "epoch": 1.5858854860186418,
      "grad_norm": 4.77905797958374,
      "kl": 0.41015625,
      "learning_rate": 6.033333333333333e-07,
      "loss": 0.0164,
      "reward": 0.22490938007831573,
      "reward_std": 0.1092853955924511,
      "reward_std_all": 0.1107526645064354,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7195362448692322,
      "rewards/forensics_tempsamp_iou_reward": 0.22490938007831573,
      "step": 596
    },
    {
      "completion_length": 39.859375,
      "epoch": 1.588548601864181,
      "grad_norm": 6.45747709274292,
      "kl": 0.4921875,
      "learning_rate": 6.026666666666667e-07,
      "loss": 0.0196,
      "reward": 0.2320239022374153,
      "reward_std": 0.11502828076481819,
      "reward_std_all": 0.12110061943531036,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7010279595851898,
      "rewards/forensics_tempsamp_iou_reward": 0.2320239022374153,
      "step": 597
    },
    {
      "completion_length": 46.96875,
      "epoch": 1.5912117177097205,
      "grad_norm": 6.074157238006592,
      "kl": 0.4248046875,
      "learning_rate": 6.019999999999999e-07,
      "loss": 0.017,
      "reward": 0.15798160433769226,
      "reward_std": 0.07751524075865746,
      "reward_std_all": 0.07763680443167686,
      "rewards-max/forensics_tempsamp_iou_reward": 0.49939800053834915,
      "rewards/forensics_tempsamp_iou_reward": 0.15798160433769226,
      "step": 598
    },
    {
      "completion_length": 36.875,
      "epoch": 1.5938748335552595,
      "grad_norm": 5.404850006103516,
      "kl": 0.576171875,
      "learning_rate": 6.013333333333334e-07,
      "loss": 0.023,
      "reward": 0.25863825529813766,
      "reward_std": 0.12059730663895607,
      "reward_std_all": 0.12805192172527313,
      "rewards-max/forensics_tempsamp_iou_reward": 0.805322140455246,
      "rewards/forensics_tempsamp_iou_reward": 0.25863825529813766,
      "step": 599
    },
    {
      "completion_length": 56.34375,
      "epoch": 1.596537949400799,
      "grad_norm": 6.603378772735596,
      "kl": 0.4541015625,
      "learning_rate": 6.006666666666666e-07,
      "loss": 0.0182,
      "reward": 0.15371407568454742,
      "reward_std": 0.0845293439924717,
      "reward_std_all": 0.09539712220430374,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6117842197418213,
      "rewards/forensics_tempsamp_iou_reward": 0.15371407568454742,
      "step": 600
    },
    {
      "completion_length": 49.25,
      "epoch": 1.5992010652463382,
      "grad_norm": 4.614071369171143,
      "kl": 0.4384765625,
      "learning_rate": 6e-07,
      "loss": 0.0175,
      "reward": 0.2042541652917862,
      "reward_std": 0.12337804958224297,
      "reward_std_all": 0.12229353934526443,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7939242720603943,
      "rewards/forensics_tempsamp_iou_reward": 0.2042541652917862,
      "step": 601
    },
    {
      "completion_length": 44.671875,
      "epoch": 1.6018641810918774,
      "grad_norm": 5.339547634124756,
      "kl": 0.57421875,
      "learning_rate": 5.993333333333333e-07,
      "loss": 0.023,
      "reward": 0.24616160243749619,
      "reward_std": 0.10631674528121948,
      "reward_std_all": 0.11292029917240143,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7743199169635773,
      "rewards/forensics_tempsamp_iou_reward": 0.24616160243749619,
      "step": 602
    },
    {
      "completion_length": 38.0,
      "epoch": 1.604527296937417,
      "grad_norm": 5.259523391723633,
      "kl": 0.638671875,
      "learning_rate": 5.986666666666667e-07,
      "loss": 0.0256,
      "reward": 0.17570561170578003,
      "reward_std": 0.1081041544675827,
      "reward_std_all": 0.11434484645724297,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6758406907320023,
      "rewards/forensics_tempsamp_iou_reward": 0.17570561170578003,
      "step": 603
    },
    {
      "completion_length": 58.5625,
      "epoch": 1.607190412782956,
      "grad_norm": 6.7216081619262695,
      "kl": 0.681640625,
      "learning_rate": 5.979999999999999e-07,
      "loss": 0.0272,
      "reward": 0.23059073090553284,
      "reward_std": 0.15572745725512505,
      "reward_std_all": 0.1441626213490963,
      "rewards-max/forensics_tempsamp_iou_reward": 0.82098388671875,
      "rewards/forensics_tempsamp_iou_reward": 0.23059073090553284,
      "step": 604
    },
    {
      "completion_length": 61.0625,
      "epoch": 1.6098535286284954,
      "grad_norm": 5.444424152374268,
      "kl": 0.630859375,
      "learning_rate": 5.973333333333334e-07,
      "loss": 0.0252,
      "reward": 0.2049470618367195,
      "reward_std": 0.13015351071953773,
      "reward_std_all": 0.1286204569041729,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7246572971343994,
      "rewards/forensics_tempsamp_iou_reward": 0.2049470618367195,
      "step": 605
    },
    {
      "completion_length": 43.796875,
      "epoch": 1.6125166444740346,
      "grad_norm": 6.045553684234619,
      "kl": 0.412109375,
      "learning_rate": 5.966666666666666e-07,
      "loss": 0.0165,
      "reward": 0.3583724945783615,
      "reward_std": 0.13811829686164856,
      "reward_std_all": 0.13901044055819511,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9913824498653412,
      "rewards/forensics_tempsamp_iou_reward": 0.3583724945783615,
      "step": 606
    },
    {
      "completion_length": 29.34375,
      "epoch": 1.6151797603195739,
      "grad_norm": 5.482343673706055,
      "kl": 0.52734375,
      "learning_rate": 5.96e-07,
      "loss": 0.0211,
      "reward": 0.23140139877796173,
      "reward_std": 0.07149109989404678,
      "reward_std_all": 0.07705347239971161,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7485544085502625,
      "rewards/forensics_tempsamp_iou_reward": 0.23140139877796173,
      "step": 607
    },
    {
      "completion_length": 32.28125,
      "epoch": 1.6178428761651131,
      "grad_norm": 5.154675483703613,
      "kl": 0.55859375,
      "learning_rate": 5.953333333333333e-07,
      "loss": 0.0224,
      "reward": 0.22885046154260635,
      "reward_std": 0.10636083781719208,
      "reward_std_all": 0.10879084840416908,
      "rewards-max/forensics_tempsamp_iou_reward": 0.630611777305603,
      "rewards/forensics_tempsamp_iou_reward": 0.22885046154260635,
      "step": 608
    },
    {
      "completion_length": 35.109375,
      "epoch": 1.6205059920106524,
      "grad_norm": 4.0740885734558105,
      "kl": 0.662109375,
      "learning_rate": 5.946666666666667e-07,
      "loss": 0.0265,
      "reward": 0.18989310413599014,
      "reward_std": 0.0744626373052597,
      "reward_std_all": 0.07811656221747398,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6113569438457489,
      "rewards/forensics_tempsamp_iou_reward": 0.18989310413599014,
      "step": 609
    },
    {
      "completion_length": 30.703125,
      "epoch": 1.6231691078561918,
      "grad_norm": 5.378664493560791,
      "kl": 0.591796875,
      "learning_rate": 5.939999999999999e-07,
      "loss": 0.0237,
      "reward": 0.27679771184921265,
      "reward_std": 0.1431259922683239,
      "reward_std_all": 0.1377328261733055,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9715625643730164,
      "rewards/forensics_tempsamp_iou_reward": 0.27679771184921265,
      "step": 610
    },
    {
      "completion_length": 26.4375,
      "epoch": 1.6258322237017309,
      "grad_norm": 5.148289680480957,
      "kl": 0.515625,
      "learning_rate": 5.933333333333334e-07,
      "loss": 0.0206,
      "reward": 0.26957976818084717,
      "reward_std": 0.0919974334537983,
      "reward_std_all": 0.09607045352458954,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7975589036941528,
      "rewards/forensics_tempsamp_iou_reward": 0.26957976818084717,
      "step": 611
    },
    {
      "completion_length": 22.703125,
      "epoch": 1.6284953395472703,
      "grad_norm": 5.807670593261719,
      "kl": 0.3623046875,
      "learning_rate": 5.926666666666667e-07,
      "loss": 0.0145,
      "reward": 0.37456341087818146,
      "reward_std": 0.130186527967453,
      "reward_std_all": 0.14818121492862701,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9510793685913086,
      "rewards/forensics_tempsamp_iou_reward": 0.37456341087818146,
      "step": 612
    },
    {
      "completion_length": 26.453125,
      "epoch": 1.6311584553928096,
      "grad_norm": 5.989894390106201,
      "kl": 0.5107421875,
      "learning_rate": 5.919999999999999e-07,
      "loss": 0.0204,
      "reward": 0.27080097049474716,
      "reward_std": 0.10591725260019302,
      "reward_std_all": 0.11020272597670555,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7046397924423218,
      "rewards/forensics_tempsamp_iou_reward": 0.27080097049474716,
      "step": 613
    },
    {
      "completion_length": 22.40625,
      "epoch": 1.6338215712383488,
      "grad_norm": 4.873281002044678,
      "kl": 0.361328125,
      "learning_rate": 5.913333333333334e-07,
      "loss": 0.0144,
      "reward": 0.27133314311504364,
      "reward_std": 0.09268786758184433,
      "reward_std_all": 0.09041234105825424,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8485919833183289,
      "rewards/forensics_tempsamp_iou_reward": 0.27133314311504364,
      "step": 614
    },
    {
      "completion_length": 26.5,
      "epoch": 1.6364846870838883,
      "grad_norm": 5.219773292541504,
      "kl": 0.5205078125,
      "learning_rate": 5.906666666666666e-07,
      "loss": 0.0208,
      "reward": 0.27519017457962036,
      "reward_std": 0.12153735384345055,
      "reward_std_all": 0.1153726726770401,
      "rewards-max/forensics_tempsamp_iou_reward": 0.928569108247757,
      "rewards/forensics_tempsamp_iou_reward": 0.27519017457962036,
      "step": 615
    },
    {
      "completion_length": 24.90625,
      "epoch": 1.6391478029294273,
      "grad_norm": 4.64023494720459,
      "kl": 0.4921875,
      "learning_rate": 5.9e-07,
      "loss": 0.0196,
      "reward": 0.269857794046402,
      "reward_std": 0.12296075373888016,
      "reward_std_all": 0.11659616231918335,
      "rewards-max/forensics_tempsamp_iou_reward": 0.960276335477829,
      "rewards/forensics_tempsamp_iou_reward": 0.269857794046402,
      "step": 616
    },
    {
      "completion_length": 22.0625,
      "epoch": 1.6418109187749668,
      "grad_norm": 7.412161350250244,
      "kl": 0.4443359375,
      "learning_rate": 5.893333333333333e-07,
      "loss": 0.0178,
      "reward": 0.31117017567157745,
      "reward_std": 0.11823153495788574,
      "reward_std_all": 0.11856427416205406,
      "rewards-max/forensics_tempsamp_iou_reward": 0.965210348367691,
      "rewards/forensics_tempsamp_iou_reward": 0.31117017567157745,
      "step": 617
    },
    {
      "completion_length": 27.671875,
      "epoch": 1.644474034620506,
      "grad_norm": 4.352371692657471,
      "kl": 0.615234375,
      "learning_rate": 5.886666666666667e-07,
      "loss": 0.0246,
      "reward": 0.23472227901220322,
      "reward_std": 0.08286194130778313,
      "reward_std_all": 0.0927373506128788,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7767676711082458,
      "rewards/forensics_tempsamp_iou_reward": 0.23472227901220322,
      "step": 618
    },
    {
      "completion_length": 17.671875,
      "epoch": 1.6471371504660453,
      "grad_norm": 5.442075729370117,
      "kl": 0.400390625,
      "learning_rate": 5.879999999999999e-07,
      "loss": 0.016,
      "reward": 0.14396006613969803,
      "reward_std": 0.05348342843353748,
      "reward_std_all": 0.053270138800144196,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6300828158855438,
      "rewards/forensics_tempsamp_iou_reward": 0.14396006613969803,
      "step": 619
    },
    {
      "completion_length": 19.078125,
      "epoch": 1.6498002663115847,
      "grad_norm": 5.535388946533203,
      "kl": 0.5126953125,
      "learning_rate": 5.873333333333334e-07,
      "loss": 0.0205,
      "reward": 0.44952210038900375,
      "reward_std": 0.09753134846687317,
      "reward_std_all": 0.09899076074361801,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9375420808792114,
      "rewards/forensics_tempsamp_iou_reward": 0.44952210038900375,
      "step": 620
    },
    {
      "completion_length": 20.03125,
      "epoch": 1.6524633821571237,
      "grad_norm": 4.764326095581055,
      "kl": 0.3857421875,
      "learning_rate": 5.866666666666666e-07,
      "loss": 0.0154,
      "reward": 0.23068895190954208,
      "reward_std": 0.11284910514950752,
      "reward_std_all": 0.09662492573261261,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9094015061855316,
      "rewards/forensics_tempsamp_iou_reward": 0.23068895190954208,
      "step": 621
    },
    {
      "completion_length": 17.953125,
      "epoch": 1.6551264980026632,
      "grad_norm": 5.654705047607422,
      "kl": 0.39599609375,
      "learning_rate": 5.86e-07,
      "loss": 0.0158,
      "reward": 0.32705114781856537,
      "reward_std": 0.07582304254174232,
      "reward_std_all": 0.08732597902417183,
      "rewards-max/forensics_tempsamp_iou_reward": 0.963668942451477,
      "rewards/forensics_tempsamp_iou_reward": 0.32705114781856537,
      "step": 622
    },
    {
      "completion_length": 24.90625,
      "epoch": 1.6577896138482024,
      "grad_norm": 5.517558574676514,
      "kl": 0.32421875,
      "learning_rate": 5.853333333333333e-07,
      "loss": 0.0129,
      "reward": 0.2234676033258438,
      "reward_std": 0.09315575286746025,
      "reward_std_all": 0.09700452536344528,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8369594514369965,
      "rewards/forensics_tempsamp_iou_reward": 0.2234676033258438,
      "step": 623
    },
    {
      "completion_length": 16.234375,
      "epoch": 1.6604527296937417,
      "grad_norm": 4.9257636070251465,
      "kl": 0.3125,
      "learning_rate": 5.846666666666667e-07,
      "loss": 0.0125,
      "reward": 0.21932245790958405,
      "reward_std": 0.04847273975610733,
      "reward_std_all": 0.053270939737558365,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9138619899749756,
      "rewards/forensics_tempsamp_iou_reward": 0.21932245790958405,
      "step": 624
    },
    {
      "completion_length": 17.90625,
      "epoch": 1.663115845539281,
      "grad_norm": 265.8256530761719,
      "kl": 5.94921875,
      "learning_rate": 5.839999999999999e-07,
      "loss": 0.2384,
      "reward": 0.2761993706226349,
      "reward_std": 0.08193692564964294,
      "reward_std_all": 0.08672548085451126,
      "rewards-max/forensics_tempsamp_iou_reward": 0.933625727891922,
      "rewards/forensics_tempsamp_iou_reward": 0.2761993706226349,
      "step": 625
    },
    {
      "completion_length": 16.78125,
      "epoch": 1.6657789613848202,
      "grad_norm": 4.736279010772705,
      "kl": 0.5556640625,
      "learning_rate": 5.833333333333334e-07,
      "loss": 0.0223,
      "reward": 0.23270411789417267,
      "reward_std": 0.07160651683807373,
      "reward_std_all": 0.08147760108113289,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6471692025661469,
      "rewards/forensics_tempsamp_iou_reward": 0.23270411789417267,
      "step": 626
    },
    {
      "completion_length": 17.125,
      "epoch": 1.6684420772303596,
      "grad_norm": 5.195702075958252,
      "kl": 0.5419921875,
      "learning_rate": 5.826666666666666e-07,
      "loss": 0.0217,
      "reward": 0.29229848086833954,
      "reward_std": 0.07160928659141064,
      "reward_std_all": 0.07778891921043396,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9646394550800323,
      "rewards/forensics_tempsamp_iou_reward": 0.29229848086833954,
      "step": 627
    },
    {
      "completion_length": 19.265625,
      "epoch": 1.6711051930758987,
      "grad_norm": 5.239990711212158,
      "kl": 0.33056640625,
      "learning_rate": 5.819999999999999e-07,
      "loss": 0.0132,
      "reward": 0.14643613249063492,
      "reward_std": 0.04028532654047012,
      "reward_std_all": 0.0553150437772274,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5461622923612595,
      "rewards/forensics_tempsamp_iou_reward": 0.14643613249063492,
      "step": 628
    },
    {
      "completion_length": 17.875,
      "epoch": 1.6737683089214381,
      "grad_norm": 4.4578447341918945,
      "kl": 0.4521484375,
      "learning_rate": 5.813333333333334e-07,
      "loss": 0.0181,
      "reward": 0.3787686675786972,
      "reward_std": 0.11365818977355957,
      "reward_std_all": 0.11955225467681885,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9829314947128296,
      "rewards/forensics_tempsamp_iou_reward": 0.3787686675786972,
      "step": 629
    },
    {
      "completion_length": 17.234375,
      "epoch": 1.6764314247669774,
      "grad_norm": 5.039796352386475,
      "kl": 0.5595703125,
      "learning_rate": 5.806666666666666e-07,
      "loss": 0.0224,
      "reward": 0.2528952956199646,
      "reward_std": 0.08018673956394196,
      "reward_std_all": 0.09492499753832817,
      "rewards-max/forensics_tempsamp_iou_reward": 0.880130797624588,
      "rewards/forensics_tempsamp_iou_reward": 0.2528952956199646,
      "step": 630
    },
    {
      "completion_length": 19.625,
      "epoch": 1.6790945406125166,
      "grad_norm": 5.4015889167785645,
      "kl": 0.431640625,
      "learning_rate": 5.8e-07,
      "loss": 0.0173,
      "reward": 0.3098211735486984,
      "reward_std": 0.10223615542054176,
      "reward_std_all": 0.11062929406762123,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8949688673019409,
      "rewards/forensics_tempsamp_iou_reward": 0.3098211735486984,
      "step": 631
    },
    {
      "completion_length": 20.796875,
      "epoch": 1.681757656458056,
      "grad_norm": 4.974618434906006,
      "kl": 0.39453125,
      "learning_rate": 5.793333333333333e-07,
      "loss": 0.0158,
      "reward": 0.27492400258779526,
      "reward_std": 0.09536436572670937,
      "reward_std_all": 0.08659148216247559,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9316239058971405,
      "rewards/forensics_tempsamp_iou_reward": 0.27492400258779526,
      "step": 632
    },
    {
      "completion_length": 19.671875,
      "epoch": 1.684420772303595,
      "grad_norm": 5.281399250030518,
      "kl": 0.4111328125,
      "learning_rate": 5.786666666666667e-07,
      "loss": 0.0164,
      "reward": 0.23401978611946106,
      "reward_std": 0.07995907217264175,
      "reward_std_all": 0.09128076583147049,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8632398247718811,
      "rewards/forensics_tempsamp_iou_reward": 0.23401978611946106,
      "step": 633
    },
    {
      "completion_length": 24.609375,
      "epoch": 1.6870838881491346,
      "grad_norm": 4.899574279785156,
      "kl": 0.564453125,
      "learning_rate": 5.779999999999999e-07,
      "loss": 0.0226,
      "reward": 0.29169556498527527,
      "reward_std": 0.1041603721678257,
      "reward_std_all": 0.12907330691814423,
      "rewards-max/forensics_tempsamp_iou_reward": 0.857887476682663,
      "rewards/forensics_tempsamp_iou_reward": 0.29169556498527527,
      "step": 634
    },
    {
      "completion_length": 21.484375,
      "epoch": 1.6897470039946738,
      "grad_norm": 5.156764507293701,
      "kl": 0.4306640625,
      "learning_rate": 5.773333333333334e-07,
      "loss": 0.0172,
      "reward": 0.22342225909233093,
      "reward_std": 0.0928741917014122,
      "reward_std_all": 0.10663270950317383,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5750360786914825,
      "rewards/forensics_tempsamp_iou_reward": 0.22342225909233093,
      "step": 635
    },
    {
      "completion_length": 19.734375,
      "epoch": 1.692410119840213,
      "grad_norm": 5.606726169586182,
      "kl": 0.703125,
      "learning_rate": 5.766666666666666e-07,
      "loss": 0.0282,
      "reward": 0.22059515118598938,
      "reward_std": 0.09804839640855789,
      "reward_std_all": 0.11201678216457367,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9030004441738129,
      "rewards/forensics_tempsamp_iou_reward": 0.22059515118598938,
      "step": 636
    },
    {
      "completion_length": 20.609375,
      "epoch": 1.6950732356857523,
      "grad_norm": 5.016711235046387,
      "kl": 0.4580078125,
      "learning_rate": 5.76e-07,
      "loss": 0.0183,
      "reward": 0.3596666604280472,
      "reward_std": 0.09518064185976982,
      "reward_std_all": 0.12671879678964615,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8682739436626434,
      "rewards/forensics_tempsamp_iou_reward": 0.3596666604280472,
      "step": 637
    },
    {
      "completion_length": 25.734375,
      "epoch": 1.6977363515312915,
      "grad_norm": 5.017139434814453,
      "kl": 0.482421875,
      "learning_rate": 5.753333333333333e-07,
      "loss": 0.0193,
      "reward": 0.3520887494087219,
      "reward_std": 0.14319634437561035,
      "reward_std_all": 0.13150574266910553,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8123107254505157,
      "rewards/forensics_tempsamp_iou_reward": 0.3520887494087219,
      "step": 638
    },
    {
      "completion_length": 19.859375,
      "epoch": 1.700399467376831,
      "grad_norm": 5.934607028961182,
      "kl": 0.556640625,
      "learning_rate": 5.746666666666667e-07,
      "loss": 0.0223,
      "reward": 0.2683071494102478,
      "reward_std": 0.10526607930660248,
      "reward_std_all": 0.10420554876327515,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7923980355262756,
      "rewards/forensics_tempsamp_iou_reward": 0.2683071494102478,
      "step": 639
    },
    {
      "completion_length": 33.125,
      "epoch": 1.70306258322237,
      "grad_norm": 5.060929775238037,
      "kl": 0.734375,
      "learning_rate": 5.739999999999999e-07,
      "loss": 0.0294,
      "reward": 0.31304575502872467,
      "reward_std": 0.14515449851751328,
      "reward_std_all": 0.15927009284496307,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8631313145160675,
      "rewards/forensics_tempsamp_iou_reward": 0.31304575502872467,
      "step": 640
    },
    {
      "completion_length": 24.609375,
      "epoch": 1.7057256990679095,
      "grad_norm": 6.148922443389893,
      "kl": 0.47265625,
      "learning_rate": 5.733333333333334e-07,
      "loss": 0.0189,
      "reward": 0.23720379173755646,
      "reward_std": 0.12511324137449265,
      "reward_std_all": 0.12738573923707008,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9187114834785461,
      "rewards/forensics_tempsamp_iou_reward": 0.23720379173755646,
      "step": 641
    },
    {
      "completion_length": 25.625,
      "epoch": 1.7083888149134487,
      "grad_norm": 6.367737770080566,
      "kl": 0.4033203125,
      "learning_rate": 5.726666666666666e-07,
      "loss": 0.0161,
      "reward": 0.19218650460243225,
      "reward_std": 0.08410846814513206,
      "reward_std_all": 0.07871096208691597,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8684961199760437,
      "rewards/forensics_tempsamp_iou_reward": 0.19218650460243225,
      "step": 642
    },
    {
      "completion_length": 33.640625,
      "epoch": 1.711051930758988,
      "grad_norm": 5.0758819580078125,
      "kl": 0.591796875,
      "learning_rate": 5.719999999999999e-07,
      "loss": 0.0237,
      "reward": 0.19955959171056747,
      "reward_std": 0.10947628691792488,
      "reward_std_all": 0.10937459394335747,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6587963104248047,
      "rewards/forensics_tempsamp_iou_reward": 0.19955959171056747,
      "step": 643
    },
    {
      "completion_length": 21.84375,
      "epoch": 1.7137150466045274,
      "grad_norm": 5.681300163269043,
      "kl": 0.603515625,
      "learning_rate": 5.713333333333333e-07,
      "loss": 0.0241,
      "reward": 0.18278536200523376,
      "reward_std": 0.06514335796236992,
      "reward_std_all": 0.07336341589689255,
      "rewards-max/forensics_tempsamp_iou_reward": 0.75811967253685,
      "rewards/forensics_tempsamp_iou_reward": 0.18278536200523376,
      "step": 644
    },
    {
      "completion_length": 24.703125,
      "epoch": 1.7163781624500665,
      "grad_norm": 6.504759788513184,
      "kl": 0.44921875,
      "learning_rate": 5.706666666666666e-07,
      "loss": 0.018,
      "reward": 0.26834118366241455,
      "reward_std": 0.060190560296177864,
      "reward_std_all": 0.060514893382787704,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8879509568214417,
      "rewards/forensics_tempsamp_iou_reward": 0.26834118366241455,
      "step": 645
    },
    {
      "completion_length": 26.546875,
      "epoch": 1.719041278295606,
      "grad_norm": 5.656229496002197,
      "kl": 0.4677734375,
      "learning_rate": 5.699999999999999e-07,
      "loss": 0.0187,
      "reward": 0.304905042052269,
      "reward_std": 0.10919343307614326,
      "reward_std_all": 0.12882079929113388,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7638993263244629,
      "rewards/forensics_tempsamp_iou_reward": 0.304905042052269,
      "step": 646
    },
    {
      "completion_length": 24.296875,
      "epoch": 1.7217043941411452,
      "grad_norm": 5.363536357879639,
      "kl": 0.4140625,
      "learning_rate": 5.693333333333333e-07,
      "loss": 0.0166,
      "reward": 0.2345857247710228,
      "reward_std": 0.09960015118122101,
      "reward_std_all": 0.09969967976212502,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9290621280670166,
      "rewards/forensics_tempsamp_iou_reward": 0.2345857247710228,
      "step": 647
    },
    {
      "completion_length": 32.03125,
      "epoch": 1.7243675099866844,
      "grad_norm": 4.521429538726807,
      "kl": 0.4609375,
      "learning_rate": 5.686666666666667e-07,
      "loss": 0.0184,
      "reward": 0.24618086963891983,
      "reward_std": 0.09159659966826439,
      "reward_std_all": 0.08728759735822678,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9253202974796295,
      "rewards/forensics_tempsamp_iou_reward": 0.24618086963891983,
      "step": 648
    },
    {
      "completion_length": 37.03125,
      "epoch": 1.7270306258322237,
      "grad_norm": 4.690366268157959,
      "kl": 0.498046875,
      "learning_rate": 5.679999999999999e-07,
      "loss": 0.0199,
      "reward": 0.1668791025876999,
      "reward_std": 0.07320558466017246,
      "reward_std_all": 0.09559902548789978,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5294191837310791,
      "rewards/forensics_tempsamp_iou_reward": 0.1668791025876999,
      "step": 649
    },
    {
      "completion_length": 42.28125,
      "epoch": 1.729693741677763,
      "grad_norm": 5.790279865264893,
      "kl": 0.431640625,
      "learning_rate": 5.673333333333334e-07,
      "loss": 0.0173,
      "reward": 0.39635588228702545,
      "reward_std": 0.1110951378941536,
      "reward_std_all": 0.11029614508152008,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9660307168960571,
      "rewards/forensics_tempsamp_iou_reward": 0.39635588228702545,
      "step": 650
    },
    {
      "completion_length": 32.09375,
      "epoch": 1.7323568575233024,
      "grad_norm": 8.351301193237305,
      "kl": 0.951171875,
      "learning_rate": 5.666666666666666e-07,
      "loss": 0.0381,
      "reward": 0.2845176011323929,
      "reward_std": 0.09438317269086838,
      "reward_std_all": 0.1178201213479042,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8610332310199738,
      "rewards/forensics_tempsamp_iou_reward": 0.2845176011323929,
      "step": 651
    },
    {
      "completion_length": 36.0625,
      "epoch": 1.7350199733688414,
      "grad_norm": 6.285243511199951,
      "kl": 0.68359375,
      "learning_rate": 5.66e-07,
      "loss": 0.0274,
      "reward": 0.32196222245693207,
      "reward_std": 0.11468749865889549,
      "reward_std_all": 0.15724355727434158,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9135896265506744,
      "rewards/forensics_tempsamp_iou_reward": 0.32196222245693207,
      "step": 652
    },
    {
      "completion_length": 27.21875,
      "epoch": 1.7376830892143809,
      "grad_norm": 5.471109867095947,
      "kl": 0.8369140625,
      "learning_rate": 5.653333333333333e-07,
      "loss": 0.0335,
      "reward": 0.22919729351997375,
      "reward_std": 0.12392263859510422,
      "reward_std_all": 0.12727534025907516,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7015529572963715,
      "rewards/forensics_tempsamp_iou_reward": 0.22919729351997375,
      "step": 653
    },
    {
      "completion_length": 31.828125,
      "epoch": 1.74034620505992,
      "grad_norm": 6.095883846282959,
      "kl": 0.517578125,
      "learning_rate": 5.646666666666667e-07,
      "loss": 0.0207,
      "reward": 0.1754152849316597,
      "reward_std": 0.07591458037495613,
      "reward_std_all": 0.09752124175429344,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5380835235118866,
      "rewards/forensics_tempsamp_iou_reward": 0.1754152849316597,
      "step": 654
    },
    {
      "completion_length": 33.328125,
      "epoch": 1.7430093209054593,
      "grad_norm": 5.8780364990234375,
      "kl": 0.4541015625,
      "learning_rate": 5.639999999999999e-07,
      "loss": 0.0181,
      "reward": 0.24257583916187286,
      "reward_std": 0.10667065531015396,
      "reward_std_all": 0.10665819048881531,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8647817373275757,
      "rewards/forensics_tempsamp_iou_reward": 0.24257583916187286,
      "step": 655
    },
    {
      "completion_length": 32.109375,
      "epoch": 1.7456724367509988,
      "grad_norm": 5.244271755218506,
      "kl": 0.45703125,
      "learning_rate": 5.633333333333334e-07,
      "loss": 0.0183,
      "reward": 0.38097378611564636,
      "reward_std": 0.08323494717478752,
      "reward_std_all": 0.0908506028354168,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8776880204677582,
      "rewards/forensics_tempsamp_iou_reward": 0.38097378611564636,
      "step": 656
    },
    {
      "completion_length": 23.609375,
      "epoch": 1.7483355525965378,
      "grad_norm": 6.450013160705566,
      "kl": 0.55859375,
      "learning_rate": 5.626666666666666e-07,
      "loss": 0.0224,
      "reward": 0.31624260544776917,
      "reward_std": 0.11249244958162308,
      "reward_std_all": 0.10706201568245888,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7829526364803314,
      "rewards/forensics_tempsamp_iou_reward": 0.31624260544776917,
      "step": 657
    },
    {
      "completion_length": 29.75,
      "epoch": 1.7509986684420773,
      "grad_norm": 6.105297088623047,
      "kl": 0.3818359375,
      "learning_rate": 5.620000000000001e-07,
      "loss": 0.0152,
      "reward": 0.30922529101371765,
      "reward_std": 0.10266237333416939,
      "reward_std_all": 0.11135182529687881,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9031613767147064,
      "rewards/forensics_tempsamp_iou_reward": 0.30922529101371765,
      "step": 658
    },
    {
      "completion_length": 33.578125,
      "epoch": 1.7536617842876165,
      "grad_norm": 5.5551533699035645,
      "kl": 0.5205078125,
      "learning_rate": 5.613333333333333e-07,
      "loss": 0.0208,
      "reward": 0.29129326343536377,
      "reward_std": 0.060592977330088615,
      "reward_std_all": 0.09436415135860443,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7708359062671661,
      "rewards/forensics_tempsamp_iou_reward": 0.29129326343536377,
      "step": 659
    },
    {
      "completion_length": 25.15625,
      "epoch": 1.7563249001331558,
      "grad_norm": 6.559389114379883,
      "kl": 0.41015625,
      "learning_rate": 5.606666666666666e-07,
      "loss": 0.0164,
      "reward": 0.3651495426893234,
      "reward_std": 0.07268904522061348,
      "reward_std_all": 0.07998219132423401,
      "rewards-max/forensics_tempsamp_iou_reward": 0.963131308555603,
      "rewards/forensics_tempsamp_iou_reward": 0.3651495426893234,
      "step": 660
    },
    {
      "completion_length": 38.640625,
      "epoch": 1.758988015978695,
      "grad_norm": 6.459887504577637,
      "kl": 0.3994140625,
      "learning_rate": 5.6e-07,
      "loss": 0.016,
      "reward": 0.3748321533203125,
      "reward_std": 0.08988529071211815,
      "reward_std_all": 0.12169113010168076,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9620689749717712,
      "rewards/forensics_tempsamp_iou_reward": 0.3748321533203125,
      "step": 661
    },
    {
      "completion_length": 34.25,
      "epoch": 1.7616511318242343,
      "grad_norm": 5.345134735107422,
      "kl": 0.4697265625,
      "learning_rate": 5.593333333333333e-07,
      "loss": 0.0188,
      "reward": 0.21560294181108475,
      "reward_std": 0.0944400429725647,
      "reward_std_all": 0.1006891094148159,
      "rewards-max/forensics_tempsamp_iou_reward": 0.889085054397583,
      "rewards/forensics_tempsamp_iou_reward": 0.21560294181108475,
      "step": 662
    },
    {
      "completion_length": 47.203125,
      "epoch": 1.7643142476697737,
      "grad_norm": 6.625260353088379,
      "kl": 0.4521484375,
      "learning_rate": 5.586666666666666e-07,
      "loss": 0.0181,
      "reward": 0.15289798006415367,
      "reward_std": 0.09277928248047829,
      "reward_std_all": 0.08409463614225388,
      "rewards-max/forensics_tempsamp_iou_reward": 0.48279570043087006,
      "rewards/forensics_tempsamp_iou_reward": 0.15289798006415367,
      "step": 663
    },
    {
      "completion_length": 33.46875,
      "epoch": 1.7669773635153128,
      "grad_norm": 8.737817764282227,
      "kl": 0.87890625,
      "learning_rate": 5.58e-07,
      "loss": 0.0351,
      "reward": 0.1571822725236416,
      "reward_std": 0.0886499248445034,
      "reward_std_all": 0.0814347043633461,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6108808517456055,
      "rewards/forensics_tempsamp_iou_reward": 0.1571822725236416,
      "step": 664
    },
    {
      "completion_length": 36.21875,
      "epoch": 1.7696404793608522,
      "grad_norm": 6.414546489715576,
      "kl": 0.4560546875,
      "learning_rate": 5.573333333333333e-07,
      "loss": 0.0182,
      "reward": 0.2195066660642624,
      "reward_std": 0.09394844248890877,
      "reward_std_all": 0.0891277827322483,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6907610595226288,
      "rewards/forensics_tempsamp_iou_reward": 0.2195066660642624,
      "step": 665
    },
    {
      "completion_length": 23.75,
      "epoch": 1.7723035952063915,
      "grad_norm": 4.812690258026123,
      "kl": 0.5009765625,
      "learning_rate": 5.566666666666666e-07,
      "loss": 0.02,
      "reward": 0.30756382644176483,
      "reward_std": 0.06258191540837288,
      "reward_std_all": 0.07328757084906101,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9351809620857239,
      "rewards/forensics_tempsamp_iou_reward": 0.30756382644176483,
      "step": 666
    },
    {
      "completion_length": 34.25,
      "epoch": 1.7749667110519307,
      "grad_norm": 6.903110027313232,
      "kl": 0.58203125,
      "learning_rate": 5.560000000000001e-07,
      "loss": 0.0233,
      "reward": 0.19729359447956085,
      "reward_std": 0.07121860049664974,
      "reward_std_all": 0.07326573505997658,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6313017904758453,
      "rewards/forensics_tempsamp_iou_reward": 0.19729359447956085,
      "step": 667
    },
    {
      "completion_length": 31.171875,
      "epoch": 1.7776298268974702,
      "grad_norm": 7.257993698120117,
      "kl": 0.4638671875,
      "learning_rate": 5.553333333333333e-07,
      "loss": 0.0186,
      "reward": 0.3384275734424591,
      "reward_std": 0.12763581424951553,
      "reward_std_all": 0.1286393217742443,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9567676782608032,
      "rewards/forensics_tempsamp_iou_reward": 0.3384275734424591,
      "step": 668
    },
    {
      "completion_length": 22.109375,
      "epoch": 1.7802929427430092,
      "grad_norm": 4.658946990966797,
      "kl": 0.37451171875,
      "learning_rate": 5.546666666666667e-07,
      "loss": 0.015,
      "reward": 0.30982711911201477,
      "reward_std": 0.0787472203373909,
      "reward_std_all": 0.0737805888056755,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8276327848434448,
      "rewards/forensics_tempsamp_iou_reward": 0.30982711911201477,
      "step": 669
    },
    {
      "completion_length": 39.765625,
      "epoch": 1.7829560585885487,
      "grad_norm": 4.760278701782227,
      "kl": 0.505859375,
      "learning_rate": 5.54e-07,
      "loss": 0.0202,
      "reward": 0.22359763830900192,
      "reward_std": 0.13808652013540268,
      "reward_std_all": 0.13092472776770592,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7436826825141907,
      "rewards/forensics_tempsamp_iou_reward": 0.22359763830900192,
      "step": 670
    },
    {
      "completion_length": 44.34375,
      "epoch": 1.785619174434088,
      "grad_norm": 4.841224670410156,
      "kl": 0.501953125,
      "learning_rate": 5.533333333333334e-07,
      "loss": 0.02,
      "reward": 0.2786852568387985,
      "reward_std": 0.10687457025051117,
      "reward_std_all": 0.10356298834085464,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9915099442005157,
      "rewards/forensics_tempsamp_iou_reward": 0.2786852568387985,
      "step": 671
    },
    {
      "completion_length": 23.203125,
      "epoch": 1.7882822902796272,
      "grad_norm": 5.633042812347412,
      "kl": 0.43359375,
      "learning_rate": 5.526666666666666e-07,
      "loss": 0.0173,
      "reward": 0.3184516876935959,
      "reward_std": 0.11364898085594177,
      "reward_std_all": 0.12825267761945724,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9314552247524261,
      "rewards/forensics_tempsamp_iou_reward": 0.3184516876935959,
      "step": 672
    },
    {
      "completion_length": 30.09375,
      "epoch": 1.7909454061251664,
      "grad_norm": 4.433487892150879,
      "kl": 0.416015625,
      "learning_rate": 5.520000000000001e-07,
      "loss": 0.0166,
      "reward": 0.17343958467245102,
      "reward_std": 0.08701244741678238,
      "reward_std_all": 0.08465303480625153,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6259277164936066,
      "rewards/forensics_tempsamp_iou_reward": 0.17343958467245102,
      "step": 673
    },
    {
      "completion_length": 33.875,
      "epoch": 1.7936085219707056,
      "grad_norm": 6.863597869873047,
      "kl": 0.4794921875,
      "learning_rate": 5.513333333333333e-07,
      "loss": 0.0192,
      "reward": 0.2132703810930252,
      "reward_std": 0.08821716904640198,
      "reward_std_all": 0.10906333848834038,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6128151267766953,
      "rewards/forensics_tempsamp_iou_reward": 0.2132703810930252,
      "step": 674
    },
    {
      "completion_length": 29.703125,
      "epoch": 1.796271637816245,
      "grad_norm": 5.942811012268066,
      "kl": 0.5009765625,
      "learning_rate": 5.506666666666666e-07,
      "loss": 0.0201,
      "reward": 0.20879629999399185,
      "reward_std": 0.06350630708038807,
      "reward_std_all": 0.06385229900479317,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7974116206169128,
      "rewards/forensics_tempsamp_iou_reward": 0.20879629999399185,
      "step": 675
    },
    {
      "completion_length": 32.84375,
      "epoch": 1.7989347536617841,
      "grad_norm": 5.378634929656982,
      "kl": 0.52734375,
      "learning_rate": 5.5e-07,
      "loss": 0.0211,
      "reward": 0.1980675756931305,
      "reward_std": 0.09126371145248413,
      "reward_std_all": 0.10372545197606087,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8316141664981842,
      "rewards/forensics_tempsamp_iou_reward": 0.1980675756931305,
      "step": 676
    },
    {
      "completion_length": 27.6875,
      "epoch": 1.8015978695073236,
      "grad_norm": 6.20062780380249,
      "kl": 0.3974609375,
      "learning_rate": 5.493333333333333e-07,
      "loss": 0.0159,
      "reward": 0.16663551703095436,
      "reward_std": 0.08009032905101776,
      "reward_std_all": 0.08253008499741554,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4312499910593033,
      "rewards/forensics_tempsamp_iou_reward": 0.16663551703095436,
      "step": 677
    },
    {
      "completion_length": 21.140625,
      "epoch": 1.8042609853528628,
      "grad_norm": 5.317364692687988,
      "kl": 0.50244140625,
      "learning_rate": 5.486666666666666e-07,
      "loss": 0.0201,
      "reward": 0.321657657623291,
      "reward_std": 0.11446502059698105,
      "reward_std_all": 0.12018522620201111,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8436107039451599,
      "rewards/forensics_tempsamp_iou_reward": 0.321657657623291,
      "step": 678
    },
    {
      "completion_length": 32.6875,
      "epoch": 1.806924101198402,
      "grad_norm": 6.73022985458374,
      "kl": 0.43359375,
      "learning_rate": 5.48e-07,
      "loss": 0.0174,
      "reward": 0.19708162546157837,
      "reward_std": 0.09468098729848862,
      "reward_std_all": 0.10125122033059597,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7644356787204742,
      "rewards/forensics_tempsamp_iou_reward": 0.19708162546157837,
      "step": 679
    },
    {
      "completion_length": 29.96875,
      "epoch": 1.8095872170439415,
      "grad_norm": 5.933170318603516,
      "kl": 0.62890625,
      "learning_rate": 5.473333333333333e-07,
      "loss": 0.0252,
      "reward": 0.3570483773946762,
      "reward_std": 0.08685445785522461,
      "reward_std_all": 0.10158070176839828,
      "rewards-max/forensics_tempsamp_iou_reward": 0.93830206990242,
      "rewards/forensics_tempsamp_iou_reward": 0.3570483773946762,
      "step": 680
    },
    {
      "completion_length": 18.28125,
      "epoch": 1.8122503328894806,
      "grad_norm": 4.066721439361572,
      "kl": 0.392578125,
      "learning_rate": 5.466666666666666e-07,
      "loss": 0.0157,
      "reward": 0.36509886384010315,
      "reward_std": 0.07284429110586643,
      "reward_std_all": 0.0708017461001873,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9734815061092377,
      "rewards/forensics_tempsamp_iou_reward": 0.36509886384010315,
      "step": 681
    },
    {
      "completion_length": 20.3125,
      "epoch": 1.81491344873502,
      "grad_norm": 6.163588047027588,
      "kl": 0.38671875,
      "learning_rate": 5.46e-07,
      "loss": 0.0155,
      "reward": 0.3640894442796707,
      "reward_std": 0.06665027886629105,
      "reward_std_all": 0.07953276485204697,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9893833100795746,
      "rewards/forensics_tempsamp_iou_reward": 0.3640894442796707,
      "step": 682
    },
    {
      "completion_length": 27.96875,
      "epoch": 1.8175765645805593,
      "grad_norm": 4.3321943283081055,
      "kl": 0.3720703125,
      "learning_rate": 5.453333333333333e-07,
      "loss": 0.0149,
      "reward": 0.32602502405643463,
      "reward_std": 0.07866596803069115,
      "reward_std_all": 0.08786382526159286,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7354721426963806,
      "rewards/forensics_tempsamp_iou_reward": 0.32602502405643463,
      "step": 683
    },
    {
      "completion_length": 48.890625,
      "epoch": 1.8202396804260985,
      "grad_norm": 4.20168399810791,
      "kl": 0.3623046875,
      "learning_rate": 5.446666666666666e-07,
      "loss": 0.0145,
      "reward": 0.25337325036525726,
      "reward_std": 0.11043676733970642,
      "reward_std_all": 0.1045827716588974,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7790933549404144,
      "rewards/forensics_tempsamp_iou_reward": 0.25337325036525726,
      "step": 684
    },
    {
      "completion_length": 16.9375,
      "epoch": 1.822902796271638,
      "grad_norm": 5.992528915405273,
      "kl": 0.2705078125,
      "learning_rate": 5.44e-07,
      "loss": 0.0109,
      "reward": 0.328800767660141,
      "reward_std": 0.11033386178314686,
      "reward_std_all": 0.10524048656225204,
      "rewards-max/forensics_tempsamp_iou_reward": 0.95772984623909,
      "rewards/forensics_tempsamp_iou_reward": 0.328800767660141,
      "step": 685
    },
    {
      "completion_length": 23.625,
      "epoch": 1.825565912117177,
      "grad_norm": 6.353752613067627,
      "kl": 0.4609375,
      "learning_rate": 5.433333333333334e-07,
      "loss": 0.0184,
      "reward": 0.2648468241095543,
      "reward_std": 0.08665069565176964,
      "reward_std_all": 0.0769798718392849,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8208062648773193,
      "rewards/forensics_tempsamp_iou_reward": 0.2648468241095543,
      "step": 686
    },
    {
      "completion_length": 24.34375,
      "epoch": 1.8282290279627165,
      "grad_norm": 7.202049255371094,
      "kl": 0.49609375,
      "learning_rate": 5.426666666666666e-07,
      "loss": 0.0199,
      "reward": 0.3779169023036957,
      "reward_std": 0.054383743554353714,
      "reward_std_all": 0.06776942126452923,
      "rewards-max/forensics_tempsamp_iou_reward": 0.940291702747345,
      "rewards/forensics_tempsamp_iou_reward": 0.3779169023036957,
      "step": 687
    },
    {
      "completion_length": 26.828125,
      "epoch": 1.8308921438082557,
      "grad_norm": 6.4211745262146,
      "kl": 0.4365234375,
      "learning_rate": 5.420000000000001e-07,
      "loss": 0.0174,
      "reward": 0.240916445851326,
      "reward_std": 0.06185879558324814,
      "reward_std_all": 0.06880464404821396,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9095344543457031,
      "rewards/forensics_tempsamp_iou_reward": 0.240916445851326,
      "step": 688
    },
    {
      "completion_length": 28.125,
      "epoch": 1.833555259653795,
      "grad_norm": 7.406388759613037,
      "kl": 0.373046875,
      "learning_rate": 5.413333333333333e-07,
      "loss": 0.0149,
      "reward": 0.26681631803512573,
      "reward_std": 0.06580425426363945,
      "reward_std_all": 0.07171830721199512,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8868643939495087,
      "rewards/forensics_tempsamp_iou_reward": 0.26681631803512573,
      "step": 689
    },
    {
      "completion_length": 23.65625,
      "epoch": 1.8362183754993342,
      "grad_norm": 6.235084056854248,
      "kl": 0.44140625,
      "learning_rate": 5.406666666666666e-07,
      "loss": 0.0177,
      "reward": 0.15451502054929733,
      "reward_std": 0.06073449179530144,
      "reward_std_all": 0.05957039259374142,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6066962778568268,
      "rewards/forensics_tempsamp_iou_reward": 0.15451502054929733,
      "step": 690
    },
    {
      "completion_length": 26.953125,
      "epoch": 1.8388814913448734,
      "grad_norm": 4.821417808532715,
      "kl": 0.396484375,
      "learning_rate": 5.4e-07,
      "loss": 0.0159,
      "reward": 0.2871055454015732,
      "reward_std": 0.10437832400202751,
      "reward_std_all": 0.10506739467382431,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9685664176940918,
      "rewards/forensics_tempsamp_iou_reward": 0.2871055454015732,
      "step": 691
    },
    {
      "completion_length": 24.328125,
      "epoch": 1.841544607190413,
      "grad_norm": 6.559597492218018,
      "kl": 0.427734375,
      "learning_rate": 5.393333333333333e-07,
      "loss": 0.0171,
      "reward": 0.31661456823349,
      "reward_std": 0.07777640968561172,
      "reward_std_all": 0.09477512538433075,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9823837280273438,
      "rewards/forensics_tempsamp_iou_reward": 0.31661456823349,
      "step": 692
    },
    {
      "completion_length": 36.15625,
      "epoch": 1.844207723035952,
      "grad_norm": 6.650701999664307,
      "kl": 0.40625,
      "learning_rate": 5.386666666666666e-07,
      "loss": 0.0163,
      "reward": 0.26999062299728394,
      "reward_std": 0.07454285211861134,
      "reward_std_all": 0.10931147634983063,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9209351539611816,
      "rewards/forensics_tempsamp_iou_reward": 0.26999062299728394,
      "step": 693
    },
    {
      "completion_length": 40.359375,
      "epoch": 1.8468708388814914,
      "grad_norm": 8.236299514770508,
      "kl": 0.3759765625,
      "learning_rate": 5.38e-07,
      "loss": 0.015,
      "reward": 0.2265503630042076,
      "reward_std": 0.046501992270350456,
      "reward_std_all": 0.051090991124510765,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6910572499036789,
      "rewards/forensics_tempsamp_iou_reward": 0.2265503630042076,
      "step": 694
    },
    {
      "completion_length": 39.265625,
      "epoch": 1.8495339547270306,
      "grad_norm": 5.919670581817627,
      "kl": 0.50390625,
      "learning_rate": 5.373333333333333e-07,
      "loss": 0.0202,
      "reward": 0.26109927892684937,
      "reward_std": 0.0904693640768528,
      "reward_std_all": 0.08936002105474472,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9440586566925049,
      "rewards/forensics_tempsamp_iou_reward": 0.26109927892684937,
      "step": 695
    },
    {
      "completion_length": 43.375,
      "epoch": 1.8521970705725699,
      "grad_norm": 6.045272350311279,
      "kl": 0.357421875,
      "learning_rate": 5.366666666666666e-07,
      "loss": 0.0143,
      "reward": 0.21811093389987946,
      "reward_std": 0.06566280126571655,
      "reward_std_all": 0.08964963629841805,
      "rewards-max/forensics_tempsamp_iou_reward": 0.880261093378067,
      "rewards/forensics_tempsamp_iou_reward": 0.21811093389987946,
      "step": 696
    },
    {
      "completion_length": 41.53125,
      "epoch": 1.8548601864181093,
      "grad_norm": 4.795764446258545,
      "kl": 0.3955078125,
      "learning_rate": 5.36e-07,
      "loss": 0.0158,
      "reward": 0.27395202964544296,
      "reward_std": 0.1290327087044716,
      "reward_std_all": 0.11590767651796341,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9374639093875885,
      "rewards/forensics_tempsamp_iou_reward": 0.27395202964544296,
      "step": 697
    },
    {
      "completion_length": 41.0,
      "epoch": 1.8575233022636484,
      "grad_norm": 6.067930221557617,
      "kl": 0.4580078125,
      "learning_rate": 5.353333333333333e-07,
      "loss": 0.0183,
      "reward": 0.26872705668210983,
      "reward_std": 0.0990775041282177,
      "reward_std_all": 0.10439306870102882,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7866122126579285,
      "rewards/forensics_tempsamp_iou_reward": 0.26872705668210983,
      "step": 698
    },
    {
      "completion_length": 37.8125,
      "epoch": 1.8601864181091878,
      "grad_norm": 4.90748929977417,
      "kl": 0.322265625,
      "learning_rate": 5.346666666666666e-07,
      "loss": 0.0129,
      "reward": 0.37610873579978943,
      "reward_std": 0.09048725292086601,
      "reward_std_all": 0.09666935727000237,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9632972776889801,
      "rewards/forensics_tempsamp_iou_reward": 0.37610873579978943,
      "step": 699
    },
    {
      "completion_length": 57.34375,
      "epoch": 1.862849533954727,
      "grad_norm": 4.683249473571777,
      "kl": 0.3408203125,
      "learning_rate": 5.34e-07,
      "loss": 0.0136,
      "reward": 0.2855464518070221,
      "reward_std": 0.0730656310915947,
      "reward_std_all": 0.08294961228966713,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9306921660900116,
      "rewards/forensics_tempsamp_iou_reward": 0.2855464518070221,
      "step": 700
    },
    {
      "completion_length": 52.25,
      "epoch": 1.8655126498002663,
      "grad_norm": 5.016707897186279,
      "kl": 0.37109375,
      "learning_rate": 5.333333333333333e-07,
      "loss": 0.0149,
      "reward": 0.2404426857829094,
      "reward_std": 0.08224598318338394,
      "reward_std_all": 0.07714950665831566,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9159186780452728,
      "rewards/forensics_tempsamp_iou_reward": 0.2404426857829094,
      "step": 701
    },
    {
      "completion_length": 61.640625,
      "epoch": 1.8681757656458056,
      "grad_norm": 4.813536167144775,
      "kl": 0.36328125,
      "learning_rate": 5.326666666666666e-07,
      "loss": 0.0145,
      "reward": 0.34249547123908997,
      "reward_std": 0.12967347726225853,
      "reward_std_all": 0.1314915418624878,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9808375239372253,
      "rewards/forensics_tempsamp_iou_reward": 0.34249547123908997,
      "step": 702
    },
    {
      "completion_length": 63.140625,
      "epoch": 1.8708388814913448,
      "grad_norm": 4.613360404968262,
      "kl": 0.482421875,
      "learning_rate": 5.32e-07,
      "loss": 0.0193,
      "reward": 0.3015991598367691,
      "reward_std": 0.08904844522476196,
      "reward_std_all": 0.10131003335118294,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9710504412651062,
      "rewards/forensics_tempsamp_iou_reward": 0.3015991598367691,
      "step": 703
    },
    {
      "completion_length": 39.046875,
      "epoch": 1.8735019973368843,
      "grad_norm": 5.018439769744873,
      "kl": 0.5703125,
      "learning_rate": 5.313333333333333e-07,
      "loss": 0.0227,
      "reward": 0.23874887824058533,
      "reward_std": 0.09623004868626595,
      "reward_std_all": 0.08909270167350769,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8506944477558136,
      "rewards/forensics_tempsamp_iou_reward": 0.23874887824058533,
      "step": 704
    },
    {
      "completion_length": 38.171875,
      "epoch": 1.8761651131824233,
      "grad_norm": 8.321675300598145,
      "kl": 0.68359375,
      "learning_rate": 5.306666666666665e-07,
      "loss": 0.0273,
      "reward": 0.17679660022258759,
      "reward_std": 0.08344232477247715,
      "reward_std_all": 0.09072573482990265,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6406079083681107,
      "rewards/forensics_tempsamp_iou_reward": 0.17679660022258759,
      "step": 705
    },
    {
      "completion_length": 45.890625,
      "epoch": 1.8788282290279628,
      "grad_norm": 5.801996231079102,
      "kl": 0.3251953125,
      "learning_rate": 5.3e-07,
      "loss": 0.013,
      "reward": 0.41353435814380646,
      "reward_std": 0.08432058244943619,
      "reward_std_all": 0.08302009850740433,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9917379915714264,
      "rewards/forensics_tempsamp_iou_reward": 0.41353435814380646,
      "step": 706
    },
    {
      "completion_length": 59.390625,
      "epoch": 1.881491344873502,
      "grad_norm": 10.180855751037598,
      "kl": 0.634765625,
      "learning_rate": 5.293333333333333e-07,
      "loss": 0.0254,
      "reward": 0.21775755286216736,
      "reward_std": 0.08394009247422218,
      "reward_std_all": 0.08784062415361404,
      "rewards-max/forensics_tempsamp_iou_reward": 0.857049286365509,
      "rewards/forensics_tempsamp_iou_reward": 0.21775755286216736,
      "step": 707
    },
    {
      "completion_length": 45.421875,
      "epoch": 1.8841544607190412,
      "grad_norm": 5.511646270751953,
      "kl": 0.564453125,
      "learning_rate": 5.286666666666666e-07,
      "loss": 0.0227,
      "reward": 0.14827970787882805,
      "reward_std": 0.06942342594265938,
      "reward_std_all": 0.06697928160429001,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5324829965829849,
      "rewards/forensics_tempsamp_iou_reward": 0.14827970787882805,
      "step": 708
    },
    {
      "completion_length": 38.65625,
      "epoch": 1.8868175765645807,
      "grad_norm": 6.597505569458008,
      "kl": 0.4873046875,
      "learning_rate": 5.28e-07,
      "loss": 0.0195,
      "reward": 0.24992069602012634,
      "reward_std": 0.08705192059278488,
      "reward_std_all": 0.09041337668895721,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9829230010509491,
      "rewards/forensics_tempsamp_iou_reward": 0.24992069602012634,
      "step": 709
    },
    {
      "completion_length": 26.15625,
      "epoch": 1.8894806924101197,
      "grad_norm": 5.434576511383057,
      "kl": 0.533203125,
      "learning_rate": 5.273333333333333e-07,
      "loss": 0.0213,
      "reward": 0.3122722655534744,
      "reward_std": 0.061223600059747696,
      "reward_std_all": 0.08212796598672867,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8416496217250824,
      "rewards/forensics_tempsamp_iou_reward": 0.3122722655534744,
      "step": 710
    },
    {
      "completion_length": 40.875,
      "epoch": 1.8921438082556592,
      "grad_norm": 6.053107261657715,
      "kl": 0.5107421875,
      "learning_rate": 5.266666666666666e-07,
      "loss": 0.0204,
      "reward": 0.31733886897563934,
      "reward_std": 0.07146182842552662,
      "reward_std_all": 0.08836972340941429,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9141460359096527,
      "rewards/forensics_tempsamp_iou_reward": 0.31733886897563934,
      "step": 711
    },
    {
      "completion_length": 44.71875,
      "epoch": 1.8948069241011984,
      "grad_norm": 6.695362567901611,
      "kl": 0.525390625,
      "learning_rate": 5.26e-07,
      "loss": 0.021,
      "reward": 0.27080947905778885,
      "reward_std": 0.06127013638615608,
      "reward_std_all": 0.0932883583009243,
      "rewards-max/forensics_tempsamp_iou_reward": 0.964193195104599,
      "rewards/forensics_tempsamp_iou_reward": 0.27080947905778885,
      "step": 712
    },
    {
      "completion_length": 46.609375,
      "epoch": 1.8974700399467377,
      "grad_norm": 4.586643695831299,
      "kl": 0.533203125,
      "learning_rate": 5.253333333333333e-07,
      "loss": 0.0214,
      "reward": 0.18554117530584335,
      "reward_std": 0.11056087166070938,
      "reward_std_all": 0.1147061251103878,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9202800393104553,
      "rewards/forensics_tempsamp_iou_reward": 0.18554117530584335,
      "step": 713
    },
    {
      "completion_length": 29.953125,
      "epoch": 1.900133155792277,
      "grad_norm": 7.601879119873047,
      "kl": 0.703125,
      "learning_rate": 5.246666666666666e-07,
      "loss": 0.0282,
      "reward": 0.27144572883844376,
      "reward_std": 0.06827457249164581,
      "reward_std_all": 0.0761001743376255,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9185703098773956,
      "rewards/forensics_tempsamp_iou_reward": 0.27144572883844376,
      "step": 714
    },
    {
      "completion_length": 47.265625,
      "epoch": 1.9027962716378162,
      "grad_norm": 5.562623500823975,
      "kl": 0.51171875,
      "learning_rate": 5.24e-07,
      "loss": 0.0205,
      "reward": 0.27866584062576294,
      "reward_std": 0.09970761463046074,
      "reward_std_all": 0.11677544564008713,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9157924354076385,
      "rewards/forensics_tempsamp_iou_reward": 0.27866584062576294,
      "step": 715
    },
    {
      "completion_length": 37.25,
      "epoch": 1.9054593874833556,
      "grad_norm": 4.620017051696777,
      "kl": 0.4033203125,
      "learning_rate": 5.233333333333333e-07,
      "loss": 0.0161,
      "reward": 0.3359920531511307,
      "reward_std": 0.08525320142507553,
      "reward_std_all": 0.09886205941438675,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9693481028079987,
      "rewards/forensics_tempsamp_iou_reward": 0.3359920531511307,
      "step": 716
    },
    {
      "completion_length": 34.484375,
      "epoch": 1.9081225033288947,
      "grad_norm": 7.481531143188477,
      "kl": 0.4453125,
      "learning_rate": 5.226666666666666e-07,
      "loss": 0.0178,
      "reward": 0.27381710708141327,
      "reward_std": 0.07706335559487343,
      "reward_std_all": 0.08828551322221756,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9006148874759674,
      "rewards/forensics_tempsamp_iou_reward": 0.27381710708141327,
      "step": 717
    },
    {
      "completion_length": 22.5,
      "epoch": 1.9107856191744341,
      "grad_norm": 6.654573917388916,
      "kl": 0.595703125,
      "learning_rate": 5.22e-07,
      "loss": 0.0238,
      "reward": 0.379924014210701,
      "reward_std": 0.132803276181221,
      "reward_std_all": 0.11994699016213417,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9940617382526398,
      "rewards/forensics_tempsamp_iou_reward": 0.379924014210701,
      "step": 718
    },
    {
      "completion_length": 32.546875,
      "epoch": 1.9134487350199734,
      "grad_norm": 6.864485263824463,
      "kl": 0.65234375,
      "learning_rate": 5.213333333333333e-07,
      "loss": 0.026,
      "reward": 0.30479809641838074,
      "reward_std": 0.12312101945281029,
      "reward_std_all": 0.11223236843943596,
      "rewards-max/forensics_tempsamp_iou_reward": 0.942065417766571,
      "rewards/forensics_tempsamp_iou_reward": 0.30479809641838074,
      "step": 719
    },
    {
      "completion_length": 39.34375,
      "epoch": 1.9161118508655126,
      "grad_norm": 4.659502983093262,
      "kl": 0.44140625,
      "learning_rate": 5.206666666666667e-07,
      "loss": 0.0176,
      "reward": 0.17691199108958244,
      "reward_std": 0.07326330989599228,
      "reward_std_all": 0.07147282175719738,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6003851741552353,
      "rewards/forensics_tempsamp_iou_reward": 0.17691199108958244,
      "step": 720
    },
    {
      "completion_length": 20.734375,
      "epoch": 1.918774966711052,
      "grad_norm": 6.191751956939697,
      "kl": 0.4208984375,
      "learning_rate": 5.2e-07,
      "loss": 0.0168,
      "reward": 0.2762850821018219,
      "reward_std": 0.13705581054091454,
      "reward_std_all": 0.12797120213508606,
      "rewards-max/forensics_tempsamp_iou_reward": 0.987094908952713,
      "rewards/forensics_tempsamp_iou_reward": 0.2762850821018219,
      "step": 721
    },
    {
      "completion_length": 21.40625,
      "epoch": 1.921438082556591,
      "grad_norm": 6.636031150817871,
      "kl": 0.4052734375,
      "learning_rate": 5.193333333333332e-07,
      "loss": 0.0162,
      "reward": 0.3879026621580124,
      "reward_std": 0.06526878289878368,
      "reward_std_all": 0.07064813934266567,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9730335772037506,
      "rewards/forensics_tempsamp_iou_reward": 0.3879026621580124,
      "step": 722
    },
    {
      "completion_length": 37.609375,
      "epoch": 1.9241011984021306,
      "grad_norm": 6.957055568695068,
      "kl": 0.4677734375,
      "learning_rate": 5.186666666666667e-07,
      "loss": 0.0187,
      "reward": 0.19211586564779282,
      "reward_std": 0.09667813219130039,
      "reward_std_all": 0.08715309016406536,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5885281413793564,
      "rewards/forensics_tempsamp_iou_reward": 0.19211586564779282,
      "step": 723
    },
    {
      "completion_length": 24.0625,
      "epoch": 1.9267643142476698,
      "grad_norm": 4.781151294708252,
      "kl": 0.453125,
      "learning_rate": 5.18e-07,
      "loss": 0.0181,
      "reward": 0.24468880146741867,
      "reward_std": 0.06357617676258087,
      "reward_std_all": 0.0906672328710556,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8982680439949036,
      "rewards/forensics_tempsamp_iou_reward": 0.24468880146741867,
      "step": 724
    },
    {
      "completion_length": 29.0,
      "epoch": 1.929427430093209,
      "grad_norm": 5.893590927124023,
      "kl": 0.6591796875,
      "learning_rate": 5.173333333333333e-07,
      "loss": 0.0264,
      "reward": 0.28768547624349594,
      "reward_std": 0.08287281915545464,
      "reward_std_all": 0.12504977732896805,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9005098640918732,
      "rewards/forensics_tempsamp_iou_reward": 0.28768547624349594,
      "step": 725
    },
    {
      "completion_length": 20.203125,
      "epoch": 1.9320905459387483,
      "grad_norm": 7.155563831329346,
      "kl": 0.486328125,
      "learning_rate": 5.166666666666667e-07,
      "loss": 0.0195,
      "reward": 0.38577814400196075,
      "reward_std": 0.14180173724889755,
      "reward_std_all": 0.13520626723766327,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9615774154663086,
      "rewards/forensics_tempsamp_iou_reward": 0.38577814400196075,
      "step": 726
    },
    {
      "completion_length": 24.96875,
      "epoch": 1.9347536617842875,
      "grad_norm": 5.90507698059082,
      "kl": 0.6259765625,
      "learning_rate": 5.16e-07,
      "loss": 0.0251,
      "reward": 0.23934589326381683,
      "reward_std": 0.09915893524885178,
      "reward_std_all": 0.1258518137037754,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9625953137874603,
      "rewards/forensics_tempsamp_iou_reward": 0.23934589326381683,
      "step": 727
    },
    {
      "completion_length": 32.625,
      "epoch": 1.937416777629827,
      "grad_norm": 5.354251861572266,
      "kl": 0.4755859375,
      "learning_rate": 5.153333333333333e-07,
      "loss": 0.019,
      "reward": 0.3611048460006714,
      "reward_std": 0.15018903091549873,
      "reward_std_all": 0.13528499007225037,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9550490975379944,
      "rewards/forensics_tempsamp_iou_reward": 0.3611048460006714,
      "step": 728
    },
    {
      "completion_length": 23.8125,
      "epoch": 1.940079893475366,
      "grad_norm": 5.126606464385986,
      "kl": 0.3154296875,
      "learning_rate": 5.146666666666667e-07,
      "loss": 0.0126,
      "reward": 0.358876571059227,
      "reward_std": 0.09409873560070992,
      "reward_std_all": 0.09543996676802635,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8715547025203705,
      "rewards/forensics_tempsamp_iou_reward": 0.358876571059227,
      "step": 729
    },
    {
      "completion_length": 27.6875,
      "epoch": 1.9427430093209055,
      "grad_norm": 6.718873977661133,
      "kl": 0.4423828125,
      "learning_rate": 5.14e-07,
      "loss": 0.0177,
      "reward": 0.18828821927309036,
      "reward_std": 0.0678953155875206,
      "reward_std_all": 0.06689763814210892,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7743548154830933,
      "rewards/forensics_tempsamp_iou_reward": 0.18828821927309036,
      "step": 730
    },
    {
      "completion_length": 18.21875,
      "epoch": 1.9454061251664447,
      "grad_norm": 7.10886287689209,
      "kl": 0.4580078125,
      "learning_rate": 5.133333333333333e-07,
      "loss": 0.0183,
      "reward": 0.2896210849285126,
      "reward_std": 0.08345565386116505,
      "reward_std_all": 0.07979157939553261,
      "rewards-max/forensics_tempsamp_iou_reward": 0.975621372461319,
      "rewards/forensics_tempsamp_iou_reward": 0.2896210849285126,
      "step": 731
    },
    {
      "completion_length": 20.453125,
      "epoch": 1.948069241011984,
      "grad_norm": 6.497289180755615,
      "kl": 0.4609375,
      "learning_rate": 5.126666666666667e-07,
      "loss": 0.0184,
      "reward": 0.25124890357255936,
      "reward_std": 0.06041656993329525,
      "reward_std_all": 0.06599946320056915,
      "rewards-max/forensics_tempsamp_iou_reward": 0.762866199016571,
      "rewards/forensics_tempsamp_iou_reward": 0.25124890357255936,
      "step": 732
    },
    {
      "completion_length": 24.046875,
      "epoch": 1.9507323568575234,
      "grad_norm": 5.386425495147705,
      "kl": 0.3671875,
      "learning_rate": 5.12e-07,
      "loss": 0.0147,
      "reward": 0.22253020852804184,
      "reward_std": 0.056473931297659874,
      "reward_std_all": 0.059712134301662445,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7786779701709747,
      "rewards/forensics_tempsamp_iou_reward": 0.22253020852804184,
      "step": 733
    },
    {
      "completion_length": 19.28125,
      "epoch": 1.9533954727030625,
      "grad_norm": 4.697003364562988,
      "kl": 0.4208984375,
      "learning_rate": 5.113333333333333e-07,
      "loss": 0.0168,
      "reward": 0.3326067328453064,
      "reward_std": 0.09414621070027351,
      "reward_std_all": 0.10704635083675385,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9646567404270172,
      "rewards/forensics_tempsamp_iou_reward": 0.3326067328453064,
      "step": 734
    },
    {
      "completion_length": 20.703125,
      "epoch": 1.956058588548602,
      "grad_norm": 4.9730963706970215,
      "kl": 0.5673828125,
      "learning_rate": 5.106666666666667e-07,
      "loss": 0.0227,
      "reward": 0.28485043346881866,
      "reward_std": 0.11003726720809937,
      "reward_std_all": 0.11037029325962067,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8233714699745178,
      "rewards/forensics_tempsamp_iou_reward": 0.28485043346881866,
      "step": 735
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.9587217043941412,
      "grad_norm": 6.520168304443359,
      "kl": 0.392578125,
      "learning_rate": 5.1e-07,
      "loss": 0.0157,
      "reward": 0.2967829704284668,
      "reward_std": 0.09326572343707085,
      "reward_std_all": 0.07994946092367172,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9673202633857727,
      "rewards/forensics_tempsamp_iou_reward": 0.2967829704284668,
      "step": 736
    },
    {
      "completion_length": 24.34375,
      "epoch": 1.9613848202396804,
      "grad_norm": 5.670892238616943,
      "kl": 0.439453125,
      "learning_rate": 5.093333333333332e-07,
      "loss": 0.0176,
      "reward": 0.2707928419113159,
      "reward_std": 0.11707255616784096,
      "reward_std_all": 0.11315668374300003,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9049272537231445,
      "rewards/forensics_tempsamp_iou_reward": 0.2707928419113159,
      "step": 737
    },
    {
      "completion_length": 59.515625,
      "epoch": 1.9640479360852197,
      "grad_norm": 4.503116607666016,
      "kl": 0.4716796875,
      "learning_rate": 5.086666666666667e-07,
      "loss": 0.0189,
      "reward": 0.18567445129156113,
      "reward_std": 0.07711577415466309,
      "reward_std_all": 0.07821493968367577,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6435586661100388,
      "rewards/forensics_tempsamp_iou_reward": 0.18567445129156113,
      "step": 738
    },
    {
      "completion_length": 17.859375,
      "epoch": 1.966711051930759,
      "grad_norm": 6.505039691925049,
      "kl": 0.4052734375,
      "learning_rate": 5.079999999999999e-07,
      "loss": 0.0162,
      "reward": 0.31389859318733215,
      "reward_std": 0.08995001018047333,
      "reward_std_all": 0.09514925628900528,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8387534022331238,
      "rewards/forensics_tempsamp_iou_reward": 0.31389859318733215,
      "step": 739
    },
    {
      "completion_length": 20.8125,
      "epoch": 1.9693741677762984,
      "grad_norm": 6.336374759674072,
      "kl": 0.3408203125,
      "learning_rate": 5.073333333333333e-07,
      "loss": 0.0137,
      "reward": 0.31697122752666473,
      "reward_std": 0.05932413041591644,
      "reward_std_all": 0.06739061325788498,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8997496664524078,
      "rewards/forensics_tempsamp_iou_reward": 0.31697122752666473,
      "step": 740
    },
    {
      "completion_length": 35.9375,
      "epoch": 1.9720372836218374,
      "grad_norm": 6.0503435134887695,
      "kl": 0.388671875,
      "learning_rate": 5.066666666666667e-07,
      "loss": 0.0156,
      "reward": 0.19466989487409592,
      "reward_std": 0.06809062883257866,
      "reward_std_all": 0.06217017024755478,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6511698216199875,
      "rewards/forensics_tempsamp_iou_reward": 0.19466989487409592,
      "step": 741
    },
    {
      "completion_length": 24.25,
      "epoch": 1.9747003994673769,
      "grad_norm": 7.539649486541748,
      "kl": 0.3466796875,
      "learning_rate": 5.06e-07,
      "loss": 0.0139,
      "reward": 0.3146063983440399,
      "reward_std": 0.07847289741039276,
      "reward_std_all": 0.0814732201397419,
      "rewards-max/forensics_tempsamp_iou_reward": 0.980733335018158,
      "rewards/forensics_tempsamp_iou_reward": 0.3146063983440399,
      "step": 742
    },
    {
      "completion_length": 27.6875,
      "epoch": 1.977363515312916,
      "grad_norm": 7.683173656463623,
      "kl": 0.7578125,
      "learning_rate": 5.053333333333333e-07,
      "loss": 0.0304,
      "reward": 0.2808585911989212,
      "reward_std": 0.10237646475434303,
      "reward_std_all": 0.10105280205607414,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9181357026100159,
      "rewards/forensics_tempsamp_iou_reward": 0.2808585911989212,
      "step": 743
    },
    {
      "completion_length": 39.109375,
      "epoch": 1.9800266311584553,
      "grad_norm": 4.3785200119018555,
      "kl": 0.44921875,
      "learning_rate": 5.046666666666667e-07,
      "loss": 0.0179,
      "reward": 0.2333238385617733,
      "reward_std": 0.05903547257184982,
      "reward_std_all": 0.056153666228055954,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6056044399738312,
      "rewards/forensics_tempsamp_iou_reward": 0.2333238385617733,
      "step": 744
    },
    {
      "completion_length": 20.9375,
      "epoch": 1.9826897470039948,
      "grad_norm": 7.760324478149414,
      "kl": 0.359375,
      "learning_rate": 5.04e-07,
      "loss": 0.0144,
      "reward": 0.26712554693222046,
      "reward_std": 0.05403153784573078,
      "reward_std_all": 0.0541615542024374,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9363425970077515,
      "rewards/forensics_tempsamp_iou_reward": 0.26712554693222046,
      "step": 745
    },
    {
      "completion_length": 34.203125,
      "epoch": 1.9853528628495338,
      "grad_norm": 4.665639400482178,
      "kl": 0.4208984375,
      "learning_rate": 5.033333333333333e-07,
      "loss": 0.0168,
      "reward": 0.3730171471834183,
      "reward_std": 0.03924519196152687,
      "reward_std_all": 0.07261967845261097,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9803114235401154,
      "rewards/forensics_tempsamp_iou_reward": 0.3730171471834183,
      "step": 746
    },
    {
      "completion_length": 21.28125,
      "epoch": 1.9880159786950733,
      "grad_norm": 9.598118782043457,
      "kl": 0.2998046875,
      "learning_rate": 5.026666666666667e-07,
      "loss": 0.012,
      "reward": 0.1929917261004448,
      "reward_std": 0.04981653392314911,
      "reward_std_all": 0.06176441162824631,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4706679731607437,
      "rewards/forensics_tempsamp_iou_reward": 0.1929917261004448,
      "step": 747
    },
    {
      "completion_length": 25.125,
      "epoch": 1.9906790945406125,
      "grad_norm": 4.922881126403809,
      "kl": 0.412109375,
      "learning_rate": 5.02e-07,
      "loss": 0.0165,
      "reward": 0.33320730924606323,
      "reward_std": 0.047695038840174675,
      "reward_std_all": 0.04805848374962807,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9912377595901489,
      "rewards/forensics_tempsamp_iou_reward": 0.33320730924606323,
      "step": 748
    },
    {
      "completion_length": 21.296875,
      "epoch": 1.9933422103861518,
      "grad_norm": 7.194376468658447,
      "kl": 0.509765625,
      "learning_rate": 5.013333333333333e-07,
      "loss": 0.0204,
      "reward": 0.2675557881593704,
      "reward_std": 0.07605774700641632,
      "reward_std_all": 0.07309763506054878,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9278545081615448,
      "rewards/forensics_tempsamp_iou_reward": 0.2675557881593704,
      "step": 749
    },
    {
      "completion_length": 27.234375,
      "epoch": 1.996005326231691,
      "grad_norm": 7.244904041290283,
      "kl": 0.5810546875,
      "learning_rate": 5.006666666666667e-07,
      "loss": 0.0233,
      "reward": 0.28918546438217163,
      "reward_std": 0.06747733801603317,
      "reward_std_all": 0.0678094457834959,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9032118022441864,
      "rewards/forensics_tempsamp_iou_reward": 0.28918546438217163,
      "step": 750
    },
    {
      "completion_length": 24.359375,
      "epoch": 1.9986684420772303,
      "grad_norm": 6.0213422775268555,
      "kl": 0.5419921875,
      "learning_rate": 5e-07,
      "loss": 0.0217,
      "reward": 0.21966619789600372,
      "reward_std": 0.08054972812533379,
      "reward_std_all": 0.09425114095211029,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6239737272262573,
      "rewards/forensics_tempsamp_iou_reward": 0.21966619789600372,
      "step": 751
    },
    {
      "completion_length": 14.0,
      "epoch": 2.0,
      "grad_norm": 4.192494869232178,
      "kl": 0.29296875,
      "learning_rate": 4.993333333333333e-07,
      "loss": 0.0063,
      "reward": 0.08275862038135529,
      "reward_std": 0.013717003166675568,
      "reward_std_all": 0.012208113446831703,
      "rewards-max/forensics_tempsamp_iou_reward": 0.08275862038135529,
      "rewards/forensics_tempsamp_iou_reward": 0.08275862038135529,
      "step": 752
    },
    {
      "completion_length": 30.59375,
      "epoch": 2.0026631158455395,
      "grad_norm": 5.415116310119629,
      "kl": 0.373046875,
      "learning_rate": 4.986666666666666e-07,
      "loss": 0.0149,
      "reward": 0.37349408864974976,
      "reward_std": 0.0779758095741272,
      "reward_std_all": 0.07995553314685822,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9041285216808319,
      "rewards/forensics_tempsamp_iou_reward": 0.37349408864974976,
      "step": 753
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.0053262316910785,
      "grad_norm": 7.6542205810546875,
      "kl": 0.5341796875,
      "learning_rate": 4.979999999999999e-07,
      "loss": 0.0214,
      "reward": 0.5081600248813629,
      "reward_std": 0.13990852981805801,
      "reward_std_all": 0.1578848585486412,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9933828413486481,
      "rewards/forensics_tempsamp_iou_reward": 0.5081600248813629,
      "step": 754
    },
    {
      "completion_length": 26.390625,
      "epoch": 2.007989347536618,
      "grad_norm": 6.116979598999023,
      "kl": 0.43359375,
      "learning_rate": 4.973333333333333e-07,
      "loss": 0.0173,
      "reward": 0.2973511666059494,
      "reward_std": 0.11919838935136795,
      "reward_std_all": 0.11319982260465622,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9589596688747406,
      "rewards/forensics_tempsamp_iou_reward": 0.2973511666059494,
      "step": 755
    },
    {
      "completion_length": 36.84375,
      "epoch": 2.010652463382157,
      "grad_norm": 6.172236442565918,
      "kl": 0.3896484375,
      "learning_rate": 4.966666666666666e-07,
      "loss": 0.0156,
      "reward": 0.23103156685829163,
      "reward_std": 0.10432883352041245,
      "reward_std_all": 0.11865005642175674,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6511635780334473,
      "rewards/forensics_tempsamp_iou_reward": 0.23103156685829163,
      "step": 756
    },
    {
      "completion_length": 48.109375,
      "epoch": 2.0133155792276964,
      "grad_norm": 6.026098251342773,
      "kl": 0.62109375,
      "learning_rate": 4.96e-07,
      "loss": 0.0248,
      "reward": 0.2579658180475235,
      "reward_std": 0.10375245288014412,
      "reward_std_all": 0.11223740503191948,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9221180081367493,
      "rewards/forensics_tempsamp_iou_reward": 0.2579658180475235,
      "step": 757
    },
    {
      "completion_length": 26.875,
      "epoch": 2.015978695073236,
      "grad_norm": 4.9076151847839355,
      "kl": 0.53515625,
      "learning_rate": 4.953333333333333e-07,
      "loss": 0.0215,
      "reward": 0.26958826184272766,
      "reward_std": 0.07973046973347664,
      "reward_std_all": 0.09041623398661613,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8813945055007935,
      "rewards/forensics_tempsamp_iou_reward": 0.26958826184272766,
      "step": 758
    },
    {
      "completion_length": 51.28125,
      "epoch": 2.018641810918775,
      "grad_norm": 5.774356365203857,
      "kl": 0.3994140625,
      "learning_rate": 4.946666666666666e-07,
      "loss": 0.016,
      "reward": 0.2736540734767914,
      "reward_std": 0.09424913302063942,
      "reward_std_all": 0.0957375168800354,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7262011766433716,
      "rewards/forensics_tempsamp_iou_reward": 0.2736540734767914,
      "step": 759
    },
    {
      "completion_length": 34.171875,
      "epoch": 2.0213049267643144,
      "grad_norm": 6.827054023742676,
      "kl": 0.48828125,
      "learning_rate": 4.94e-07,
      "loss": 0.0195,
      "reward": 0.22518093883991241,
      "reward_std": 0.0673053814098239,
      "reward_std_all": 0.06986132822930813,
      "rewards-max/forensics_tempsamp_iou_reward": 0.641874372959137,
      "rewards/forensics_tempsamp_iou_reward": 0.22518093883991241,
      "step": 760
    },
    {
      "completion_length": 39.421875,
      "epoch": 2.0239680426098534,
      "grad_norm": 10.806450843811035,
      "kl": 0.8466796875,
      "learning_rate": 4.933333333333333e-07,
      "loss": 0.0339,
      "reward": 0.23034009337425232,
      "reward_std": 0.07736076787114143,
      "reward_std_all": 0.08667415007948875,
      "rewards-max/forensics_tempsamp_iou_reward": 0.885682076215744,
      "rewards/forensics_tempsamp_iou_reward": 0.23034009337425232,
      "step": 761
    },
    {
      "completion_length": 30.671875,
      "epoch": 2.026631158455393,
      "grad_norm": 6.995960712432861,
      "kl": 0.544921875,
      "learning_rate": 4.926666666666667e-07,
      "loss": 0.0218,
      "reward": 0.2692921683192253,
      "reward_std": 0.09713537245988846,
      "reward_std_all": 0.0903329961001873,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8721751570701599,
      "rewards/forensics_tempsamp_iou_reward": 0.2692921683192253,
      "step": 762
    },
    {
      "completion_length": 28.640625,
      "epoch": 2.029294274300932,
      "grad_norm": 5.96353816986084,
      "kl": 0.6640625,
      "learning_rate": 4.92e-07,
      "loss": 0.0265,
      "reward": 0.19313093274831772,
      "reward_std": 0.06710563227534294,
      "reward_std_all": 0.063824862241745,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8329641819000244,
      "rewards/forensics_tempsamp_iou_reward": 0.19313093274831772,
      "step": 763
    },
    {
      "completion_length": 30.828125,
      "epoch": 2.0319573901464714,
      "grad_norm": 5.470616340637207,
      "kl": 0.607421875,
      "learning_rate": 4.913333333333334e-07,
      "loss": 0.0243,
      "reward": 0.35097917914390564,
      "reward_std": 0.05522290617227554,
      "reward_std_all": 0.0693231150507927,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9699074029922485,
      "rewards/forensics_tempsamp_iou_reward": 0.35097917914390564,
      "step": 764
    },
    {
      "completion_length": 39.28125,
      "epoch": 2.034620505992011,
      "grad_norm": 5.056976795196533,
      "kl": 0.771484375,
      "learning_rate": 4.906666666666666e-07,
      "loss": 0.0309,
      "reward": 0.17264170944690704,
      "reward_std": 0.06661534495651722,
      "reward_std_all": 0.0678353738039732,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6705882400274277,
      "rewards/forensics_tempsamp_iou_reward": 0.17264170944690704,
      "step": 765
    },
    {
      "completion_length": 43.78125,
      "epoch": 2.03728362183755,
      "grad_norm": 4.845458507537842,
      "kl": 0.421875,
      "learning_rate": 4.9e-07,
      "loss": 0.0169,
      "reward": 0.22932226210832596,
      "reward_std": 0.08610499650239944,
      "reward_std_all": 0.08428502082824707,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8924897313117981,
      "rewards/forensics_tempsamp_iou_reward": 0.22932226210832596,
      "step": 766
    },
    {
      "completion_length": 28.59375,
      "epoch": 2.0399467376830893,
      "grad_norm": 4.902706146240234,
      "kl": 0.376953125,
      "learning_rate": 4.893333333333333e-07,
      "loss": 0.0151,
      "reward": 0.21237194538116455,
      "reward_std": 0.0851559191942215,
      "reward_std_all": 0.0809682309627533,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8342756032943726,
      "rewards/forensics_tempsamp_iou_reward": 0.21237194538116455,
      "step": 767
    },
    {
      "completion_length": 38.765625,
      "epoch": 2.0426098535286283,
      "grad_norm": 5.692898273468018,
      "kl": 0.3837890625,
      "learning_rate": 4.886666666666667e-07,
      "loss": 0.0154,
      "reward": 0.23438191413879395,
      "reward_std": 0.07338238880038261,
      "reward_std_all": 0.07176540791988373,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8215160965919495,
      "rewards/forensics_tempsamp_iou_reward": 0.23438191413879395,
      "step": 768
    },
    {
      "completion_length": 37.53125,
      "epoch": 2.045272969374168,
      "grad_norm": 6.046560287475586,
      "kl": 0.3798828125,
      "learning_rate": 4.879999999999999e-07,
      "loss": 0.0152,
      "reward": 0.2767695188522339,
      "reward_std": 0.09122221916913986,
      "reward_std_all": 0.09268498793244362,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8816419541835785,
      "rewards/forensics_tempsamp_iou_reward": 0.2767695188522339,
      "step": 769
    },
    {
      "completion_length": 26.140625,
      "epoch": 2.0479360852197073,
      "grad_norm": 5.497118949890137,
      "kl": 0.640625,
      "learning_rate": 4.873333333333333e-07,
      "loss": 0.0256,
      "reward": 0.4272046238183975,
      "reward_std": 0.12514221668243408,
      "reward_std_all": 0.12506991624832153,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9837891161441803,
      "rewards/forensics_tempsamp_iou_reward": 0.4272046238183975,
      "step": 770
    },
    {
      "completion_length": 39.40625,
      "epoch": 2.0505992010652463,
      "grad_norm": 4.993990421295166,
      "kl": 0.3798828125,
      "learning_rate": 4.866666666666666e-07,
      "loss": 0.0152,
      "reward": 0.21344498544931412,
      "reward_std": 0.07815133780241013,
      "reward_std_all": 0.07584847137331963,
      "rewards-max/forensics_tempsamp_iou_reward": 0.819918692111969,
      "rewards/forensics_tempsamp_iou_reward": 0.21344498544931412,
      "step": 771
    },
    {
      "completion_length": 28.453125,
      "epoch": 2.0532623169107858,
      "grad_norm": 4.536665439605713,
      "kl": 0.451171875,
      "learning_rate": 4.86e-07,
      "loss": 0.0181,
      "reward": 0.31233961135149,
      "reward_std": 0.07487349584698677,
      "reward_std_all": 0.06975237280130386,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9971590936183929,
      "rewards/forensics_tempsamp_iou_reward": 0.31233961135149,
      "step": 772
    },
    {
      "completion_length": 40.765625,
      "epoch": 2.0559254327563248,
      "grad_norm": 5.917294979095459,
      "kl": 0.37109375,
      "learning_rate": 4.853333333333333e-07,
      "loss": 0.0148,
      "reward": 0.394231379032135,
      "reward_std": 0.11161822825670242,
      "reward_std_all": 0.11809047311544418,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9777494370937347,
      "rewards/forensics_tempsamp_iou_reward": 0.394231379032135,
      "step": 773
    },
    {
      "completion_length": 61.171875,
      "epoch": 2.0585885486018642,
      "grad_norm": 4.230681896209717,
      "kl": 0.513671875,
      "learning_rate": 4.846666666666667e-07,
      "loss": 0.0205,
      "reward": 0.22916606813669205,
      "reward_std": 0.12413223087787628,
      "reward_std_all": 0.11878752708435059,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7360910922288895,
      "rewards/forensics_tempsamp_iou_reward": 0.22916606813669205,
      "step": 774
    },
    {
      "completion_length": 48.53125,
      "epoch": 2.0612516644474033,
      "grad_norm": 5.386083602905273,
      "kl": 0.4248046875,
      "learning_rate": 4.839999999999999e-07,
      "loss": 0.017,
      "reward": 0.21202101558446884,
      "reward_std": 0.08784666657447815,
      "reward_std_all": 0.1045566126704216,
      "rewards-max/forensics_tempsamp_iou_reward": 0.761075496673584,
      "rewards/forensics_tempsamp_iou_reward": 0.21202101558446884,
      "step": 775
    },
    {
      "completion_length": 38.734375,
      "epoch": 2.0639147802929427,
      "grad_norm": 6.067046165466309,
      "kl": 0.708984375,
      "learning_rate": 4.833333333333333e-07,
      "loss": 0.0283,
      "reward": 0.27861686050891876,
      "reward_std": 0.11626618728041649,
      "reward_std_all": 0.12133388966321945,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8751309514045715,
      "rewards/forensics_tempsamp_iou_reward": 0.27861686050891876,
      "step": 776
    },
    {
      "completion_length": 49.1875,
      "epoch": 2.066577896138482,
      "grad_norm": 17.608755111694336,
      "kl": 0.880859375,
      "learning_rate": 4.826666666666666e-07,
      "loss": 0.0353,
      "reward": 0.2712481692433357,
      "reward_std": 0.09591212868690491,
      "reward_std_all": 0.10216937959194183,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7754823267459869,
      "rewards/forensics_tempsamp_iou_reward": 0.2712481692433357,
      "step": 777
    },
    {
      "completion_length": 27.578125,
      "epoch": 2.069241011984021,
      "grad_norm": 6.117976188659668,
      "kl": 0.5458984375,
      "learning_rate": 4.82e-07,
      "loss": 0.0218,
      "reward": 0.16621402651071548,
      "reward_std": 0.051579799503088,
      "reward_std_all": 0.05222955159842968,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6741211712360382,
      "rewards/forensics_tempsamp_iou_reward": 0.16621402651071548,
      "step": 778
    },
    {
      "completion_length": 38.1875,
      "epoch": 2.0719041278295607,
      "grad_norm": 4.440543174743652,
      "kl": 0.5166015625,
      "learning_rate": 4.813333333333334e-07,
      "loss": 0.0207,
      "reward": 0.21951719745993614,
      "reward_std": 0.08504664897918701,
      "reward_std_all": 0.07870283164083958,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5793465673923492,
      "rewards/forensics_tempsamp_iou_reward": 0.21951719745993614,
      "step": 779
    },
    {
      "completion_length": 45.78125,
      "epoch": 2.0745672436750997,
      "grad_norm": 3.963068962097168,
      "kl": 0.296875,
      "learning_rate": 4.806666666666667e-07,
      "loss": 0.0119,
      "reward": 0.3114546462893486,
      "reward_std": 0.10207319259643555,
      "reward_std_all": 0.11323631927371025,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9576867818832397,
      "rewards/forensics_tempsamp_iou_reward": 0.3114546462893486,
      "step": 780
    },
    {
      "completion_length": 33.703125,
      "epoch": 2.077230359520639,
      "grad_norm": 5.980926513671875,
      "kl": 0.5029296875,
      "learning_rate": 4.8e-07,
      "loss": 0.0201,
      "reward": 0.245530903339386,
      "reward_std": 0.09020311012864113,
      "reward_std_all": 0.09587197750806808,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7758107483386993,
      "rewards/forensics_tempsamp_iou_reward": 0.245530903339386,
      "step": 781
    },
    {
      "completion_length": 38.96875,
      "epoch": 2.0798934753661786,
      "grad_norm": 4.696355819702148,
      "kl": 0.4287109375,
      "learning_rate": 4.793333333333333e-07,
      "loss": 0.0172,
      "reward": 0.284845232963562,
      "reward_std": 0.06805543601512909,
      "reward_std_all": 0.06289765611290932,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9974250495433807,
      "rewards/forensics_tempsamp_iou_reward": 0.284845232963562,
      "step": 782
    },
    {
      "completion_length": 44.46875,
      "epoch": 2.0825565912117177,
      "grad_norm": 5.348301887512207,
      "kl": 0.3046875,
      "learning_rate": 4.786666666666667e-07,
      "loss": 0.0122,
      "reward": 0.2912895679473877,
      "reward_std": 0.12373141944408417,
      "reward_std_all": 0.1208425834774971,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9865109026432037,
      "rewards/forensics_tempsamp_iou_reward": 0.2912895679473877,
      "step": 783
    },
    {
      "completion_length": 35.5625,
      "epoch": 2.085219707057257,
      "grad_norm": 4.852180480957031,
      "kl": 0.26611328125,
      "learning_rate": 4.779999999999999e-07,
      "loss": 0.0107,
      "reward": 0.33774198591709137,
      "reward_std": 0.10911816731095314,
      "reward_std_all": 0.0993361584842205,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9827235639095306,
      "rewards/forensics_tempsamp_iou_reward": 0.33774198591709137,
      "step": 784
    },
    {
      "completion_length": 32.1875,
      "epoch": 2.087882822902796,
      "grad_norm": 6.294316291809082,
      "kl": 0.705078125,
      "learning_rate": 4.773333333333333e-07,
      "loss": 0.0282,
      "reward": 0.22326859831809998,
      "reward_std": 0.10427584499120712,
      "reward_std_all": 0.10015933960676193,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8935875594615936,
      "rewards/forensics_tempsamp_iou_reward": 0.22326859831809998,
      "step": 785
    },
    {
      "completion_length": 36.75,
      "epoch": 2.0905459387483356,
      "grad_norm": 4.391331195831299,
      "kl": 0.5029296875,
      "learning_rate": 4.7666666666666667e-07,
      "loss": 0.0201,
      "reward": 0.23225557804107666,
      "reward_std": 0.06666375324130058,
      "reward_std_all": 0.06345595233142376,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7600111365318298,
      "rewards/forensics_tempsamp_iou_reward": 0.23225557804107666,
      "step": 786
    },
    {
      "completion_length": 60.921875,
      "epoch": 2.0932090545938746,
      "grad_norm": 6.141987323760986,
      "kl": 0.447265625,
      "learning_rate": 4.76e-07,
      "loss": 0.0179,
      "reward": 0.3523828685283661,
      "reward_std": 0.11406201124191284,
      "reward_std_all": 0.1038224846124649,
      "rewards-max/forensics_tempsamp_iou_reward": 0.961597353219986,
      "rewards/forensics_tempsamp_iou_reward": 0.3523828685283661,
      "step": 787
    },
    {
      "completion_length": 39.859375,
      "epoch": 2.095872170439414,
      "grad_norm": 5.846104145050049,
      "kl": 0.4296875,
      "learning_rate": 4.7533333333333333e-07,
      "loss": 0.0172,
      "reward": 0.32590997219085693,
      "reward_std": 0.08377372100949287,
      "reward_std_all": 0.109867874532938,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9729612767696381,
      "rewards/forensics_tempsamp_iou_reward": 0.32590997219085693,
      "step": 788
    },
    {
      "completion_length": 24.78125,
      "epoch": 2.0985352862849536,
      "grad_norm": 7.404065132141113,
      "kl": 0.5205078125,
      "learning_rate": 4.746666666666667e-07,
      "loss": 0.0208,
      "reward": 0.21377234160900116,
      "reward_std": 0.05623985081911087,
      "reward_std_all": 0.06057043373584747,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8093249499797821,
      "rewards/forensics_tempsamp_iou_reward": 0.21377234160900116,
      "step": 789
    },
    {
      "completion_length": 30.609375,
      "epoch": 2.1011984021304926,
      "grad_norm": 8.054015159606934,
      "kl": 0.404296875,
      "learning_rate": 4.7399999999999993e-07,
      "loss": 0.0162,
      "reward": 0.5206746459007263,
      "reward_std": 0.1284557431936264,
      "reward_std_all": 0.13102825358510017,
      "rewards-max/forensics_tempsamp_iou_reward": 0.977223128080368,
      "rewards/forensics_tempsamp_iou_reward": 0.5206746459007263,
      "step": 790
    },
    {
      "completion_length": 45.6875,
      "epoch": 2.103861517976032,
      "grad_norm": 6.153932571411133,
      "kl": 0.60546875,
      "learning_rate": 4.733333333333333e-07,
      "loss": 0.0243,
      "reward": 0.15040885657072067,
      "reward_std": 0.06901400536298752,
      "reward_std_all": 0.08707733824849129,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5602668970823288,
      "rewards/forensics_tempsamp_iou_reward": 0.15040885657072067,
      "step": 791
    },
    {
      "completion_length": 25.875,
      "epoch": 2.106524633821571,
      "grad_norm": 7.088293075561523,
      "kl": 0.3271484375,
      "learning_rate": 4.7266666666666664e-07,
      "loss": 0.0131,
      "reward": 0.403921440243721,
      "reward_std": 0.09618882834911346,
      "reward_std_all": 0.11893000453710556,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9841398894786835,
      "rewards/forensics_tempsamp_iou_reward": 0.403921440243721,
      "step": 792
    },
    {
      "completion_length": 29.390625,
      "epoch": 2.1091877496671105,
      "grad_norm": 5.56721830368042,
      "kl": 0.4267578125,
      "learning_rate": 4.7199999999999994e-07,
      "loss": 0.0171,
      "reward": 0.3348863795399666,
      "reward_std": 0.106162179261446,
      "reward_std_all": 0.10465763509273529,
      "rewards-max/forensics_tempsamp_iou_reward": 0.982917308807373,
      "rewards/forensics_tempsamp_iou_reward": 0.3348863795399666,
      "step": 793
    },
    {
      "completion_length": 55.5625,
      "epoch": 2.11185086551265,
      "grad_norm": 4.6698808670043945,
      "kl": 0.5107421875,
      "learning_rate": 4.713333333333333e-07,
      "loss": 0.0204,
      "reward": 0.2916248142719269,
      "reward_std": 0.10005054622888565,
      "reward_std_all": 0.10066862404346466,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9611640572547913,
      "rewards/forensics_tempsamp_iou_reward": 0.2916248142719269,
      "step": 794
    },
    {
      "completion_length": 42.15625,
      "epoch": 2.114513981358189,
      "grad_norm": 5.183050632476807,
      "kl": 0.5283203125,
      "learning_rate": 4.7066666666666665e-07,
      "loss": 0.0212,
      "reward": 0.40978188812732697,
      "reward_std": 0.10805238410830498,
      "reward_std_all": 0.10372176766395569,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9755744934082031,
      "rewards/forensics_tempsamp_iou_reward": 0.40978188812732697,
      "step": 795
    },
    {
      "completion_length": 46.203125,
      "epoch": 2.1171770972037285,
      "grad_norm": 7.377845764160156,
      "kl": 0.6181640625,
      "learning_rate": 4.6999999999999995e-07,
      "loss": 0.0248,
      "reward": 0.22559157758951187,
      "reward_std": 0.08124585077166557,
      "reward_std_all": 0.10456870123744011,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7893617153167725,
      "rewards/forensics_tempsamp_iou_reward": 0.22559157758951187,
      "step": 796
    },
    {
      "completion_length": 35.34375,
      "epoch": 2.1198402130492675,
      "grad_norm": 5.291655540466309,
      "kl": 0.384765625,
      "learning_rate": 4.693333333333333e-07,
      "loss": 0.0154,
      "reward": 0.34405848383903503,
      "reward_std": 0.06335341930389404,
      "reward_std_all": 0.07027833163738251,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9707766771316528,
      "rewards/forensics_tempsamp_iou_reward": 0.34405848383903503,
      "step": 797
    },
    {
      "completion_length": 36.671875,
      "epoch": 2.122503328894807,
      "grad_norm": 7.224287509918213,
      "kl": 0.4775390625,
      "learning_rate": 4.6866666666666665e-07,
      "loss": 0.0191,
      "reward": 0.25918877869844437,
      "reward_std": 0.11200052499771118,
      "reward_std_all": 0.10206843912601471,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7734777927398682,
      "rewards/forensics_tempsamp_iou_reward": 0.25918877869844437,
      "step": 798
    },
    {
      "completion_length": 36.53125,
      "epoch": 2.125166444740346,
      "grad_norm": 8.524933815002441,
      "kl": 0.5009765625,
      "learning_rate": 4.68e-07,
      "loss": 0.02,
      "reward": 0.18537231534719467,
      "reward_std": 0.08364395331591368,
      "reward_std_all": 0.07763499207794666,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6079783216118813,
      "rewards/forensics_tempsamp_iou_reward": 0.18537231534719467,
      "step": 799
    },
    {
      "completion_length": 31.546875,
      "epoch": 2.1278295605858855,
      "grad_norm": 5.450706958770752,
      "kl": 0.35302734375,
      "learning_rate": 4.673333333333333e-07,
      "loss": 0.0141,
      "reward": 0.15300791710615158,
      "reward_std": 0.053572921082377434,
      "reward_std_all": 0.051050757989287376,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7275339663028717,
      "rewards/forensics_tempsamp_iou_reward": 0.15300791710615158,
      "step": 800
    },
    {
      "completion_length": 46.0625,
      "epoch": 2.130492676431425,
      "grad_norm": 5.011477947235107,
      "kl": 0.30126953125,
      "learning_rate": 4.6666666666666666e-07,
      "loss": 0.0121,
      "reward": 0.16974283754825592,
      "reward_std": 0.04546356946229935,
      "reward_std_all": 0.04333732649683952,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7026733607053757,
      "rewards/forensics_tempsamp_iou_reward": 0.16974283754825592,
      "step": 801
    },
    {
      "completion_length": 42.9375,
      "epoch": 2.133155792276964,
      "grad_norm": 4.932240009307861,
      "kl": 0.3603515625,
      "learning_rate": 4.66e-07,
      "loss": 0.0144,
      "reward": 0.2132987305521965,
      "reward_std": 0.06131386198103428,
      "reward_std_all": 0.0673750676214695,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9273670017719269,
      "rewards/forensics_tempsamp_iou_reward": 0.2132987305521965,
      "step": 802
    },
    {
      "completion_length": 31.203125,
      "epoch": 2.1358189081225034,
      "grad_norm": 6.091581344604492,
      "kl": 0.439453125,
      "learning_rate": 4.653333333333333e-07,
      "loss": 0.0176,
      "reward": 0.30617838352918625,
      "reward_std": 0.06997818872332573,
      "reward_std_all": 0.08030526340007782,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8427419364452362,
      "rewards/forensics_tempsamp_iou_reward": 0.30617838352918625,
      "step": 803
    },
    {
      "completion_length": 59.21875,
      "epoch": 2.1384820239680424,
      "grad_norm": 5.8460893630981445,
      "kl": 0.31640625,
      "learning_rate": 4.6466666666666667e-07,
      "loss": 0.0126,
      "reward": 0.2759132385253906,
      "reward_std": 0.07359474152326584,
      "reward_std_all": 0.0851428285241127,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9452271461486816,
      "rewards/forensics_tempsamp_iou_reward": 0.2759132385253906,
      "step": 804
    },
    {
      "completion_length": 36.921875,
      "epoch": 2.141145139813582,
      "grad_norm": 5.443902015686035,
      "kl": 0.626953125,
      "learning_rate": 4.64e-07,
      "loss": 0.025,
      "reward": 0.2623017728328705,
      "reward_std": 0.10656324028968811,
      "reward_std_all": 0.11406804248690605,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9971055686473846,
      "rewards/forensics_tempsamp_iou_reward": 0.2623017728328705,
      "step": 805
    },
    {
      "completion_length": 45.671875,
      "epoch": 2.1438082556591214,
      "grad_norm": 4.921936988830566,
      "kl": 0.58984375,
      "learning_rate": 4.633333333333333e-07,
      "loss": 0.0236,
      "reward": 0.2273828163743019,
      "reward_std": 0.09076546132564545,
      "reward_std_all": 0.09119147434830666,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9346801340579987,
      "rewards/forensics_tempsamp_iou_reward": 0.2273828163743019,
      "step": 806
    },
    {
      "completion_length": 44.671875,
      "epoch": 2.1464713715046604,
      "grad_norm": 4.898108005523682,
      "kl": 0.4052734375,
      "learning_rate": 4.6266666666666663e-07,
      "loss": 0.0162,
      "reward": 0.24944762885570526,
      "reward_std": 0.09784168377518654,
      "reward_std_all": 0.10053782537579536,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8391370177268982,
      "rewards/forensics_tempsamp_iou_reward": 0.24944762885570526,
      "step": 807
    },
    {
      "completion_length": 36.953125,
      "epoch": 2.1491344873502,
      "grad_norm": 5.260880947113037,
      "kl": 0.4169921875,
      "learning_rate": 4.62e-07,
      "loss": 0.0167,
      "reward": 0.3972347229719162,
      "reward_std": 0.08770488388836384,
      "reward_std_all": 0.08221221156418324,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9685302078723907,
      "rewards/forensics_tempsamp_iou_reward": 0.3972347229719162,
      "step": 808
    },
    {
      "completion_length": 35.296875,
      "epoch": 2.151797603195739,
      "grad_norm": 7.490433692932129,
      "kl": 0.65625,
      "learning_rate": 4.613333333333333e-07,
      "loss": 0.0263,
      "reward": 0.3201395571231842,
      "reward_std": 0.07684233784675598,
      "reward_std_all": 0.08459336310625076,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9803921580314636,
      "rewards/forensics_tempsamp_iou_reward": 0.3201395571231842,
      "step": 809
    },
    {
      "completion_length": 45.25,
      "epoch": 2.1544607190412783,
      "grad_norm": 6.662381649017334,
      "kl": 0.5693359375,
      "learning_rate": 4.6066666666666664e-07,
      "loss": 0.0227,
      "reward": 0.22007624059915543,
      "reward_std": 0.07211453095078468,
      "reward_std_all": 0.07920156046748161,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8439519703388214,
      "rewards/forensics_tempsamp_iou_reward": 0.22007624059915543,
      "step": 810
    },
    {
      "completion_length": 60.03125,
      "epoch": 2.157123834886818,
      "grad_norm": 8.012365341186523,
      "kl": 0.5439453125,
      "learning_rate": 4.6e-07,
      "loss": 0.0217,
      "reward": 0.2333141565322876,
      "reward_std": 0.07449319958686829,
      "reward_std_all": 0.07550327107310295,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7179940044879913,
      "rewards/forensics_tempsamp_iou_reward": 0.2333141565322876,
      "step": 811
    },
    {
      "completion_length": 79.0625,
      "epoch": 2.159786950732357,
      "grad_norm": 5.223391532897949,
      "kl": 0.4619140625,
      "learning_rate": 4.593333333333333e-07,
      "loss": 0.0185,
      "reward": 0.25437435507774353,
      "reward_std": 0.12100997567176819,
      "reward_std_all": 0.13336803019046783,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9360088407993317,
      "rewards/forensics_tempsamp_iou_reward": 0.25437435507774353,
      "step": 812
    },
    {
      "completion_length": 34.6875,
      "epoch": 2.1624500665778963,
      "grad_norm": 7.097558498382568,
      "kl": 0.404296875,
      "learning_rate": 4.5866666666666664e-07,
      "loss": 0.0162,
      "reward": 0.1585419774055481,
      "reward_std": 0.026662983931601048,
      "reward_std_all": 0.028596585616469383,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7716259062290192,
      "rewards/forensics_tempsamp_iou_reward": 0.1585419774055481,
      "step": 813
    },
    {
      "completion_length": 32.5,
      "epoch": 2.1651131824234353,
      "grad_norm": 4.407137393951416,
      "kl": 0.498046875,
      "learning_rate": 4.58e-07,
      "loss": 0.0199,
      "reward": 0.3425905555486679,
      "reward_std": 0.10222461074590683,
      "reward_std_all": 0.12321403250098228,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9748620092868805,
      "rewards/forensics_tempsamp_iou_reward": 0.3425905555486679,
      "step": 814
    },
    {
      "completion_length": 33.453125,
      "epoch": 2.1677762982689748,
      "grad_norm": 4.821294784545898,
      "kl": 0.412109375,
      "learning_rate": 4.573333333333333e-07,
      "loss": 0.0165,
      "reward": 0.33110544085502625,
      "reward_std": 0.08177119493484497,
      "reward_std_all": 0.07917829230427742,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9209187924861908,
      "rewards/forensics_tempsamp_iou_reward": 0.33110544085502625,
      "step": 815
    },
    {
      "completion_length": 79.671875,
      "epoch": 2.170439414114514,
      "grad_norm": 3.7101287841796875,
      "kl": 0.478515625,
      "learning_rate": 4.5666666666666665e-07,
      "loss": 0.0192,
      "reward": 0.17578162252902985,
      "reward_std": 0.07023680955171585,
      "reward_std_all": 0.09050814062356949,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6964200139045715,
      "rewards/forensics_tempsamp_iou_reward": 0.17578162252902985,
      "step": 816
    },
    {
      "completion_length": 48.671875,
      "epoch": 2.1731025299600533,
      "grad_norm": 5.204427242279053,
      "kl": 0.478515625,
      "learning_rate": 4.56e-07,
      "loss": 0.0192,
      "reward": 0.2417317032814026,
      "reward_std": 0.11912758648395538,
      "reward_std_all": 0.13158101588487625,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8889739215373993,
      "rewards/forensics_tempsamp_iou_reward": 0.2417317032814026,
      "step": 817
    },
    {
      "completion_length": 44.59375,
      "epoch": 2.1757656458055927,
      "grad_norm": 7.95546293258667,
      "kl": 0.705078125,
      "learning_rate": 4.553333333333333e-07,
      "loss": 0.0282,
      "reward": 0.15433666855096817,
      "reward_std": 0.10862037539482117,
      "reward_std_all": 0.10609481111168861,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5513628721237183,
      "rewards/forensics_tempsamp_iou_reward": 0.15433666855096817,
      "step": 818
    },
    {
      "completion_length": 38.328125,
      "epoch": 2.1784287616511318,
      "grad_norm": 4.6286516189575195,
      "kl": 0.630859375,
      "learning_rate": 4.5466666666666666e-07,
      "loss": 0.0252,
      "reward": 0.28907206654548645,
      "reward_std": 0.08454467728734016,
      "reward_std_all": 0.08212633058428764,
      "rewards-max/forensics_tempsamp_iou_reward": 0.975452184677124,
      "rewards/forensics_tempsamp_iou_reward": 0.28907206654548645,
      "step": 819
    },
    {
      "completion_length": 51.84375,
      "epoch": 2.181091877496671,
      "grad_norm": 5.177523136138916,
      "kl": 0.35546875,
      "learning_rate": 4.54e-07,
      "loss": 0.0142,
      "reward": 0.2340269833803177,
      "reward_std": 0.08980882167816162,
      "reward_std_all": 0.11138535663485527,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8178012371063232,
      "rewards/forensics_tempsamp_iou_reward": 0.2340269833803177,
      "step": 820
    },
    {
      "completion_length": 33.59375,
      "epoch": 2.1837549933422102,
      "grad_norm": 5.383734226226807,
      "kl": 0.4130859375,
      "learning_rate": 4.5333333333333326e-07,
      "loss": 0.0165,
      "reward": 0.23724743723869324,
      "reward_std": 0.08516483753919601,
      "reward_std_all": 0.08916985243558884,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8341970145702362,
      "rewards/forensics_tempsamp_iou_reward": 0.23724743723869324,
      "step": 821
    },
    {
      "completion_length": 42.203125,
      "epoch": 2.1864181091877497,
      "grad_norm": 6.778890609741211,
      "kl": 0.5078125,
      "learning_rate": 4.526666666666666e-07,
      "loss": 0.0203,
      "reward": 0.3818547874689102,
      "reward_std": 0.11053429543972015,
      "reward_std_all": 0.11877790093421936,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9640117287635803,
      "rewards/forensics_tempsamp_iou_reward": 0.3818547874689102,
      "step": 822
    },
    {
      "completion_length": 32.078125,
      "epoch": 2.1890812250332887,
      "grad_norm": 5.7676682472229,
      "kl": 0.615234375,
      "learning_rate": 4.5199999999999997e-07,
      "loss": 0.0246,
      "reward": 0.1405675671994686,
      "reward_std": 0.05124250054359436,
      "reward_std_all": 0.05962708778679371,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6069179773330688,
      "rewards/forensics_tempsamp_iou_reward": 0.1405675671994686,
      "step": 823
    },
    {
      "completion_length": 89.828125,
      "epoch": 2.191744340878828,
      "grad_norm": 10.022123336791992,
      "kl": 0.6875,
      "learning_rate": 4.5133333333333327e-07,
      "loss": 0.0275,
      "reward": 0.1619940847158432,
      "reward_std": 0.08384360745549202,
      "reward_std_all": 0.10440436005592346,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5606698542833328,
      "rewards/forensics_tempsamp_iou_reward": 0.1619940847158432,
      "step": 824
    },
    {
      "completion_length": 50.71875,
      "epoch": 2.1944074567243677,
      "grad_norm": 6.99566125869751,
      "kl": 0.6787109375,
      "learning_rate": 4.506666666666666e-07,
      "loss": 0.0272,
      "reward": 0.16809745132923126,
      "reward_std": 0.09545797854661942,
      "reward_std_all": 0.10460169985890388,
      "rewards-max/forensics_tempsamp_iou_reward": 0.659407377243042,
      "rewards/forensics_tempsamp_iou_reward": 0.16809745132923126,
      "step": 825
    },
    {
      "completion_length": 63.296875,
      "epoch": 2.1970705725699067,
      "grad_norm": 5.517711162567139,
      "kl": 0.4755859375,
      "learning_rate": 4.5e-07,
      "loss": 0.019,
      "reward": 0.23652932047843933,
      "reward_std": 0.10362926125526428,
      "reward_std_all": 0.10171295329928398,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9759325385093689,
      "rewards/forensics_tempsamp_iou_reward": 0.23652932047843933,
      "step": 826
    },
    {
      "completion_length": 89.65625,
      "epoch": 2.199733688415446,
      "grad_norm": 4.436985015869141,
      "kl": 0.45703125,
      "learning_rate": 4.493333333333333e-07,
      "loss": 0.0183,
      "reward": 0.13801654428243637,
      "reward_std": 0.07088514044880867,
      "reward_std_all": 0.07327225804328918,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5471343994140625,
      "rewards/forensics_tempsamp_iou_reward": 0.13801654428243637,
      "step": 827
    },
    {
      "completion_length": 38.8125,
      "epoch": 2.202396804260985,
      "grad_norm": 6.24175500869751,
      "kl": 0.353515625,
      "learning_rate": 4.4866666666666663e-07,
      "loss": 0.0141,
      "reward": 0.3560017794370651,
      "reward_std": 0.1038176603615284,
      "reward_std_all": 0.11717261001467705,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9212391972541809,
      "rewards/forensics_tempsamp_iou_reward": 0.3560017794370651,
      "step": 828
    },
    {
      "completion_length": 38.765625,
      "epoch": 2.2050599201065246,
      "grad_norm": 6.498541831970215,
      "kl": 0.4072265625,
      "learning_rate": 4.48e-07,
      "loss": 0.0163,
      "reward": 0.3248281329870224,
      "reward_std": 0.12074107676744461,
      "reward_std_all": 0.13251762837171555,
      "rewards-max/forensics_tempsamp_iou_reward": 0.921398937702179,
      "rewards/forensics_tempsamp_iou_reward": 0.3248281329870224,
      "step": 829
    },
    {
      "completion_length": 66.40625,
      "epoch": 2.207723035952064,
      "grad_norm": 4.668221950531006,
      "kl": 0.4365234375,
      "learning_rate": 4.4733333333333334e-07,
      "loss": 0.0175,
      "reward": 0.23823676258325577,
      "reward_std": 0.11507652327418327,
      "reward_std_all": 0.12319403886795044,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8325614333152771,
      "rewards/forensics_tempsamp_iou_reward": 0.23823676258325577,
      "step": 830
    },
    {
      "completion_length": 32.34375,
      "epoch": 2.210386151797603,
      "grad_norm": 7.207661151885986,
      "kl": 0.40234375,
      "learning_rate": 4.4666666666666664e-07,
      "loss": 0.0161,
      "reward": 0.23020119220018387,
      "reward_std": 0.08983100205659866,
      "reward_std_all": 0.09071028605103493,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8419388234615326,
      "rewards/forensics_tempsamp_iou_reward": 0.23020119220018387,
      "step": 831
    },
    {
      "completion_length": 29.640625,
      "epoch": 2.2130492676431426,
      "grad_norm": 5.4816718101501465,
      "kl": 0.5068359375,
      "learning_rate": 4.46e-07,
      "loss": 0.0203,
      "reward": 0.23378518968820572,
      "reward_std": 0.09025180339813232,
      "reward_std_all": 0.08405204117298126,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6597612947225571,
      "rewards/forensics_tempsamp_iou_reward": 0.23378518968820572,
      "step": 832
    },
    {
      "completion_length": 44.96875,
      "epoch": 2.2157123834886816,
      "grad_norm": 5.360189914703369,
      "kl": 0.40625,
      "learning_rate": 4.4533333333333335e-07,
      "loss": 0.0162,
      "reward": 0.38289298117160797,
      "reward_std": 0.08976875245571136,
      "reward_std_all": 0.10048411786556244,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9542421698570251,
      "rewards/forensics_tempsamp_iou_reward": 0.38289298117160797,
      "step": 833
    },
    {
      "completion_length": 30.59375,
      "epoch": 2.218375499334221,
      "grad_norm": 6.067938804626465,
      "kl": 0.2841796875,
      "learning_rate": 4.4466666666666665e-07,
      "loss": 0.0114,
      "reward": 0.3850521072745323,
      "reward_std": 0.06843089312314987,
      "reward_std_all": 0.06481123715639114,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9915531575679779,
      "rewards/forensics_tempsamp_iou_reward": 0.3850521072745323,
      "step": 834
    },
    {
      "completion_length": 34.171875,
      "epoch": 2.2210386151797605,
      "grad_norm": 7.164780139923096,
      "kl": 0.49609375,
      "learning_rate": 4.44e-07,
      "loss": 0.0198,
      "reward": 0.2621460109949112,
      "reward_std": 0.11725062131881714,
      "reward_std_all": 0.11859038472175598,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9444180727005005,
      "rewards/forensics_tempsamp_iou_reward": 0.2621460109949112,
      "step": 835
    },
    {
      "completion_length": 34.015625,
      "epoch": 2.2237017310252996,
      "grad_norm": 4.66994047164917,
      "kl": 0.3388671875,
      "learning_rate": 4.4333333333333336e-07,
      "loss": 0.0136,
      "reward": 0.3629397004842758,
      "reward_std": 0.08053024113178253,
      "reward_std_all": 0.09825421497225761,
      "rewards-max/forensics_tempsamp_iou_reward": 0.981259673833847,
      "rewards/forensics_tempsamp_iou_reward": 0.3629397004842758,
      "step": 836
    },
    {
      "completion_length": 39.234375,
      "epoch": 2.226364846870839,
      "grad_norm": 4.98631477355957,
      "kl": 0.326171875,
      "learning_rate": 4.426666666666666e-07,
      "loss": 0.0131,
      "reward": 0.3676644265651703,
      "reward_std": 0.061146095395088196,
      "reward_std_all": 0.08922955021262169,
      "rewards-max/forensics_tempsamp_iou_reward": 0.997648686170578,
      "rewards/forensics_tempsamp_iou_reward": 0.3676644265651703,
      "step": 837
    },
    {
      "completion_length": 24.90625,
      "epoch": 2.229027962716378,
      "grad_norm": 6.073675632476807,
      "kl": 0.5146484375,
      "learning_rate": 4.4199999999999996e-07,
      "loss": 0.0206,
      "reward": 0.21828176826238632,
      "reward_std": 0.06108706071972847,
      "reward_std_all": 0.05699107423424721,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7255081385374069,
      "rewards/forensics_tempsamp_iou_reward": 0.21828176826238632,
      "step": 838
    },
    {
      "completion_length": 28.40625,
      "epoch": 2.2316910785619175,
      "grad_norm": 6.069279193878174,
      "kl": 0.3671875,
      "learning_rate": 4.413333333333333e-07,
      "loss": 0.0147,
      "reward": 0.3971099704504013,
      "reward_std": 0.07990911230444908,
      "reward_std_all": 0.0820915624499321,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9836382567882538,
      "rewards/forensics_tempsamp_iou_reward": 0.3971099704504013,
      "step": 839
    },
    {
      "completion_length": 27.03125,
      "epoch": 2.2343541944074565,
      "grad_norm": 5.870733261108398,
      "kl": 0.560546875,
      "learning_rate": 4.406666666666666e-07,
      "loss": 0.0224,
      "reward": 0.30318233370780945,
      "reward_std": 0.052640787325799465,
      "reward_std_all": 0.06587598100304604,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8574332892894745,
      "rewards/forensics_tempsamp_iou_reward": 0.30318233370780945,
      "step": 840
    },
    {
      "completion_length": 41.140625,
      "epoch": 2.237017310252996,
      "grad_norm": 4.843681335449219,
      "kl": 0.361328125,
      "learning_rate": 4.3999999999999997e-07,
      "loss": 0.0145,
      "reward": 0.2672797068953514,
      "reward_std": 0.0977373793721199,
      "reward_std_all": 0.0999925397336483,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8866955935955048,
      "rewards/forensics_tempsamp_iou_reward": 0.2672797068953514,
      "step": 841
    },
    {
      "completion_length": 43.453125,
      "epoch": 2.2396804260985355,
      "grad_norm": 7.217106819152832,
      "kl": 0.4140625,
      "learning_rate": 4.393333333333333e-07,
      "loss": 0.0165,
      "reward": 0.2420312762260437,
      "reward_std": 0.07552868686616421,
      "reward_std_all": 0.07264917343854904,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7000000029802322,
      "rewards/forensics_tempsamp_iou_reward": 0.2420312762260437,
      "step": 842
    },
    {
      "completion_length": 34.46875,
      "epoch": 2.2423435419440745,
      "grad_norm": 5.2975969314575195,
      "kl": 0.3935546875,
      "learning_rate": 4.386666666666666e-07,
      "loss": 0.0157,
      "reward": 0.27301549166440964,
      "reward_std": 0.0858931839466095,
      "reward_std_all": 0.08243649825453758,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8995072245597839,
      "rewards/forensics_tempsamp_iou_reward": 0.27301549166440964,
      "step": 843
    },
    {
      "completion_length": 53.375,
      "epoch": 2.245006657789614,
      "grad_norm": 5.8954620361328125,
      "kl": 0.7919921875,
      "learning_rate": 4.38e-07,
      "loss": 0.0316,
      "reward": 0.2185906022787094,
      "reward_std": 0.09180477261543274,
      "reward_std_all": 0.09067746996879578,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8580270409584045,
      "rewards/forensics_tempsamp_iou_reward": 0.2185906022787094,
      "step": 844
    },
    {
      "completion_length": 54.390625,
      "epoch": 2.247669773635153,
      "grad_norm": 6.244306564331055,
      "kl": 0.521484375,
      "learning_rate": 4.3733333333333333e-07,
      "loss": 0.0209,
      "reward": 0.23982834815979004,
      "reward_std": 0.09562679007649422,
      "reward_std_all": 0.10286502540111542,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7969541847705841,
      "rewards/forensics_tempsamp_iou_reward": 0.23982834815979004,
      "step": 845
    },
    {
      "completion_length": 42.90625,
      "epoch": 2.2503328894806924,
      "grad_norm": 7.060049057006836,
      "kl": 0.65234375,
      "learning_rate": 4.3666666666666663e-07,
      "loss": 0.0261,
      "reward": 0.2494412437081337,
      "reward_std": 0.08849278092384338,
      "reward_std_all": 0.0901668593287468,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8033712208271027,
      "rewards/forensics_tempsamp_iou_reward": 0.2494412437081337,
      "step": 846
    },
    {
      "completion_length": 33.4375,
      "epoch": 2.2529960053262315,
      "grad_norm": 6.241236209869385,
      "kl": 0.5341796875,
      "learning_rate": 4.36e-07,
      "loss": 0.0214,
      "reward": 0.35564441978931427,
      "reward_std": 0.1099066361784935,
      "reward_std_all": 0.13103488087654114,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9630320072174072,
      "rewards/forensics_tempsamp_iou_reward": 0.35564441978931427,
      "step": 847
    },
    {
      "completion_length": 31.84375,
      "epoch": 2.255659121171771,
      "grad_norm": 5.575371742248535,
      "kl": 0.609375,
      "learning_rate": 4.3533333333333334e-07,
      "loss": 0.0244,
      "reward": 0.3064082860946655,
      "reward_std": 0.08684190362691879,
      "reward_std_all": 0.0922834500670433,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9661084711551666,
      "rewards/forensics_tempsamp_iou_reward": 0.3064082860946655,
      "step": 848
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.2583222370173104,
      "grad_norm": 5.864142417907715,
      "kl": 0.447265625,
      "learning_rate": 4.3466666666666664e-07,
      "loss": 0.0179,
      "reward": 0.22903825342655182,
      "reward_std": 0.09742971509695053,
      "reward_std_all": 0.09084172546863556,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6937229335308075,
      "rewards/forensics_tempsamp_iou_reward": 0.22903825342655182,
      "step": 849
    },
    {
      "completion_length": 21.71875,
      "epoch": 2.2609853528628494,
      "grad_norm": 8.855807304382324,
      "kl": 0.5888671875,
      "learning_rate": 4.34e-07,
      "loss": 0.0235,
      "reward": 0.32416267693042755,
      "reward_std": 0.09835025668144226,
      "reward_std_all": 0.093924880027771,
      "rewards-max/forensics_tempsamp_iou_reward": 0.952006071805954,
      "rewards/forensics_tempsamp_iou_reward": 0.32416267693042755,
      "step": 850
    },
    {
      "completion_length": 44.796875,
      "epoch": 2.263648468708389,
      "grad_norm": 5.393848419189453,
      "kl": 0.4111328125,
      "learning_rate": 4.3333333333333335e-07,
      "loss": 0.0164,
      "reward": 0.35277026891708374,
      "reward_std": 0.12594353780150414,
      "reward_std_all": 0.13301274180412292,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9494506120681763,
      "rewards/forensics_tempsamp_iou_reward": 0.35277026891708374,
      "step": 851
    },
    {
      "completion_length": 25.15625,
      "epoch": 2.266311584553928,
      "grad_norm": 7.736822128295898,
      "kl": 0.58203125,
      "learning_rate": 4.3266666666666665e-07,
      "loss": 0.0233,
      "reward": 0.390127032995224,
      "reward_std": 0.09702950716018677,
      "reward_std_all": 0.10150617733597755,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9650193750858307,
      "rewards/forensics_tempsamp_iou_reward": 0.390127032995224,
      "step": 852
    },
    {
      "completion_length": 44.609375,
      "epoch": 2.2689747003994674,
      "grad_norm": 6.051205158233643,
      "kl": 0.4990234375,
      "learning_rate": 4.3199999999999995e-07,
      "loss": 0.02,
      "reward": 0.36472825706005096,
      "reward_std": 0.08319973573088646,
      "reward_std_all": 0.10061004012823105,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9691919088363647,
      "rewards/forensics_tempsamp_iou_reward": 0.36472825706005096,
      "step": 853
    },
    {
      "completion_length": 47.90625,
      "epoch": 2.271637816245007,
      "grad_norm": 5.211728572845459,
      "kl": 0.4052734375,
      "learning_rate": 4.313333333333333e-07,
      "loss": 0.0162,
      "reward": 0.2249244824051857,
      "reward_std": 0.06513313204050064,
      "reward_std_all": 0.08202052116394043,
      "rewards-max/forensics_tempsamp_iou_reward": 0.903775542974472,
      "rewards/forensics_tempsamp_iou_reward": 0.2249244824051857,
      "step": 854
    },
    {
      "completion_length": 42.9375,
      "epoch": 2.274300932090546,
      "grad_norm": 5.83925724029541,
      "kl": 0.4375,
      "learning_rate": 4.306666666666666e-07,
      "loss": 0.0175,
      "reward": 0.2567717730998993,
      "reward_std": 0.07487558200955391,
      "reward_std_all": 0.07434553653001785,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7270010411739349,
      "rewards/forensics_tempsamp_iou_reward": 0.2567717730998993,
      "step": 855
    },
    {
      "completion_length": 30.90625,
      "epoch": 2.2769640479360853,
      "grad_norm": 5.352628707885742,
      "kl": 0.630859375,
      "learning_rate": 4.2999999999999996e-07,
      "loss": 0.0252,
      "reward": 0.3297174423933029,
      "reward_std": 0.07313139922916889,
      "reward_std_all": 0.07061155326664448,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7896275222301483,
      "rewards/forensics_tempsamp_iou_reward": 0.3297174423933029,
      "step": 856
    },
    {
      "completion_length": 53.390625,
      "epoch": 2.2796271637816243,
      "grad_norm": 5.9668660163879395,
      "kl": 0.353515625,
      "learning_rate": 4.293333333333333e-07,
      "loss": 0.0141,
      "reward": 0.24308453127741814,
      "reward_std": 0.07300189882516861,
      "reward_std_all": 0.07630341313779354,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8062991499900818,
      "rewards/forensics_tempsamp_iou_reward": 0.24308453127741814,
      "step": 857
    },
    {
      "completion_length": 34.765625,
      "epoch": 2.282290279627164,
      "grad_norm": 4.142416477203369,
      "kl": 0.380859375,
      "learning_rate": 4.286666666666666e-07,
      "loss": 0.0152,
      "reward": 0.33041897416114807,
      "reward_std": 0.12180507928133011,
      "reward_std_all": 0.13378888368606567,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9630503058433533,
      "rewards/forensics_tempsamp_iou_reward": 0.33041897416114807,
      "step": 858
    },
    {
      "completion_length": 55.109375,
      "epoch": 2.2849533954727033,
      "grad_norm": 6.933821201324463,
      "kl": 0.6728515625,
      "learning_rate": 4.2799999999999997e-07,
      "loss": 0.0269,
      "reward": 0.26730482280254364,
      "reward_std": 0.06906852126121521,
      "reward_std_all": 0.07811716198921204,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6666534692049026,
      "rewards/forensics_tempsamp_iou_reward": 0.26730482280254364,
      "step": 859
    },
    {
      "completion_length": 37.71875,
      "epoch": 2.2876165113182423,
      "grad_norm": 10.527251243591309,
      "kl": 0.9453125,
      "learning_rate": 4.273333333333333e-07,
      "loss": 0.0377,
      "reward": 0.30088871717453003,
      "reward_std": 0.1077643632888794,
      "reward_std_all": 0.10984247550368309,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9852348566055298,
      "rewards/forensics_tempsamp_iou_reward": 0.30088871717453003,
      "step": 860
    },
    {
      "completion_length": 54.46875,
      "epoch": 2.2902796271637818,
      "grad_norm": 7.499032497406006,
      "kl": 0.7578125,
      "learning_rate": 4.266666666666667e-07,
      "loss": 0.0303,
      "reward": 0.20920944213867188,
      "reward_std": 0.09408693760633469,
      "reward_std_all": 0.088776595890522,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6474159359931946,
      "rewards/forensics_tempsamp_iou_reward": 0.20920944213867188,
      "step": 861
    },
    {
      "completion_length": 30.78125,
      "epoch": 2.2929427430093208,
      "grad_norm": 5.101268291473389,
      "kl": 0.529296875,
      "learning_rate": 4.26e-07,
      "loss": 0.0212,
      "reward": 0.3111117333173752,
      "reward_std": 0.10912150144577026,
      "reward_std_all": 0.1062324121594429,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9968897104263306,
      "rewards/forensics_tempsamp_iou_reward": 0.3111117333173752,
      "step": 862
    },
    {
      "completion_length": 53.625,
      "epoch": 2.2956058588548602,
      "grad_norm": 5.313534259796143,
      "kl": 0.48046875,
      "learning_rate": 4.2533333333333333e-07,
      "loss": 0.0192,
      "reward": 0.37641242146492004,
      "reward_std": 0.08595328964293003,
      "reward_std_all": 0.11532390490174294,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9893253445625305,
      "rewards/forensics_tempsamp_iou_reward": 0.37641242146492004,
      "step": 863
    },
    {
      "completion_length": 43.75,
      "epoch": 2.2982689747003997,
      "grad_norm": 6.913459777832031,
      "kl": 0.810546875,
      "learning_rate": 4.246666666666667e-07,
      "loss": 0.0324,
      "reward": 0.19047874212265015,
      "reward_std": 0.0817375909537077,
      "reward_std_all": 0.08896639943122864,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5346968024969101,
      "rewards/forensics_tempsamp_iou_reward": 0.19047874212265015,
      "step": 864
    },
    {
      "completion_length": 33.796875,
      "epoch": 2.3009320905459387,
      "grad_norm": 5.0748677253723145,
      "kl": 0.4794921875,
      "learning_rate": 4.24e-07,
      "loss": 0.0192,
      "reward": 0.3903079181909561,
      "reward_std": 0.10754971951246262,
      "reward_std_all": 0.1043216697871685,
      "rewards-max/forensics_tempsamp_iou_reward": 0.992341548204422,
      "rewards/forensics_tempsamp_iou_reward": 0.3903079181909561,
      "step": 865
    },
    {
      "completion_length": 36.078125,
      "epoch": 2.303595206391478,
      "grad_norm": 5.06741189956665,
      "kl": 0.45068359375,
      "learning_rate": 4.2333333333333334e-07,
      "loss": 0.018,
      "reward": 0.1414983794093132,
      "reward_std": 0.07521897926926613,
      "reward_std_all": 0.07504778355360031,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6720583289861679,
      "rewards/forensics_tempsamp_iou_reward": 0.1414983794093132,
      "step": 866
    },
    {
      "completion_length": 40.65625,
      "epoch": 2.306258322237017,
      "grad_norm": 6.157499313354492,
      "kl": 0.568359375,
      "learning_rate": 4.226666666666667e-07,
      "loss": 0.0227,
      "reward": 0.30542197823524475,
      "reward_std": 0.07516897842288017,
      "reward_std_all": 0.07376082241535187,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9834111332893372,
      "rewards/forensics_tempsamp_iou_reward": 0.30542197823524475,
      "step": 867
    },
    {
      "completion_length": 40.3125,
      "epoch": 2.3089214380825567,
      "grad_norm": 6.085176467895508,
      "kl": 0.4990234375,
      "learning_rate": 4.2199999999999994e-07,
      "loss": 0.0199,
      "reward": 0.2988104075193405,
      "reward_std": 0.09029753506183624,
      "reward_std_all": 0.09278979152441025,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9795168042182922,
      "rewards/forensics_tempsamp_iou_reward": 0.2988104075193405,
      "step": 868
    },
    {
      "completion_length": 38.390625,
      "epoch": 2.3115845539280957,
      "grad_norm": 5.566412925720215,
      "kl": 0.4599609375,
      "learning_rate": 4.213333333333333e-07,
      "loss": 0.0184,
      "reward": 0.26651497930288315,
      "reward_std": 0.06788551062345505,
      "reward_std_all": 0.07899314165115356,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9539473652839661,
      "rewards/forensics_tempsamp_iou_reward": 0.26651497930288315,
      "step": 869
    },
    {
      "completion_length": 43.46875,
      "epoch": 2.314247669773635,
      "grad_norm": 4.693703651428223,
      "kl": 0.4921875,
      "learning_rate": 4.2066666666666665e-07,
      "loss": 0.0197,
      "reward": 0.20520421117544174,
      "reward_std": 0.10819090902805328,
      "reward_std_all": 0.11265263706445694,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5929521843791008,
      "rewards/forensics_tempsamp_iou_reward": 0.20520421117544174,
      "step": 870
    },
    {
      "completion_length": 57.9375,
      "epoch": 2.316910785619174,
      "grad_norm": 7.49887752532959,
      "kl": 0.421875,
      "learning_rate": 4.1999999999999995e-07,
      "loss": 0.0169,
      "reward": 0.3187476843595505,
      "reward_std": 0.08186979591846466,
      "reward_std_all": 0.10885677486658096,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9205686450004578,
      "rewards/forensics_tempsamp_iou_reward": 0.3187476843595505,
      "step": 871
    },
    {
      "completion_length": 27.65625,
      "epoch": 2.3195739014647137,
      "grad_norm": 6.075897693634033,
      "kl": 0.58203125,
      "learning_rate": 4.193333333333333e-07,
      "loss": 0.0233,
      "reward": 0.29150501638650894,
      "reward_std": 0.0882476270198822,
      "reward_std_all": 0.08990779891610146,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9797466993331909,
      "rewards/forensics_tempsamp_iou_reward": 0.29150501638650894,
      "step": 872
    },
    {
      "completion_length": 42.84375,
      "epoch": 2.322237017310253,
      "grad_norm": 5.880300521850586,
      "kl": 0.67578125,
      "learning_rate": 4.1866666666666666e-07,
      "loss": 0.027,
      "reward": 0.31014464795589447,
      "reward_std": 0.1456722468137741,
      "reward_std_all": 0.15486204996705055,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8849672973155975,
      "rewards/forensics_tempsamp_iou_reward": 0.31014464795589447,
      "step": 873
    },
    {
      "completion_length": 60.25,
      "epoch": 2.324900133155792,
      "grad_norm": 5.872572422027588,
      "kl": 0.3681640625,
      "learning_rate": 4.1799999999999996e-07,
      "loss": 0.0147,
      "reward": 0.2107282355427742,
      "reward_std": 0.06889225170016289,
      "reward_std_all": 0.0711451880633831,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7432456016540527,
      "rewards/forensics_tempsamp_iou_reward": 0.2107282355427742,
      "step": 874
    },
    {
      "completion_length": 57.25,
      "epoch": 2.3275632490013316,
      "grad_norm": 7.228302478790283,
      "kl": 0.537109375,
      "learning_rate": 4.173333333333333e-07,
      "loss": 0.0215,
      "reward": 0.2964966893196106,
      "reward_std": 0.13694075495004654,
      "reward_std_all": 0.12334014847874641,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9931192696094513,
      "rewards/forensics_tempsamp_iou_reward": 0.2964966893196106,
      "step": 875
    },
    {
      "completion_length": 79.359375,
      "epoch": 2.3302263648468706,
      "grad_norm": 5.679161071777344,
      "kl": 0.4560546875,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.0182,
      "reward": 0.2726709842681885,
      "reward_std": 0.09194334596395493,
      "reward_std_all": 0.09582710266113281,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9665115177631378,
      "rewards/forensics_tempsamp_iou_reward": 0.2726709842681885,
      "step": 876
    },
    {
      "completion_length": 47.765625,
      "epoch": 2.33288948069241,
      "grad_norm": 8.926897048950195,
      "kl": 0.484375,
      "learning_rate": 4.1599999999999997e-07,
      "loss": 0.0194,
      "reward": 0.35726699233055115,
      "reward_std": 0.06896435469388962,
      "reward_std_all": 0.07928021997213364,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9774734377861023,
      "rewards/forensics_tempsamp_iou_reward": 0.35726699233055115,
      "step": 877
    },
    {
      "completion_length": 44.359375,
      "epoch": 2.3355525965379496,
      "grad_norm": 7.470283031463623,
      "kl": 0.5068359375,
      "learning_rate": 4.153333333333333e-07,
      "loss": 0.0203,
      "reward": 0.340262770652771,
      "reward_std": 0.09896693751215935,
      "reward_std_all": 0.09561905637383461,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9465041160583496,
      "rewards/forensics_tempsamp_iou_reward": 0.340262770652771,
      "step": 878
    },
    {
      "completion_length": 38.546875,
      "epoch": 2.3382157123834886,
      "grad_norm": 6.0342254638671875,
      "kl": 0.388671875,
      "learning_rate": 4.146666666666667e-07,
      "loss": 0.0156,
      "reward": 0.2642209827899933,
      "reward_std": 0.07034052349627018,
      "reward_std_all": 0.06544638983905315,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9508727490901947,
      "rewards/forensics_tempsamp_iou_reward": 0.2642209827899933,
      "step": 879
    },
    {
      "completion_length": 69.234375,
      "epoch": 2.340878828229028,
      "grad_norm": 6.43641471862793,
      "kl": 0.62109375,
      "learning_rate": 4.14e-07,
      "loss": 0.0248,
      "reward": 0.23198461532592773,
      "reward_std": 0.09669613093137741,
      "reward_std_all": 0.09531350433826447,
      "rewards-max/forensics_tempsamp_iou_reward": 0.774127870798111,
      "rewards/forensics_tempsamp_iou_reward": 0.23198461532592773,
      "step": 880
    },
    {
      "completion_length": 63.84375,
      "epoch": 2.343541944074567,
      "grad_norm": 5.417963981628418,
      "kl": 0.56640625,
      "learning_rate": 4.1333333333333333e-07,
      "loss": 0.0226,
      "reward": 0.2797366827726364,
      "reward_std": 0.08256861567497253,
      "reward_std_all": 0.083042673766613,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9546680152416229,
      "rewards/forensics_tempsamp_iou_reward": 0.2797366827726364,
      "step": 881
    },
    {
      "completion_length": 74.1875,
      "epoch": 2.3462050599201065,
      "grad_norm": 7.872814655303955,
      "kl": 0.4755859375,
      "learning_rate": 4.126666666666667e-07,
      "loss": 0.019,
      "reward": 0.15053772926330566,
      "reward_std": 0.101092129945755,
      "reward_std_all": 0.10467489808797836,
      "rewards-max/forensics_tempsamp_iou_reward": 0.550555557012558,
      "rewards/forensics_tempsamp_iou_reward": 0.15053772926330566,
      "step": 882
    },
    {
      "completion_length": 34.828125,
      "epoch": 2.348868175765646,
      "grad_norm": 7.422289848327637,
      "kl": 0.427734375,
      "learning_rate": 4.12e-07,
      "loss": 0.0171,
      "reward": 0.3486548215150833,
      "reward_std": 0.06692353636026382,
      "reward_std_all": 0.08048749715089798,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9675783216953278,
      "rewards/forensics_tempsamp_iou_reward": 0.3486548215150833,
      "step": 883
    },
    {
      "completion_length": 39.296875,
      "epoch": 2.351531291611185,
      "grad_norm": 8.192822456359863,
      "kl": 0.412109375,
      "learning_rate": 4.113333333333333e-07,
      "loss": 0.0165,
      "reward": 0.2597833722829819,
      "reward_std": 0.06677369400858879,
      "reward_std_all": 0.06609195284545422,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9006548523902893,
      "rewards/forensics_tempsamp_iou_reward": 0.2597833722829819,
      "step": 884
    },
    {
      "completion_length": 85.15625,
      "epoch": 2.3541944074567245,
      "grad_norm": 6.161942958831787,
      "kl": 0.626953125,
      "learning_rate": 4.1066666666666664e-07,
      "loss": 0.0251,
      "reward": 0.23760729283094406,
      "reward_std": 0.0769445076584816,
      "reward_std_all": 0.07174713164567947,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8584899008274078,
      "rewards/forensics_tempsamp_iou_reward": 0.23760729283094406,
      "step": 885
    },
    {
      "completion_length": 59.578125,
      "epoch": 2.3568575233022635,
      "grad_norm": 6.121006965637207,
      "kl": 0.609375,
      "learning_rate": 4.0999999999999994e-07,
      "loss": 0.0244,
      "reward": 0.24782009422779083,
      "reward_std": 0.08912544697523117,
      "reward_std_all": 0.12383615225553513,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9504286646842957,
      "rewards/forensics_tempsamp_iou_reward": 0.24782009422779083,
      "step": 886
    },
    {
      "completion_length": 111.578125,
      "epoch": 2.359520639147803,
      "grad_norm": 810.1111450195312,
      "kl": 13.123046875,
      "learning_rate": 4.093333333333333e-07,
      "loss": 0.5247,
      "reward": 0.19217262417078018,
      "reward_std": 0.09281064197421074,
      "reward_std_all": 0.08904868736863136,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7614163458347321,
      "rewards/forensics_tempsamp_iou_reward": 0.19217262417078018,
      "step": 887
    },
    {
      "completion_length": 80.59375,
      "epoch": 2.3621837549933424,
      "grad_norm": 7.697288513183594,
      "kl": 0.65625,
      "learning_rate": 4.0866666666666665e-07,
      "loss": 0.0262,
      "reward": 0.3161894232034683,
      "reward_std": 0.11257318779826164,
      "reward_std_all": 0.1141250990331173,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9456469118595123,
      "rewards/forensics_tempsamp_iou_reward": 0.3161894232034683,
      "step": 888
    },
    {
      "completion_length": 30.875,
      "epoch": 2.3648468708388815,
      "grad_norm": 5.407928466796875,
      "kl": 0.33203125,
      "learning_rate": 4.0799999999999995e-07,
      "loss": 0.0132,
      "reward": 0.31990912556648254,
      "reward_std": 0.10487033613026142,
      "reward_std_all": 0.09664353728294373,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8986573219299316,
      "rewards/forensics_tempsamp_iou_reward": 0.31990912556648254,
      "step": 889
    },
    {
      "completion_length": 52.125,
      "epoch": 2.367509986684421,
      "grad_norm": 5.972710609436035,
      "kl": 0.4794921875,
      "learning_rate": 4.073333333333333e-07,
      "loss": 0.0192,
      "reward": 0.34489668905735016,
      "reward_std": 0.09327398985624313,
      "reward_std_all": 0.09650982730090618,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9377104341983795,
      "rewards/forensics_tempsamp_iou_reward": 0.34489668905735016,
      "step": 890
    },
    {
      "completion_length": 49.734375,
      "epoch": 2.37017310252996,
      "grad_norm": 5.205358982086182,
      "kl": 0.4609375,
      "learning_rate": 4.0666666666666666e-07,
      "loss": 0.0184,
      "reward": 0.31038014590740204,
      "reward_std": 0.08258211798965931,
      "reward_std_all": 0.07674823515117168,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9800394475460052,
      "rewards/forensics_tempsamp_iou_reward": 0.31038014590740204,
      "step": 891
    },
    {
      "completion_length": 42.984375,
      "epoch": 2.3728362183754994,
      "grad_norm": 8.083590507507324,
      "kl": 0.44921875,
      "learning_rate": 4.06e-07,
      "loss": 0.018,
      "reward": 0.3600379079580307,
      "reward_std": 0.08767618983983994,
      "reward_std_all": 0.09006603062152863,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9766823947429657,
      "rewards/forensics_tempsamp_iou_reward": 0.3600379079580307,
      "step": 892
    },
    {
      "completion_length": 39.703125,
      "epoch": 2.3754993342210384,
      "grad_norm": 5.462181091308594,
      "kl": 0.43359375,
      "learning_rate": 4.053333333333333e-07,
      "loss": 0.0173,
      "reward": 0.21135813742876053,
      "reward_std": 0.08140208572149277,
      "reward_std_all": 0.0778360441327095,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6037089079618454,
      "rewards/forensics_tempsamp_iou_reward": 0.21135813742876053,
      "step": 893
    },
    {
      "completion_length": 37.625,
      "epoch": 2.378162450066578,
      "grad_norm": 7.13931941986084,
      "kl": 0.4912109375,
      "learning_rate": 4.0466666666666666e-07,
      "loss": 0.0196,
      "reward": 0.22006727755069733,
      "reward_std": 0.10083040967583656,
      "reward_std_all": 0.0956919826567173,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6551251858472824,
      "rewards/forensics_tempsamp_iou_reward": 0.22006727755069733,
      "step": 894
    },
    {
      "completion_length": 39.859375,
      "epoch": 2.3808255659121174,
      "grad_norm": 5.327638149261475,
      "kl": 0.5927734375,
      "learning_rate": 4.04e-07,
      "loss": 0.0237,
      "reward": 0.23377423733472824,
      "reward_std": 0.0845579281449318,
      "reward_std_all": 0.0836777314543724,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8049961924552917,
      "rewards/forensics_tempsamp_iou_reward": 0.23377423733472824,
      "step": 895
    },
    {
      "completion_length": 58.171875,
      "epoch": 2.3834886817576564,
      "grad_norm": 5.474289894104004,
      "kl": 0.421875,
      "learning_rate": 4.033333333333333e-07,
      "loss": 0.0169,
      "reward": 0.2931336909532547,
      "reward_std": 0.08647238463163376,
      "reward_std_all": 0.0848810188472271,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9910996854305267,
      "rewards/forensics_tempsamp_iou_reward": 0.2931336909532547,
      "step": 896
    },
    {
      "completion_length": 27.390625,
      "epoch": 2.386151797603196,
      "grad_norm": 6.4578423500061035,
      "kl": 0.4248046875,
      "learning_rate": 4.0266666666666667e-07,
      "loss": 0.017,
      "reward": 0.3935161828994751,
      "reward_std": 0.1113305352628231,
      "reward_std_all": 0.10223431140184402,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9905429482460022,
      "rewards/forensics_tempsamp_iou_reward": 0.3935161828994751,
      "step": 897
    },
    {
      "completion_length": 47.453125,
      "epoch": 2.388814913448735,
      "grad_norm": 6.243274688720703,
      "kl": 0.3876953125,
      "learning_rate": 4.02e-07,
      "loss": 0.0155,
      "reward": 0.24096472561359406,
      "reward_std": 0.07085787504911423,
      "reward_std_all": 0.10222640261054039,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9648643136024475,
      "rewards/forensics_tempsamp_iou_reward": 0.24096472561359406,
      "step": 898
    },
    {
      "completion_length": 23.328125,
      "epoch": 2.3914780292942743,
      "grad_norm": 6.101861476898193,
      "kl": 0.642578125,
      "learning_rate": 4.0133333333333333e-07,
      "loss": 0.0257,
      "reward": 0.35691919922828674,
      "reward_std": 0.11310684308409691,
      "reward_std_all": 0.13187549263238907,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9892487227916718,
      "rewards/forensics_tempsamp_iou_reward": 0.35691919922828674,
      "step": 899
    },
    {
      "completion_length": 20.90625,
      "epoch": 2.3941411451398134,
      "grad_norm": 5.4951653480529785,
      "kl": 0.4609375,
      "learning_rate": 4.0066666666666663e-07,
      "loss": 0.0185,
      "reward": 0.3251018822193146,
      "reward_std": 0.0836048573255539,
      "reward_std_all": 0.07887672632932663,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8910569846630096,
      "rewards/forensics_tempsamp_iou_reward": 0.3251018822193146,
      "step": 900
    },
    {
      "completion_length": 36.390625,
      "epoch": 2.396804260985353,
      "grad_norm": 6.2277655601501465,
      "kl": 0.326171875,
      "learning_rate": 4e-07,
      "loss": 0.013,
      "reward": 0.3162272274494171,
      "reward_std": 0.08978329598903656,
      "reward_std_all": 0.0913698673248291,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9612706899642944,
      "rewards/forensics_tempsamp_iou_reward": 0.3162272274494171,
      "step": 901
    },
    {
      "completion_length": 32.90625,
      "epoch": 2.3994673768308923,
      "grad_norm": 6.379018783569336,
      "kl": 0.4677734375,
      "learning_rate": 3.993333333333333e-07,
      "loss": 0.0187,
      "reward": 0.37078744173049927,
      "reward_std": 0.14797286689281464,
      "reward_std_all": 0.15021532773971558,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9369885325431824,
      "rewards/forensics_tempsamp_iou_reward": 0.37078744173049927,
      "step": 902
    },
    {
      "completion_length": 41.125,
      "epoch": 2.4021304926764313,
      "grad_norm": 5.905031204223633,
      "kl": 0.396484375,
      "learning_rate": 3.9866666666666664e-07,
      "loss": 0.0159,
      "reward": 0.3491959273815155,
      "reward_std": 0.10671911388635635,
      "reward_std_all": 0.10803025960922241,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8295209109783173,
      "rewards/forensics_tempsamp_iou_reward": 0.3491959273815155,
      "step": 903
    },
    {
      "completion_length": 29.59375,
      "epoch": 2.4047936085219708,
      "grad_norm": 6.569928169250488,
      "kl": 0.5,
      "learning_rate": 3.98e-07,
      "loss": 0.02,
      "reward": 0.18312940001487732,
      "reward_std": 0.08156169950962067,
      "reward_std_all": 0.09251194074749947,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5213315188884735,
      "rewards/forensics_tempsamp_iou_reward": 0.18312940001487732,
      "step": 904
    },
    {
      "completion_length": 46.8125,
      "epoch": 2.40745672436751,
      "grad_norm": 7.611531734466553,
      "kl": 0.5205078125,
      "learning_rate": 3.973333333333333e-07,
      "loss": 0.0208,
      "reward": 0.20239371061325073,
      "reward_std": 0.1311246007680893,
      "reward_std_all": 0.12419088184833527,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8565941154956818,
      "rewards/forensics_tempsamp_iou_reward": 0.20239371061325073,
      "step": 905
    },
    {
      "completion_length": 47.453125,
      "epoch": 2.4101198402130493,
      "grad_norm": 6.315272808074951,
      "kl": 0.6171875,
      "learning_rate": 3.9666666666666665e-07,
      "loss": 0.0247,
      "reward": 0.2783162072300911,
      "reward_std": 0.11246883869171143,
      "reward_std_all": 0.10211987048387527,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9903253614902496,
      "rewards/forensics_tempsamp_iou_reward": 0.2783162072300911,
      "step": 906
    },
    {
      "completion_length": 23.671875,
      "epoch": 2.4127829560585887,
      "grad_norm": 6.787199974060059,
      "kl": 0.392578125,
      "learning_rate": 3.96e-07,
      "loss": 0.0157,
      "reward": 0.4101473540067673,
      "reward_std": 0.06797648593783379,
      "reward_std_all": 0.09025510773062706,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9665538370609283,
      "rewards/forensics_tempsamp_iou_reward": 0.4101473540067673,
      "step": 907
    },
    {
      "completion_length": 51.03125,
      "epoch": 2.4154460719041277,
      "grad_norm": 5.324568748474121,
      "kl": 0.3056640625,
      "learning_rate": 3.953333333333333e-07,
      "loss": 0.0122,
      "reward": 0.16159364581108093,
      "reward_std": 0.06669994629919529,
      "reward_std_all": 0.07256781682372093,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5508778095245361,
      "rewards/forensics_tempsamp_iou_reward": 0.16159364581108093,
      "step": 908
    },
    {
      "completion_length": 37.125,
      "epoch": 2.418109187749667,
      "grad_norm": 5.278027534484863,
      "kl": 0.6640625,
      "learning_rate": 3.9466666666666665e-07,
      "loss": 0.0266,
      "reward": 0.28635431826114655,
      "reward_std": 0.129747673869133,
      "reward_std_all": 0.12999269366264343,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9581988155841827,
      "rewards/forensics_tempsamp_iou_reward": 0.28635431826114655,
      "step": 909
    },
    {
      "completion_length": 22.625,
      "epoch": 2.4207723035952062,
      "grad_norm": 5.719862937927246,
      "kl": 0.3515625,
      "learning_rate": 3.94e-07,
      "loss": 0.014,
      "reward": 0.34562693536281586,
      "reward_std": 0.0915300827473402,
      "reward_std_all": 0.08543691970407963,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9805673658847809,
      "rewards/forensics_tempsamp_iou_reward": 0.34562693536281586,
      "step": 910
    },
    {
      "completion_length": 26.1875,
      "epoch": 2.4234354194407457,
      "grad_norm": 4.396015167236328,
      "kl": 0.37890625,
      "learning_rate": 3.933333333333333e-07,
      "loss": 0.0151,
      "reward": 0.2580402195453644,
      "reward_std": 0.088039081543684,
      "reward_std_all": 0.08196760341525078,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9647796452045441,
      "rewards/forensics_tempsamp_iou_reward": 0.2580402195453644,
      "step": 911
    },
    {
      "completion_length": 25.578125,
      "epoch": 2.426098535286285,
      "grad_norm": 7.271711826324463,
      "kl": 0.697265625,
      "learning_rate": 3.9266666666666666e-07,
      "loss": 0.0279,
      "reward": 0.23424356803297997,
      "reward_std": 0.09910854697227478,
      "reward_std_all": 0.09917643293738365,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6814102530479431,
      "rewards/forensics_tempsamp_iou_reward": 0.23424356803297997,
      "step": 912
    },
    {
      "completion_length": 16.921875,
      "epoch": 2.428761651131824,
      "grad_norm": 8.416755676269531,
      "kl": 0.4755859375,
      "learning_rate": 3.92e-07,
      "loss": 0.019,
      "reward": 0.3718503713607788,
      "reward_std": 0.08317684754729271,
      "reward_std_all": 0.09072202816605568,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9858193397521973,
      "rewards/forensics_tempsamp_iou_reward": 0.3718503713607788,
      "step": 913
    },
    {
      "completion_length": 19.859375,
      "epoch": 2.4314247669773636,
      "grad_norm": 5.389240264892578,
      "kl": 0.6650390625,
      "learning_rate": 3.913333333333333e-07,
      "loss": 0.0266,
      "reward": 0.21016937494277954,
      "reward_std": 0.07779601216316223,
      "reward_std_all": 0.07631132565438747,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6318692713975906,
      "rewards/forensics_tempsamp_iou_reward": 0.21016937494277954,
      "step": 914
    },
    {
      "completion_length": 31.65625,
      "epoch": 2.4340878828229027,
      "grad_norm": 6.9287214279174805,
      "kl": 0.44921875,
      "learning_rate": 3.906666666666666e-07,
      "loss": 0.018,
      "reward": 0.32024459540843964,
      "reward_std": 0.12095813825726509,
      "reward_std_all": 0.11026836186647415,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9503490328788757,
      "rewards/forensics_tempsamp_iou_reward": 0.32024459540843964,
      "step": 915
    },
    {
      "completion_length": 15.3125,
      "epoch": 2.436750998668442,
      "grad_norm": 8.413219451904297,
      "kl": 0.51953125,
      "learning_rate": 3.8999999999999997e-07,
      "loss": 0.0208,
      "reward": 0.2926226556301117,
      "reward_std": 0.064427325502038,
      "reward_std_all": 0.07116679847240448,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9112149477005005,
      "rewards/forensics_tempsamp_iou_reward": 0.2926226556301117,
      "step": 916
    },
    {
      "completion_length": 46.390625,
      "epoch": 2.4394141145139816,
      "grad_norm": 5.2434892654418945,
      "kl": 0.4814453125,
      "learning_rate": 3.8933333333333327e-07,
      "loss": 0.0193,
      "reward": 0.2416633442044258,
      "reward_std": 0.083160150796175,
      "reward_std_all": 0.08645711466670036,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6709554195404053,
      "rewards/forensics_tempsamp_iou_reward": 0.2416633442044258,
      "step": 917
    },
    {
      "completion_length": 23.390625,
      "epoch": 2.4420772303595206,
      "grad_norm": 5.822032928466797,
      "kl": 0.4267578125,
      "learning_rate": 3.8866666666666663e-07,
      "loss": 0.017,
      "reward": 0.28407566994428635,
      "reward_std": 0.06637348234653473,
      "reward_std_all": 0.07720473781228065,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9972527325153351,
      "rewards/forensics_tempsamp_iou_reward": 0.28407566994428635,
      "step": 918
    },
    {
      "completion_length": 33.421875,
      "epoch": 2.44474034620506,
      "grad_norm": 5.743676662445068,
      "kl": 0.5029296875,
      "learning_rate": 3.88e-07,
      "loss": 0.0201,
      "reward": 0.1755281314253807,
      "reward_std": 0.06401870027184486,
      "reward_std_all": 0.07202344760298729,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6603794097900391,
      "rewards/forensics_tempsamp_iou_reward": 0.1755281314253807,
      "step": 919
    },
    {
      "completion_length": 24.78125,
      "epoch": 2.447403462050599,
      "grad_norm": 5.890852928161621,
      "kl": 0.537109375,
      "learning_rate": 3.873333333333333e-07,
      "loss": 0.0214,
      "reward": 0.21520961821079254,
      "reward_std": 0.08665891364216805,
      "reward_std_all": 0.0775669775903225,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8752976357936859,
      "rewards/forensics_tempsamp_iou_reward": 0.21520961821079254,
      "step": 920
    },
    {
      "completion_length": 37.625,
      "epoch": 2.4500665778961386,
      "grad_norm": 7.4759135246276855,
      "kl": 0.498046875,
      "learning_rate": 3.8666666666666664e-07,
      "loss": 0.0199,
      "reward": 0.31555457040667534,
      "reward_std": 0.07890450954437256,
      "reward_std_all": 0.08734829723834991,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7392700761556625,
      "rewards/forensics_tempsamp_iou_reward": 0.31555457040667534,
      "step": 921
    },
    {
      "completion_length": 27.953125,
      "epoch": 2.4527296937416776,
      "grad_norm": 5.370597839355469,
      "kl": 0.5908203125,
      "learning_rate": 3.86e-07,
      "loss": 0.0236,
      "reward": 0.29374152421951294,
      "reward_std": 0.066433385014534,
      "reward_std_all": 0.06908846646547318,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8850435316562653,
      "rewards/forensics_tempsamp_iou_reward": 0.29374152421951294,
      "step": 922
    },
    {
      "completion_length": 57.296875,
      "epoch": 2.455392809587217,
      "grad_norm": 7.997413635253906,
      "kl": 0.724609375,
      "learning_rate": 3.8533333333333334e-07,
      "loss": 0.029,
      "reward": 0.3122631087899208,
      "reward_std": 0.18239450454711914,
      "reward_std_all": 0.17765148729085922,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9160052835941315,
      "rewards/forensics_tempsamp_iou_reward": 0.3122631087899208,
      "step": 923
    },
    {
      "completion_length": 40.953125,
      "epoch": 2.458055925432756,
      "grad_norm": 6.11227560043335,
      "kl": 0.4755859375,
      "learning_rate": 3.8466666666666664e-07,
      "loss": 0.019,
      "reward": 0.24444279074668884,
      "reward_std": 0.06222495809197426,
      "reward_std_all": 0.06827708706259727,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8690700829029083,
      "rewards/forensics_tempsamp_iou_reward": 0.24444279074668884,
      "step": 924
    },
    {
      "completion_length": 28.8125,
      "epoch": 2.4607190412782955,
      "grad_norm": 6.307239532470703,
      "kl": 0.41015625,
      "learning_rate": 3.84e-07,
      "loss": 0.0164,
      "reward": 0.25870002806186676,
      "reward_std": 0.07798102498054504,
      "reward_std_all": 0.08726754784584045,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5796363800764084,
      "rewards/forensics_tempsamp_iou_reward": 0.25870002806186676,
      "step": 925
    },
    {
      "completion_length": 33.328125,
      "epoch": 2.463382157123835,
      "grad_norm": 5.33472204208374,
      "kl": 0.533203125,
      "learning_rate": 3.8333333333333335e-07,
      "loss": 0.0213,
      "reward": 0.2547152563929558,
      "reward_std": 0.09182215668261051,
      "reward_std_all": 0.09452477656304836,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9833013415336609,
      "rewards/forensics_tempsamp_iou_reward": 0.2547152563929558,
      "step": 926
    },
    {
      "completion_length": 28.625,
      "epoch": 2.466045272969374,
      "grad_norm": 6.795247554779053,
      "kl": 0.4228515625,
      "learning_rate": 3.8266666666666665e-07,
      "loss": 0.0169,
      "reward": 0.39052383601665497,
      "reward_std": 0.09515273198485374,
      "reward_std_all": 0.08440715819597244,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9564394056797028,
      "rewards/forensics_tempsamp_iou_reward": 0.39052383601665497,
      "step": 927
    },
    {
      "completion_length": 43.359375,
      "epoch": 2.4687083888149135,
      "grad_norm": 6.186047554016113,
      "kl": 0.34765625,
      "learning_rate": 3.82e-07,
      "loss": 0.0139,
      "reward": 0.34587013721466064,
      "reward_std": 0.061279296875,
      "reward_std_all": 0.06971406564116478,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9621800482273102,
      "rewards/forensics_tempsamp_iou_reward": 0.34587013721466064,
      "step": 928
    },
    {
      "completion_length": 19.140625,
      "epoch": 2.4713715046604525,
      "grad_norm": 6.984518051147461,
      "kl": 0.28515625,
      "learning_rate": 3.8133333333333336e-07,
      "loss": 0.0114,
      "reward": 0.2887784317135811,
      "reward_std": 0.03935005608946085,
      "reward_std_all": 0.05447212606668472,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7284739464521408,
      "rewards/forensics_tempsamp_iou_reward": 0.2887784317135811,
      "step": 929
    },
    {
      "completion_length": 34.0625,
      "epoch": 2.474034620505992,
      "grad_norm": 6.061614036560059,
      "kl": 0.4248046875,
      "learning_rate": 3.8066666666666666e-07,
      "loss": 0.017,
      "reward": 0.36988624930381775,
      "reward_std": 0.07536594569683075,
      "reward_std_all": 0.0916764922440052,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9981927573680878,
      "rewards/forensics_tempsamp_iou_reward": 0.36988624930381775,
      "step": 930
    },
    {
      "completion_length": 28.0625,
      "epoch": 2.4766977363515315,
      "grad_norm": 5.410984992980957,
      "kl": 0.4990234375,
      "learning_rate": 3.7999999999999996e-07,
      "loss": 0.0199,
      "reward": 0.44282734394073486,
      "reward_std": 0.11304908245801926,
      "reward_std_all": 0.10991458967328072,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9820887446403503,
      "rewards/forensics_tempsamp_iou_reward": 0.44282734394073486,
      "step": 931
    },
    {
      "completion_length": 20.359375,
      "epoch": 2.4793608521970705,
      "grad_norm": 7.453874588012695,
      "kl": 0.2412109375,
      "learning_rate": 3.793333333333333e-07,
      "loss": 0.0096,
      "reward": 0.4445604383945465,
      "reward_std": 0.0851195640861988,
      "reward_std_all": 0.07988504320383072,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9962776601314545,
      "rewards/forensics_tempsamp_iou_reward": 0.4445604383945465,
      "step": 932
    },
    {
      "completion_length": 43.65625,
      "epoch": 2.48202396804261,
      "grad_norm": 5.5403218269348145,
      "kl": 0.318359375,
      "learning_rate": 3.786666666666666e-07,
      "loss": 0.0128,
      "reward": 0.34625986218452454,
      "reward_std": 0.12256054952740669,
      "reward_std_all": 0.11625587195158005,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9873116910457611,
      "rewards/forensics_tempsamp_iou_reward": 0.34625986218452454,
      "step": 933
    },
    {
      "completion_length": 38.765625,
      "epoch": 2.484687083888149,
      "grad_norm": 8.244385719299316,
      "kl": 0.408203125,
      "learning_rate": 3.7799999999999997e-07,
      "loss": 0.0163,
      "reward": 0.33232617378234863,
      "reward_std": 0.12219759821891785,
      "reward_std_all": 0.1354283057153225,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9980689883232117,
      "rewards/forensics_tempsamp_iou_reward": 0.33232617378234863,
      "step": 934
    },
    {
      "completion_length": 52.21875,
      "epoch": 2.4873501997336884,
      "grad_norm": 10.12186336517334,
      "kl": 0.4296875,
      "learning_rate": 3.773333333333333e-07,
      "loss": 0.0172,
      "reward": 0.2772548496723175,
      "reward_std": 0.048910679295659065,
      "reward_std_all": 0.04917246848344803,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7044702172279358,
      "rewards/forensics_tempsamp_iou_reward": 0.2772548496723175,
      "step": 935
    },
    {
      "completion_length": 51.75,
      "epoch": 2.490013315579228,
      "grad_norm": 10.682339668273926,
      "kl": 0.86328125,
      "learning_rate": 3.766666666666666e-07,
      "loss": 0.0345,
      "reward": 0.2878611236810684,
      "reward_std": 0.09276965260505676,
      "reward_std_all": 0.08309646509587765,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9957351982593536,
      "rewards/forensics_tempsamp_iou_reward": 0.2878611236810684,
      "step": 936
    },
    {
      "completion_length": 37.4375,
      "epoch": 2.492676431424767,
      "grad_norm": 5.840885639190674,
      "kl": 0.556640625,
      "learning_rate": 3.76e-07,
      "loss": 0.0222,
      "reward": 0.24873816967010498,
      "reward_std": 0.07511314004659653,
      "reward_std_all": 0.06300836428999901,
      "rewards-max/forensics_tempsamp_iou_reward": 0.738631084561348,
      "rewards/forensics_tempsamp_iou_reward": 0.24873816967010498,
      "step": 937
    },
    {
      "completion_length": 41.8125,
      "epoch": 2.4953395472703064,
      "grad_norm": 69.11461639404297,
      "kl": 3.2470703125,
      "learning_rate": 3.7533333333333333e-07,
      "loss": 0.1301,
      "reward": 0.3440101146697998,
      "reward_std": 0.09194321557879448,
      "reward_std_all": 0.0853460356593132,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9949193596839905,
      "rewards/forensics_tempsamp_iou_reward": 0.3440101146697998,
      "step": 938
    },
    {
      "completion_length": 69.390625,
      "epoch": 2.4980026631158454,
      "grad_norm": 5.784310340881348,
      "kl": 0.62109375,
      "learning_rate": 3.7466666666666663e-07,
      "loss": 0.0248,
      "reward": 0.2946823090314865,
      "reward_std": 0.10934996232390404,
      "reward_std_all": 0.1072925552725792,
      "rewards-max/forensics_tempsamp_iou_reward": 0.81742724776268,
      "rewards/forensics_tempsamp_iou_reward": 0.2946823090314865,
      "step": 939
    },
    {
      "completion_length": 41.0625,
      "epoch": 2.500665778961385,
      "grad_norm": 7.327580451965332,
      "kl": 0.701171875,
      "learning_rate": 3.74e-07,
      "loss": 0.0281,
      "reward": 0.3621450215578079,
      "reward_std": 0.08033144846558571,
      "reward_std_all": 0.07477520406246185,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9819150269031525,
      "rewards/forensics_tempsamp_iou_reward": 0.3621450215578079,
      "step": 940
    },
    {
      "completion_length": 53.25,
      "epoch": 2.5033288948069243,
      "grad_norm": 6.704052925109863,
      "kl": 0.427734375,
      "learning_rate": 3.7333333333333334e-07,
      "loss": 0.0171,
      "reward": 0.4722006171941757,
      "reward_std": 0.11716754734516144,
      "reward_std_all": 0.12117672711610794,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9672055840492249,
      "rewards/forensics_tempsamp_iou_reward": 0.4722006171941757,
      "step": 941
    },
    {
      "completion_length": 39.296875,
      "epoch": 2.5059920106524634,
      "grad_norm": 5.885894775390625,
      "kl": 0.6669921875,
      "learning_rate": 3.7266666666666664e-07,
      "loss": 0.0267,
      "reward": 0.23138877749443054,
      "reward_std": 0.06203344650566578,
      "reward_std_all": 0.06342266499996185,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7508503496646881,
      "rewards/forensics_tempsamp_iou_reward": 0.23138877749443054,
      "step": 942
    },
    {
      "completion_length": 63.375,
      "epoch": 2.508655126498003,
      "grad_norm": 4.820390224456787,
      "kl": 0.4951171875,
      "learning_rate": 3.72e-07,
      "loss": 0.0198,
      "reward": 0.26044774800539017,
      "reward_std": 0.06620652601122856,
      "reward_std_all": 0.0632922500371933,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9867225587368011,
      "rewards/forensics_tempsamp_iou_reward": 0.26044774800539017,
      "step": 943
    },
    {
      "completion_length": 41.8125,
      "epoch": 2.511318242343542,
      "grad_norm": 7.124270439147949,
      "kl": 0.4326171875,
      "learning_rate": 3.7133333333333335e-07,
      "loss": 0.0173,
      "reward": 0.3784666508436203,
      "reward_std": 0.04590476118028164,
      "reward_std_all": 0.048056794330477715,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9836035072803497,
      "rewards/forensics_tempsamp_iou_reward": 0.3784666508436203,
      "step": 944
    },
    {
      "completion_length": 37.046875,
      "epoch": 2.5139813581890813,
      "grad_norm": 13.065274238586426,
      "kl": 1.126953125,
      "learning_rate": 3.7066666666666665e-07,
      "loss": 0.0451,
      "reward": 0.23879273980855942,
      "reward_std": 0.052393365651369095,
      "reward_std_all": 0.06046467833220959,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9919524788856506,
      "rewards/forensics_tempsamp_iou_reward": 0.23879273980855942,
      "step": 945
    },
    {
      "completion_length": 58.15625,
      "epoch": 2.5166444740346208,
      "grad_norm": 4.060372829437256,
      "kl": 0.41748046875,
      "learning_rate": 3.7e-07,
      "loss": 0.0167,
      "reward": 0.2830073684453964,
      "reward_std": 0.09457614459097385,
      "reward_std_all": 0.08746591210365295,
      "rewards-max/forensics_tempsamp_iou_reward": 0.91104856133461,
      "rewards/forensics_tempsamp_iou_reward": 0.2830073684453964,
      "step": 946
    },
    {
      "completion_length": 42.34375,
      "epoch": 2.51930758988016,
      "grad_norm": 10.219375610351562,
      "kl": 0.5107421875,
      "learning_rate": 3.693333333333333e-07,
      "loss": 0.0205,
      "reward": 0.34564320743083954,
      "reward_std": 0.11518901214003563,
      "reward_std_all": 0.10970089212059975,
      "rewards-max/forensics_tempsamp_iou_reward": 0.987097293138504,
      "rewards/forensics_tempsamp_iou_reward": 0.34564320743083954,
      "step": 947
    },
    {
      "completion_length": 66.625,
      "epoch": 2.521970705725699,
      "grad_norm": 6.688936710357666,
      "kl": 0.3720703125,
      "learning_rate": 3.686666666666666e-07,
      "loss": 0.0149,
      "reward": 0.2344306856393814,
      "reward_std": 0.05011484771966934,
      "reward_std_all": 0.05269969813525677,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9903306066989899,
      "rewards/forensics_tempsamp_iou_reward": 0.2344306856393814,
      "step": 948
    },
    {
      "completion_length": 46.484375,
      "epoch": 2.5246338215712383,
      "grad_norm": 5.394065856933594,
      "kl": 0.4853515625,
      "learning_rate": 3.6799999999999996e-07,
      "loss": 0.0194,
      "reward": 0.25250137597322464,
      "reward_std": 0.06935447081923485,
      "reward_std_all": 0.06982208602130413,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9754960238933563,
      "rewards/forensics_tempsamp_iou_reward": 0.25250137597322464,
      "step": 949
    },
    {
      "completion_length": 48.71875,
      "epoch": 2.5272969374167777,
      "grad_norm": 6.006037712097168,
      "kl": 0.509765625,
      "learning_rate": 3.673333333333333e-07,
      "loss": 0.0204,
      "reward": 0.2692633867263794,
      "reward_std": 0.05686100386083126,
      "reward_std_all": 0.05868298560380936,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7521964907646179,
      "rewards/forensics_tempsamp_iou_reward": 0.2692633867263794,
      "step": 950
    },
    {
      "completion_length": 25.125,
      "epoch": 2.5299600532623168,
      "grad_norm": 6.561493396759033,
      "kl": 0.615234375,
      "learning_rate": 3.666666666666666e-07,
      "loss": 0.0246,
      "reward": 0.24900680780410767,
      "reward_std": 0.03287660237401724,
      "reward_std_all": 0.03964542597532272,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6213109344244003,
      "rewards/forensics_tempsamp_iou_reward": 0.24900680780410767,
      "step": 951
    },
    {
      "completion_length": 54.1875,
      "epoch": 2.5326231691078562,
      "grad_norm": 5.721921920776367,
      "kl": 0.564453125,
      "learning_rate": 3.6599999999999997e-07,
      "loss": 0.0226,
      "reward": 0.1744316816329956,
      "reward_std": 0.05773061513900757,
      "reward_std_all": 0.0651552639901638,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5176575183868408,
      "rewards/forensics_tempsamp_iou_reward": 0.1744316816329956,
      "step": 952
    },
    {
      "completion_length": 38.875,
      "epoch": 2.5352862849533953,
      "grad_norm": 7.515442848205566,
      "kl": 0.3740234375,
      "learning_rate": 3.653333333333333e-07,
      "loss": 0.015,
      "reward": 0.1854657605290413,
      "reward_std": 0.05420428887009621,
      "reward_std_all": 0.052685339003801346,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6184080243110657,
      "rewards/forensics_tempsamp_iou_reward": 0.1854657605290413,
      "step": 953
    },
    {
      "completion_length": 33.078125,
      "epoch": 2.5379494007989347,
      "grad_norm": 6.751503944396973,
      "kl": 0.45703125,
      "learning_rate": 3.646666666666666e-07,
      "loss": 0.0183,
      "reward": 0.3058309778571129,
      "reward_std": 0.09385723993182182,
      "reward_std_all": 0.09576119109988213,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9913918673992157,
      "rewards/forensics_tempsamp_iou_reward": 0.3058309778571129,
      "step": 954
    },
    {
      "completion_length": 58.328125,
      "epoch": 2.540612516644474,
      "grad_norm": 252.77223205566406,
      "kl": 7.0546875,
      "learning_rate": 3.64e-07,
      "loss": 0.2817,
      "reward": 0.20067497342824936,
      "reward_std": 0.06229238957166672,
      "reward_std_all": 0.08440165966749191,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5196428596973419,
      "rewards/forensics_tempsamp_iou_reward": 0.20067497342824936,
      "step": 955
    },
    {
      "completion_length": 77.734375,
      "epoch": 2.543275632490013,
      "grad_norm": 5.908304691314697,
      "kl": 0.5771484375,
      "learning_rate": 3.6333333333333333e-07,
      "loss": 0.0231,
      "reward": 0.24154987931251526,
      "reward_std": 0.06236104294657707,
      "reward_std_all": 0.07651265524327755,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9718327820301056,
      "rewards/forensics_tempsamp_iou_reward": 0.24154987931251526,
      "step": 956
    },
    {
      "completion_length": 57.703125,
      "epoch": 2.5459387483355527,
      "grad_norm": 6.564271450042725,
      "kl": 0.44921875,
      "learning_rate": 3.626666666666667e-07,
      "loss": 0.0179,
      "reward": 0.2456197589635849,
      "reward_std": 0.07918777503073215,
      "reward_std_all": 0.07331929728388786,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6316663026809692,
      "rewards/forensics_tempsamp_iou_reward": 0.2456197589635849,
      "step": 957
    },
    {
      "completion_length": 31.40625,
      "epoch": 2.5486018641810917,
      "grad_norm": 6.664885997772217,
      "kl": 0.4453125,
      "learning_rate": 3.62e-07,
      "loss": 0.0178,
      "reward": 0.2742168679833412,
      "reward_std": 0.0738939568400383,
      "reward_std_all": 0.07367544993758202,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6793407201766968,
      "rewards/forensics_tempsamp_iou_reward": 0.2742168679833412,
      "step": 958
    },
    {
      "completion_length": 65.390625,
      "epoch": 2.551264980026631,
      "grad_norm": 6.19660758972168,
      "kl": 0.736328125,
      "learning_rate": 3.6133333333333334e-07,
      "loss": 0.0294,
      "reward": 0.3511640131473541,
      "reward_std": 0.14138377085328102,
      "reward_std_all": 0.12734385579824448,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9809378385543823,
      "rewards/forensics_tempsamp_iou_reward": 0.3511640131473541,
      "step": 959
    },
    {
      "completion_length": 66.890625,
      "epoch": 2.5539280958721706,
      "grad_norm": 6.318402290344238,
      "kl": 0.92578125,
      "learning_rate": 3.606666666666667e-07,
      "loss": 0.0371,
      "reward": 0.22571583837270737,
      "reward_std": 0.09404229372739792,
      "reward_std_all": 0.09671733528375626,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8571685552597046,
      "rewards/forensics_tempsamp_iou_reward": 0.22571583837270737,
      "step": 960
    },
    {
      "completion_length": 46.9375,
      "epoch": 2.5565912117177096,
      "grad_norm": 5.446471691131592,
      "kl": 0.4130859375,
      "learning_rate": 3.6e-07,
      "loss": 0.0165,
      "reward": 0.3447328358888626,
      "reward_std": 0.10741181671619415,
      "reward_std_all": 0.10821763426065445,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8861005306243896,
      "rewards/forensics_tempsamp_iou_reward": 0.3447328358888626,
      "step": 961
    },
    {
      "completion_length": 26.484375,
      "epoch": 2.559254327563249,
      "grad_norm": 7.953296661376953,
      "kl": 0.513671875,
      "learning_rate": 3.5933333333333335e-07,
      "loss": 0.0205,
      "reward": 0.21060866862535477,
      "reward_std": 0.06149807199835777,
      "reward_std_all": 0.06173959746956825,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6481359452009201,
      "rewards/forensics_tempsamp_iou_reward": 0.21060866862535477,
      "step": 962
    },
    {
      "completion_length": 27.140625,
      "epoch": 2.561917443408788,
      "grad_norm": 8.28184986114502,
      "kl": 0.982421875,
      "learning_rate": 3.5866666666666665e-07,
      "loss": 0.0394,
      "reward": 0.3776250183582306,
      "reward_std": 0.09987549856305122,
      "reward_std_all": 0.09996691718697548,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9977628588676453,
      "rewards/forensics_tempsamp_iou_reward": 0.3776250183582306,
      "step": 963
    },
    {
      "completion_length": 67.84375,
      "epoch": 2.5645805592543276,
      "grad_norm": 12.450820922851562,
      "kl": 0.970703125,
      "learning_rate": 3.5799999999999995e-07,
      "loss": 0.0388,
      "reward": 0.22640091180801392,
      "reward_std": 0.12606889754533768,
      "reward_std_all": 0.1145310141146183,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9016752541065216,
      "rewards/forensics_tempsamp_iou_reward": 0.22640091180801392,
      "step": 964
    },
    {
      "completion_length": 53.25,
      "epoch": 2.567243675099867,
      "grad_norm": 5.916290283203125,
      "kl": 0.40234375,
      "learning_rate": 3.573333333333333e-07,
      "loss": 0.0161,
      "reward": 0.25607719272375107,
      "reward_std": 0.07659618929028511,
      "reward_std_all": 0.07919362373650074,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9702380895614624,
      "rewards/forensics_tempsamp_iou_reward": 0.25607719272375107,
      "step": 965
    },
    {
      "completion_length": 37.828125,
      "epoch": 2.569906790945406,
      "grad_norm": 7.34855842590332,
      "kl": 0.501953125,
      "learning_rate": 3.5666666666666666e-07,
      "loss": 0.0201,
      "reward": 0.19295836240053177,
      "reward_std": 0.06412428431212902,
      "reward_std_all": 0.07812713086605072,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5955947786569595,
      "rewards/forensics_tempsamp_iou_reward": 0.19295836240053177,
      "step": 966
    },
    {
      "completion_length": 22.53125,
      "epoch": 2.5725699067909455,
      "grad_norm": 5.974421501159668,
      "kl": 0.439453125,
      "learning_rate": 3.5599999999999996e-07,
      "loss": 0.0176,
      "reward": 0.33358214795589447,
      "reward_std": 0.07093764282763004,
      "reward_std_all": 0.1031434964388609,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9705592095851898,
      "rewards/forensics_tempsamp_iou_reward": 0.33358214795589447,
      "step": 967
    },
    {
      "completion_length": 20.9375,
      "epoch": 2.5752330226364846,
      "grad_norm": 6.125857830047607,
      "kl": 0.62109375,
      "learning_rate": 3.553333333333333e-07,
      "loss": 0.0249,
      "reward": 0.29045017063617706,
      "reward_std": 0.11740705370903015,
      "reward_std_all": 0.12096959352493286,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8519510924816132,
      "rewards/forensics_tempsamp_iou_reward": 0.29045017063617706,
      "step": 968
    },
    {
      "completion_length": 38.15625,
      "epoch": 2.577896138482024,
      "grad_norm": 6.3913044929504395,
      "kl": 0.75,
      "learning_rate": 3.5466666666666667e-07,
      "loss": 0.0299,
      "reward": 0.2489553689956665,
      "reward_std": 0.08736833557486534,
      "reward_std_all": 0.0872049555182457,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9753044247627258,
      "rewards/forensics_tempsamp_iou_reward": 0.2489553689956665,
      "step": 969
    },
    {
      "completion_length": 40.4375,
      "epoch": 2.5805592543275635,
      "grad_norm": 6.142925262451172,
      "kl": 0.455078125,
      "learning_rate": 3.5399999999999997e-07,
      "loss": 0.0183,
      "reward": 0.3954838365316391,
      "reward_std": 0.07545557618141174,
      "reward_std_all": 0.08833637088537216,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9779246747493744,
      "rewards/forensics_tempsamp_iou_reward": 0.3954838365316391,
      "step": 970
    },
    {
      "completion_length": 25.25,
      "epoch": 2.5832223701731025,
      "grad_norm": 6.380289077758789,
      "kl": 0.5751953125,
      "learning_rate": 3.533333333333333e-07,
      "loss": 0.023,
      "reward": 0.3173702210187912,
      "reward_std": 0.05658508837223053,
      "reward_std_all": 0.07894808053970337,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9676689505577087,
      "rewards/forensics_tempsamp_iou_reward": 0.3173702210187912,
      "step": 971
    },
    {
      "completion_length": 42.328125,
      "epoch": 2.5858854860186415,
      "grad_norm": 6.16045618057251,
      "kl": 0.677734375,
      "learning_rate": 3.526666666666667e-07,
      "loss": 0.0271,
      "reward": 0.19116999208927155,
      "reward_std": 0.08925343677401543,
      "reward_std_all": 0.08122733235359192,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9473315179347992,
      "rewards/forensics_tempsamp_iou_reward": 0.19116999208927155,
      "step": 972
    },
    {
      "completion_length": 47.703125,
      "epoch": 2.588548601864181,
      "grad_norm": 5.7039337158203125,
      "kl": 0.5068359375,
      "learning_rate": 3.52e-07,
      "loss": 0.0203,
      "reward": 0.24119698256254196,
      "reward_std": 0.08429114520549774,
      "reward_std_all": 0.09251216053962708,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6671982556581497,
      "rewards/forensics_tempsamp_iou_reward": 0.24119698256254196,
      "step": 973
    },
    {
      "completion_length": 32.21875,
      "epoch": 2.5912117177097205,
      "grad_norm": 8.448829650878906,
      "kl": 0.4306640625,
      "learning_rate": 3.5133333333333333e-07,
      "loss": 0.0172,
      "reward": 0.3213553726673126,
      "reward_std": 0.09118778631091118,
      "reward_std_all": 0.10907182469964027,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8416694700717926,
      "rewards/forensics_tempsamp_iou_reward": 0.3213553726673126,
      "step": 974
    },
    {
      "completion_length": 36.5,
      "epoch": 2.5938748335552595,
      "grad_norm": 6.013685703277588,
      "kl": 0.58984375,
      "learning_rate": 3.506666666666667e-07,
      "loss": 0.0235,
      "reward": 0.22557908296585083,
      "reward_std": 0.05762239918112755,
      "reward_std_all": 0.07233357056975365,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7547276318073273,
      "rewards/forensics_tempsamp_iou_reward": 0.22557908296585083,
      "step": 975
    },
    {
      "completion_length": 43.828125,
      "epoch": 2.596537949400799,
      "grad_norm": 4.915727615356445,
      "kl": 0.58984375,
      "learning_rate": 3.5e-07,
      "loss": 0.0236,
      "reward": 0.3333210349082947,
      "reward_std": 0.078699741512537,
      "reward_std_all": 0.0867949053645134,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9905875623226166,
      "rewards/forensics_tempsamp_iou_reward": 0.3333210349082947,
      "step": 976
    },
    {
      "completion_length": 30.03125,
      "epoch": 2.599201065246338,
      "grad_norm": 6.7379326820373535,
      "kl": 0.451171875,
      "learning_rate": 3.4933333333333334e-07,
      "loss": 0.0181,
      "reward": 0.33543406426906586,
      "reward_std": 0.08703918009996414,
      "reward_std_all": 0.08465433865785599,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8595041334629059,
      "rewards/forensics_tempsamp_iou_reward": 0.33543406426906586,
      "step": 977
    },
    {
      "completion_length": 48.921875,
      "epoch": 2.6018641810918774,
      "grad_norm": 6.665098190307617,
      "kl": 0.4453125,
      "learning_rate": 3.4866666666666664e-07,
      "loss": 0.0178,
      "reward": 0.3066778779029846,
      "reward_std": 0.09177310392260551,
      "reward_std_all": 0.08684222400188446,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7625000178813934,
      "rewards/forensics_tempsamp_iou_reward": 0.3066778779029846,
      "step": 978
    },
    {
      "completion_length": 42.65625,
      "epoch": 2.604527296937417,
      "grad_norm": 5.011829853057861,
      "kl": 0.625,
      "learning_rate": 3.4799999999999994e-07,
      "loss": 0.025,
      "reward": 0.3362949937582016,
      "reward_std": 0.07262931391596794,
      "reward_std_all": 0.10622439160943031,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9781609177589417,
      "rewards/forensics_tempsamp_iou_reward": 0.3362949937582016,
      "step": 979
    },
    {
      "completion_length": 34.859375,
      "epoch": 2.607190412782956,
      "grad_norm": 8.118428230285645,
      "kl": 0.716796875,
      "learning_rate": 3.473333333333333e-07,
      "loss": 0.0286,
      "reward": 0.1847938448190689,
      "reward_std": 0.07887633144855499,
      "reward_std_all": 0.08274958282709122,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7379354983568192,
      "rewards/forensics_tempsamp_iou_reward": 0.1847938448190689,
      "step": 980
    },
    {
      "completion_length": 40.65625,
      "epoch": 2.6098535286284954,
      "grad_norm": 4.8700151443481445,
      "kl": 0.638671875,
      "learning_rate": 3.4666666666666665e-07,
      "loss": 0.0256,
      "reward": 0.2858477607369423,
      "reward_std": 0.0971948429942131,
      "reward_std_all": 0.08936383947730064,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9717786610126495,
      "rewards/forensics_tempsamp_iou_reward": 0.2858477607369423,
      "step": 981
    },
    {
      "completion_length": 26.828125,
      "epoch": 2.6125166444740344,
      "grad_norm": 5.252438068389893,
      "kl": 0.44140625,
      "learning_rate": 3.4599999999999995e-07,
      "loss": 0.0176,
      "reward": 0.2972627431154251,
      "reward_std": 0.07582629658281803,
      "reward_std_all": 0.0706704631447792,
      "rewards-max/forensics_tempsamp_iou_reward": 0.831769585609436,
      "rewards/forensics_tempsamp_iou_reward": 0.2972627431154251,
      "step": 982
    },
    {
      "completion_length": 17.8125,
      "epoch": 2.615179760319574,
      "grad_norm": 6.838748455047607,
      "kl": 0.462890625,
      "learning_rate": 3.453333333333333e-07,
      "loss": 0.0185,
      "reward": 0.2565012201666832,
      "reward_std": 0.046172309666872025,
      "reward_std_all": 0.045774360187351704,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9936606287956238,
      "rewards/forensics_tempsamp_iou_reward": 0.2565012201666832,
      "step": 983
    },
    {
      "completion_length": 24.25,
      "epoch": 2.6178428761651134,
      "grad_norm": 6.2594218254089355,
      "kl": 0.578125,
      "learning_rate": 3.4466666666666666e-07,
      "loss": 0.0231,
      "reward": 0.33059607446193695,
      "reward_std": 0.08465803414583206,
      "reward_std_all": 0.08342967182397842,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9919320642948151,
      "rewards/forensics_tempsamp_iou_reward": 0.33059607446193695,
      "step": 984
    },
    {
      "completion_length": 24.390625,
      "epoch": 2.6205059920106524,
      "grad_norm": 5.340566635131836,
      "kl": 0.3974609375,
      "learning_rate": 3.4399999999999996e-07,
      "loss": 0.0159,
      "reward": 0.30077409744262695,
      "reward_std": 0.11494114995002747,
      "reward_std_all": 0.11344562098383904,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8952624797821045,
      "rewards/forensics_tempsamp_iou_reward": 0.30077409744262695,
      "step": 985
    },
    {
      "completion_length": 35.21875,
      "epoch": 2.623169107856192,
      "grad_norm": 5.590181827545166,
      "kl": 0.470703125,
      "learning_rate": 3.433333333333333e-07,
      "loss": 0.0189,
      "reward": 0.28907522559165955,
      "reward_std": 0.05976417288184166,
      "reward_std_all": 0.07222860679030418,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9543659090995789,
      "rewards/forensics_tempsamp_iou_reward": 0.28907522559165955,
      "step": 986
    },
    {
      "completion_length": 27.984375,
      "epoch": 2.625832223701731,
      "grad_norm": 6.179065227508545,
      "kl": 0.365234375,
      "learning_rate": 3.4266666666666666e-07,
      "loss": 0.0146,
      "reward": 0.25790348649024963,
      "reward_std": 0.07658427953720093,
      "reward_std_all": 0.08050088211894035,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7030423432588577,
      "rewards/forensics_tempsamp_iou_reward": 0.25790348649024963,
      "step": 987
    },
    {
      "completion_length": 31.71875,
      "epoch": 2.6284953395472703,
      "grad_norm": 5.1240644454956055,
      "kl": 0.4111328125,
      "learning_rate": 3.42e-07,
      "loss": 0.0164,
      "reward": 0.22565945982933044,
      "reward_std": 0.08213484659790993,
      "reward_std_all": 0.07987243682146072,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7421759366989136,
      "rewards/forensics_tempsamp_iou_reward": 0.22565945982933044,
      "step": 988
    },
    {
      "completion_length": 44.359375,
      "epoch": 2.63115845539281,
      "grad_norm": 6.971096038818359,
      "kl": 0.3759765625,
      "learning_rate": 3.413333333333333e-07,
      "loss": 0.0151,
      "reward": 0.2666497677564621,
      "reward_std": 0.0672510378062725,
      "reward_std_all": 0.06989135965704918,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6760550737380981,
      "rewards/forensics_tempsamp_iou_reward": 0.2666497677564621,
      "step": 989
    },
    {
      "completion_length": 43.609375,
      "epoch": 2.633821571238349,
      "grad_norm": 6.278562545776367,
      "kl": 0.4521484375,
      "learning_rate": 3.4066666666666667e-07,
      "loss": 0.0181,
      "reward": 0.23590926826000214,
      "reward_std": 0.0545866247266531,
      "reward_std_all": 0.0699337087571621,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9913194477558136,
      "rewards/forensics_tempsamp_iou_reward": 0.23590926826000214,
      "step": 990
    },
    {
      "completion_length": 31.921875,
      "epoch": 2.6364846870838883,
      "grad_norm": 6.648685932159424,
      "kl": 0.443359375,
      "learning_rate": 3.4000000000000003e-07,
      "loss": 0.0178,
      "reward": 0.17625433579087257,
      "reward_std": 0.03350256197154522,
      "reward_std_all": 0.07115729711949825,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6083401590585709,
      "rewards/forensics_tempsamp_iou_reward": 0.17625433579087257,
      "step": 991
    },
    {
      "completion_length": 39.671875,
      "epoch": 2.6391478029294273,
      "grad_norm": 4.709296703338623,
      "kl": 0.4384765625,
      "learning_rate": 3.3933333333333333e-07,
      "loss": 0.0175,
      "reward": 0.3734120726585388,
      "reward_std": 0.07333650067448616,
      "reward_std_all": 0.08245967328548431,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9780050218105316,
      "rewards/forensics_tempsamp_iou_reward": 0.3734120726585388,
      "step": 992
    },
    {
      "completion_length": 51.484375,
      "epoch": 2.6418109187749668,
      "grad_norm": 4.397525310516357,
      "kl": 0.5078125,
      "learning_rate": 3.386666666666667e-07,
      "loss": 0.0203,
      "reward": 0.30992943048477173,
      "reward_std": 0.09154203534126282,
      "reward_std_all": 0.09107815101742744,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9804059863090515,
      "rewards/forensics_tempsamp_iou_reward": 0.30992943048477173,
      "step": 993
    },
    {
      "completion_length": 28.34375,
      "epoch": 2.6444740346205062,
      "grad_norm": 7.53649377822876,
      "kl": 0.5869140625,
      "learning_rate": 3.38e-07,
      "loss": 0.0235,
      "reward": 0.30876074731349945,
      "reward_std": 0.06724560260772705,
      "reward_std_all": 0.06967539712786674,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9313282072544098,
      "rewards/forensics_tempsamp_iou_reward": 0.30876074731349945,
      "step": 994
    },
    {
      "completion_length": 42.9375,
      "epoch": 2.6471371504660453,
      "grad_norm": 5.218784809112549,
      "kl": 0.48828125,
      "learning_rate": 3.373333333333333e-07,
      "loss": 0.0195,
      "reward": 0.32770322263240814,
      "reward_std": 0.11738253012299538,
      "reward_std_all": 0.11971966177225113,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9880183041095734,
      "rewards/forensics_tempsamp_iou_reward": 0.32770322263240814,
      "step": 995
    },
    {
      "completion_length": 62.59375,
      "epoch": 2.6498002663115847,
      "grad_norm": 4.918638706207275,
      "kl": 0.5390625,
      "learning_rate": 3.3666666666666664e-07,
      "loss": 0.0216,
      "reward": 0.20582620799541473,
      "reward_std": 0.0955886971205473,
      "reward_std_all": 0.08875411748886108,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8593451380729675,
      "rewards/forensics_tempsamp_iou_reward": 0.20582620799541473,
      "step": 996
    },
    {
      "completion_length": 59.28125,
      "epoch": 2.6524633821571237,
      "grad_norm": 5.636105537414551,
      "kl": 0.5693359375,
      "learning_rate": 3.36e-07,
      "loss": 0.0227,
      "reward": 0.2693793550133705,
      "reward_std": 0.09514287672936916,
      "reward_std_all": 0.08982370793819427,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7676282227039337,
      "rewards/forensics_tempsamp_iou_reward": 0.2693793550133705,
      "step": 997
    },
    {
      "completion_length": 38.46875,
      "epoch": 2.655126498002663,
      "grad_norm": 6.659896373748779,
      "kl": 0.6162109375,
      "learning_rate": 3.353333333333333e-07,
      "loss": 0.0247,
      "reward": 0.41722919046878815,
      "reward_std": 0.11047535017132759,
      "reward_std_all": 0.1044410951435566,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9278696775436401,
      "rewards/forensics_tempsamp_iou_reward": 0.41722919046878815,
      "step": 998
    },
    {
      "completion_length": 42.4375,
      "epoch": 2.6577896138482027,
      "grad_norm": 5.606866836547852,
      "kl": 0.599609375,
      "learning_rate": 3.3466666666666665e-07,
      "loss": 0.024,
      "reward": 0.4536001533269882,
      "reward_std": 0.11948133260011673,
      "reward_std_all": 0.11603637039661407,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9900184571743011,
      "rewards/forensics_tempsamp_iou_reward": 0.4536001533269882,
      "step": 999
    },
    {
      "completion_length": 32.390625,
      "epoch": 2.6604527296937417,
      "grad_norm": 6.356039524078369,
      "kl": 0.62890625,
      "learning_rate": 3.34e-07,
      "loss": 0.0252,
      "reward": 0.3892970234155655,
      "reward_std": 0.04442036524415016,
      "reward_std_all": 0.07239383645355701,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9916722178459167,
      "rewards/forensics_tempsamp_iou_reward": 0.3892970234155655,
      "step": 1000
    },
    {
      "completion_length": 70.109375,
      "epoch": 2.6631158455392807,
      "grad_norm": 4.0008087158203125,
      "kl": 0.6103515625,
      "learning_rate": 3.333333333333333e-07,
      "loss": 0.0245,
      "reward": 0.38170285522937775,
      "reward_std": 0.12065034732222557,
      "reward_std_all": 0.12301044166088104,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9731794893741608,
      "rewards/forensics_tempsamp_iou_reward": 0.38170285522937775,
      "step": 1001
    },
    {
      "completion_length": 36.421875,
      "epoch": 2.66577896138482,
      "grad_norm": 5.899105072021484,
      "kl": 0.5107421875,
      "learning_rate": 3.3266666666666665e-07,
      "loss": 0.0204,
      "reward": 0.42207056283950806,
      "reward_std": 0.06395825743675232,
      "reward_std_all": 0.0686293113976717,
      "rewards-max/forensics_tempsamp_iou_reward": 0.994429349899292,
      "rewards/forensics_tempsamp_iou_reward": 0.42207056283950806,
      "step": 1002
    },
    {
      "completion_length": 41.96875,
      "epoch": 2.6684420772303596,
      "grad_norm": 5.696472644805908,
      "kl": 0.78125,
      "learning_rate": 3.32e-07,
      "loss": 0.0313,
      "reward": 0.2534469962120056,
      "reward_std": 0.057186832651495934,
      "reward_std_all": 0.06556735374033451,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8999465703964233,
      "rewards/forensics_tempsamp_iou_reward": 0.2534469962120056,
      "step": 1003
    },
    {
      "completion_length": 50.703125,
      "epoch": 2.6711051930758987,
      "grad_norm": 11.314850807189941,
      "kl": 0.921875,
      "learning_rate": 3.313333333333333e-07,
      "loss": 0.037,
      "reward": 0.29028770327568054,
      "reward_std": 0.09779027849435806,
      "reward_std_all": 0.08428268507122993,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9633740782737732,
      "rewards/forensics_tempsamp_iou_reward": 0.29028770327568054,
      "step": 1004
    },
    {
      "completion_length": 77.0625,
      "epoch": 2.673768308921438,
      "grad_norm": 7.3065714836120605,
      "kl": 0.583984375,
      "learning_rate": 3.3066666666666666e-07,
      "loss": 0.0234,
      "reward": 0.3869807720184326,
      "reward_std": 0.043634163215756416,
      "reward_std_all": 0.04337439686059952,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9747927188873291,
      "rewards/forensics_tempsamp_iou_reward": 0.3869807720184326,
      "step": 1005
    },
    {
      "completion_length": 57.453125,
      "epoch": 2.676431424766977,
      "grad_norm": 6.1997551918029785,
      "kl": 0.67578125,
      "learning_rate": 3.3e-07,
      "loss": 0.027,
      "reward": 0.4087709039449692,
      "reward_std": 0.08508365228772163,
      "reward_std_all": 0.07537402771413326,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9910555481910706,
      "rewards/forensics_tempsamp_iou_reward": 0.4087709039449692,
      "step": 1006
    },
    {
      "completion_length": 74.71875,
      "epoch": 2.6790945406125166,
      "grad_norm": 6.1492509841918945,
      "kl": 0.6298828125,
      "learning_rate": 3.293333333333333e-07,
      "loss": 0.0252,
      "reward": 0.27826617658138275,
      "reward_std": 0.08629635907709599,
      "reward_std_all": 0.1043778732419014,
      "rewards-max/forensics_tempsamp_iou_reward": 0.976684182882309,
      "rewards/forensics_tempsamp_iou_reward": 0.27826617658138275,
      "step": 1007
    },
    {
      "completion_length": 83.75,
      "epoch": 2.681757656458056,
      "grad_norm": 8.03593635559082,
      "kl": 0.900390625,
      "learning_rate": 3.2866666666666667e-07,
      "loss": 0.0361,
      "reward": 0.25339002907276154,
      "reward_std": 0.12035540118813515,
      "reward_std_all": 0.12707314640283585,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9896039366722107,
      "rewards/forensics_tempsamp_iou_reward": 0.25339002907276154,
      "step": 1008
    },
    {
      "completion_length": 42.671875,
      "epoch": 2.684420772303595,
      "grad_norm": 4.246826171875,
      "kl": 0.3974609375,
      "learning_rate": 3.28e-07,
      "loss": 0.0159,
      "reward": 0.29050345718860626,
      "reward_std": 0.038707561790943146,
      "reward_std_all": 0.03458742797374725,
      "rewards-max/forensics_tempsamp_iou_reward": 0.988812655210495,
      "rewards/forensics_tempsamp_iou_reward": 0.29050345718860626,
      "step": 1009
    },
    {
      "completion_length": 57.65625,
      "epoch": 2.6870838881491346,
      "grad_norm": 6.675752639770508,
      "kl": 0.576171875,
      "learning_rate": 3.2733333333333327e-07,
      "loss": 0.023,
      "reward": 0.30860334634780884,
      "reward_std": 0.059035757556557655,
      "reward_std_all": 0.0545190405100584,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9783520698547363,
      "rewards/forensics_tempsamp_iou_reward": 0.30860334634780884,
      "step": 1010
    },
    {
      "completion_length": 41.859375,
      "epoch": 2.6897470039946736,
      "grad_norm": 6.053644180297852,
      "kl": 0.4521484375,
      "learning_rate": 3.2666666666666663e-07,
      "loss": 0.0181,
      "reward": 0.3537987917661667,
      "reward_std": 0.04723053053021431,
      "reward_std_all": 0.043859174475073814,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9525345861911774,
      "rewards/forensics_tempsamp_iou_reward": 0.3537987917661667,
      "step": 1011
    },
    {
      "completion_length": 47.5625,
      "epoch": 2.692410119840213,
      "grad_norm": 6.0329155921936035,
      "kl": 0.40625,
      "learning_rate": 3.26e-07,
      "loss": 0.0162,
      "reward": 0.28444603085517883,
      "reward_std": 0.09686193615198135,
      "reward_std_all": 0.10299382731318474,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9638460278511047,
      "rewards/forensics_tempsamp_iou_reward": 0.28444603085517883,
      "step": 1012
    },
    {
      "completion_length": 31.515625,
      "epoch": 2.6950732356857525,
      "grad_norm": 14.029623985290527,
      "kl": 0.880859375,
      "learning_rate": 3.253333333333333e-07,
      "loss": 0.0353,
      "reward": 0.21229057013988495,
      "reward_std": 0.028658791445195675,
      "reward_std_all": 0.038003990426659584,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7146829813718796,
      "rewards/forensics_tempsamp_iou_reward": 0.21229057013988495,
      "step": 1013
    },
    {
      "completion_length": 70.8125,
      "epoch": 2.6977363515312915,
      "grad_norm": 5.704909801483154,
      "kl": 0.4248046875,
      "learning_rate": 3.2466666666666664e-07,
      "loss": 0.017,
      "reward": 0.21491631865501404,
      "reward_std": 0.07215389609336853,
      "reward_std_all": 0.07394881919026375,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7166125625371933,
      "rewards/forensics_tempsamp_iou_reward": 0.21491631865501404,
      "step": 1014
    },
    {
      "completion_length": 57.5,
      "epoch": 2.700399467376831,
      "grad_norm": 6.348427772521973,
      "kl": 0.3232421875,
      "learning_rate": 3.24e-07,
      "loss": 0.013,
      "reward": 0.3180218040943146,
      "reward_std": 0.08496332168579102,
      "reward_std_all": 0.08189404383301735,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9964240193367004,
      "rewards/forensics_tempsamp_iou_reward": 0.3180218040943146,
      "step": 1015
    },
    {
      "completion_length": 37.90625,
      "epoch": 2.70306258322237,
      "grad_norm": 6.859036922454834,
      "kl": 0.564453125,
      "learning_rate": 3.233333333333333e-07,
      "loss": 0.0226,
      "reward": 0.18374799937009811,
      "reward_std": 0.01918818149715662,
      "reward_std_all": 0.01917664334177971,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6002224683761597,
      "rewards/forensics_tempsamp_iou_reward": 0.18374799937009811,
      "step": 1016
    },
    {
      "completion_length": 19.109375,
      "epoch": 2.7057256990679095,
      "grad_norm": 6.589335918426514,
      "kl": 0.2802734375,
      "learning_rate": 3.2266666666666664e-07,
      "loss": 0.0112,
      "reward": 0.3809352517127991,
      "reward_std": 0.04315316304564476,
      "reward_std_all": 0.042172614485025406,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9949630498886108,
      "rewards/forensics_tempsamp_iou_reward": 0.3809352517127991,
      "step": 1017
    },
    {
      "completion_length": 59.765625,
      "epoch": 2.708388814913449,
      "grad_norm": 6.387355804443359,
      "kl": 0.5185546875,
      "learning_rate": 3.22e-07,
      "loss": 0.0207,
      "reward": 0.2316252514719963,
      "reward_std": 0.08725753240287304,
      "reward_std_all": 0.08476256020367146,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9632448554039001,
      "rewards/forensics_tempsamp_iou_reward": 0.2316252514719963,
      "step": 1018
    },
    {
      "completion_length": 22.609375,
      "epoch": 2.711051930758988,
      "grad_norm": 6.533299446105957,
      "kl": 0.650390625,
      "learning_rate": 3.2133333333333335e-07,
      "loss": 0.026,
      "reward": 0.39448580145835876,
      "reward_std": 0.06116886530071497,
      "reward_std_all": 0.05745343305170536,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9876568019390106,
      "rewards/forensics_tempsamp_iou_reward": 0.39448580145835876,
      "step": 1019
    },
    {
      "completion_length": 55.875,
      "epoch": 2.7137150466045274,
      "grad_norm": 5.541659832000732,
      "kl": 0.5400390625,
      "learning_rate": 3.2066666666666665e-07,
      "loss": 0.0216,
      "reward": 0.43568500876426697,
      "reward_std": 0.13501647859811783,
      "reward_std_all": 0.17866604030132294,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9941160082817078,
      "rewards/forensics_tempsamp_iou_reward": 0.43568500876426697,
      "step": 1020
    },
    {
      "completion_length": 37.46875,
      "epoch": 2.7163781624500665,
      "grad_norm": 6.564080238342285,
      "kl": 0.513671875,
      "learning_rate": 3.2e-07,
      "loss": 0.0206,
      "reward": 0.33975230157375336,
      "reward_std": 0.07769939303398132,
      "reward_std_all": 0.07753438502550125,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9836063385009766,
      "rewards/forensics_tempsamp_iou_reward": 0.33975230157375336,
      "step": 1021
    },
    {
      "completion_length": 19.734375,
      "epoch": 2.719041278295606,
      "grad_norm": 5.197702407836914,
      "kl": 0.439453125,
      "learning_rate": 3.1933333333333336e-07,
      "loss": 0.0176,
      "reward": 0.4938911646604538,
      "reward_std": 0.06892986595630646,
      "reward_std_all": 0.07396415621042252,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9795134663581848,
      "rewards/forensics_tempsamp_iou_reward": 0.4938911646604538,
      "step": 1022
    },
    {
      "completion_length": 33.78125,
      "epoch": 2.7217043941411454,
      "grad_norm": 6.782731056213379,
      "kl": 0.38671875,
      "learning_rate": 3.1866666666666666e-07,
      "loss": 0.0154,
      "reward": 0.3450550436973572,
      "reward_std": 0.05235268548130989,
      "reward_std_all": 0.08931868523359299,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9985145330429077,
      "rewards/forensics_tempsamp_iou_reward": 0.3450550436973572,
      "step": 1023
    },
    {
      "completion_length": 16.59375,
      "epoch": 2.7243675099866844,
      "grad_norm": 5.945672035217285,
      "kl": 0.6982421875,
      "learning_rate": 3.18e-07,
      "loss": 0.0279,
      "reward": 0.20264064520597458,
      "reward_std": 0.07513284496963024,
      "reward_std_all": 0.06646958738565445,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8078208565711975,
      "rewards/forensics_tempsamp_iou_reward": 0.20264064520597458,
      "step": 1024
    },
    {
      "completion_length": 36.421875,
      "epoch": 2.7270306258322234,
      "grad_norm": 7.514369010925293,
      "kl": 0.4619140625,
      "learning_rate": 3.173333333333333e-07,
      "loss": 0.0185,
      "reward": 0.3581099808216095,
      "reward_std": 0.1361307092010975,
      "reward_std_all": 0.12814082577824593,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9961835443973541,
      "rewards/forensics_tempsamp_iou_reward": 0.3581099808216095,
      "step": 1025
    },
    {
      "completion_length": 57.359375,
      "epoch": 2.729693741677763,
      "grad_norm": 4.282844543457031,
      "kl": 0.4453125,
      "learning_rate": 3.166666666666666e-07,
      "loss": 0.0178,
      "reward": 0.26015686243772507,
      "reward_std": 0.1250225529074669,
      "reward_std_all": 0.11622392013669014,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8387716710567474,
      "rewards/forensics_tempsamp_iou_reward": 0.26015686243772507,
      "step": 1026
    },
    {
      "completion_length": 48.234375,
      "epoch": 2.7323568575233024,
      "grad_norm": 7.920551300048828,
      "kl": 0.6796875,
      "learning_rate": 3.1599999999999997e-07,
      "loss": 0.0272,
      "reward": 0.18253637105226517,
      "reward_std": 0.05385348200798035,
      "reward_std_all": 0.04812268726527691,
      "rewards-max/forensics_tempsamp_iou_reward": 0.795391708612442,
      "rewards/forensics_tempsamp_iou_reward": 0.18253637105226517,
      "step": 1027
    },
    {
      "completion_length": 41.59375,
      "epoch": 2.7350199733688414,
      "grad_norm": 6.17524528503418,
      "kl": 0.462890625,
      "learning_rate": 3.153333333333333e-07,
      "loss": 0.0185,
      "reward": 0.31797409802675247,
      "reward_std": 0.12384946271777153,
      "reward_std_all": 0.12104190140962601,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8519929051399231,
      "rewards/forensics_tempsamp_iou_reward": 0.31797409802675247,
      "step": 1028
    },
    {
      "completion_length": 57.734375,
      "epoch": 2.737683089214381,
      "grad_norm": 6.445511817932129,
      "kl": 0.404296875,
      "learning_rate": 3.146666666666666e-07,
      "loss": 0.0162,
      "reward": 0.30687879025936127,
      "reward_std": 0.08286375179886818,
      "reward_std_all": 0.1060541644692421,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9811982214450836,
      "rewards/forensics_tempsamp_iou_reward": 0.30687879025936127,
      "step": 1029
    },
    {
      "completion_length": 48.9375,
      "epoch": 2.74034620505992,
      "grad_norm": 5.9301652908325195,
      "kl": 0.552734375,
      "learning_rate": 3.14e-07,
      "loss": 0.0221,
      "reward": 0.4000554829835892,
      "reward_std": 0.050710730254650116,
      "reward_std_all": 0.0518084354698658,
      "rewards-max/forensics_tempsamp_iou_reward": 0.989398866891861,
      "rewards/forensics_tempsamp_iou_reward": 0.4000554829835892,
      "step": 1030
    },
    {
      "completion_length": 33.671875,
      "epoch": 2.7430093209054593,
      "grad_norm": 11.409177780151367,
      "kl": 1.142578125,
      "learning_rate": 3.1333333333333333e-07,
      "loss": 0.0458,
      "reward": 0.2839377373456955,
      "reward_std": 0.09522148221731186,
      "reward_std_all": 0.08880370110273361,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9878485202789307,
      "rewards/forensics_tempsamp_iou_reward": 0.2839377373456955,
      "step": 1031
    },
    {
      "completion_length": 82.84375,
      "epoch": 2.745672436750999,
      "grad_norm": 8.850500106811523,
      "kl": 1.05859375,
      "learning_rate": 3.1266666666666663e-07,
      "loss": 0.0424,
      "reward": 0.27799302339553833,
      "reward_std": 0.10505277663469315,
      "reward_std_all": 0.09540077671408653,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9667994976043701,
      "rewards/forensics_tempsamp_iou_reward": 0.27799302339553833,
      "step": 1032
    },
    {
      "completion_length": 65.125,
      "epoch": 2.748335552596538,
      "grad_norm": 7.680317401885986,
      "kl": 0.701171875,
      "learning_rate": 3.12e-07,
      "loss": 0.028,
      "reward": 0.2799750193953514,
      "reward_std": 0.1188180036842823,
      "reward_std_all": 0.12345144152641296,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9847733676433563,
      "rewards/forensics_tempsamp_iou_reward": 0.2799750193953514,
      "step": 1033
    },
    {
      "completion_length": 53.109375,
      "epoch": 2.7509986684420773,
      "grad_norm": 8.186453819274902,
      "kl": 0.8984375,
      "learning_rate": 3.1133333333333334e-07,
      "loss": 0.036,
      "reward": 0.2984434440732002,
      "reward_std": 0.067399175837636,
      "reward_std_all": 0.08166131004691124,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9897143244743347,
      "rewards/forensics_tempsamp_iou_reward": 0.2984434440732002,
      "step": 1034
    },
    {
      "completion_length": 49.703125,
      "epoch": 2.7536617842876163,
      "grad_norm": 11.916816711425781,
      "kl": 0.55859375,
      "learning_rate": 3.1066666666666664e-07,
      "loss": 0.0223,
      "reward": 0.38422271609306335,
      "reward_std": 0.07975426688790321,
      "reward_std_all": 0.09260157495737076,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9733875691890717,
      "rewards/forensics_tempsamp_iou_reward": 0.38422271609306335,
      "step": 1035
    },
    {
      "completion_length": 78.28125,
      "epoch": 2.756324900133156,
      "grad_norm": 6.266043186187744,
      "kl": 0.54296875,
      "learning_rate": 3.1e-07,
      "loss": 0.0217,
      "reward": 0.29408174753189087,
      "reward_std": 0.05946051701903343,
      "reward_std_all": 0.06843561679124832,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9535790979862213,
      "rewards/forensics_tempsamp_iou_reward": 0.29408174753189087,
      "step": 1036
    },
    {
      "completion_length": 42.4375,
      "epoch": 2.7589880159786953,
      "grad_norm": 7.104606628417969,
      "kl": 0.822265625,
      "learning_rate": 3.0933333333333335e-07,
      "loss": 0.033,
      "reward": 0.2074204981327057,
      "reward_std": 0.08506253734230995,
      "reward_std_all": 0.09360184893012047,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7991460561752319,
      "rewards/forensics_tempsamp_iou_reward": 0.2074204981327057,
      "step": 1037
    },
    {
      "completion_length": 31.21875,
      "epoch": 2.7616511318242343,
      "grad_norm": 6.207160472869873,
      "kl": 0.44140625,
      "learning_rate": 3.0866666666666665e-07,
      "loss": 0.0177,
      "reward": 0.34364065527915955,
      "reward_std": 0.0815894640982151,
      "reward_std_all": 0.07672438770532608,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9870305359363556,
      "rewards/forensics_tempsamp_iou_reward": 0.34364065527915955,
      "step": 1038
    },
    {
      "completion_length": 60.0625,
      "epoch": 2.7643142476697737,
      "grad_norm": 8.145869255065918,
      "kl": 0.56640625,
      "learning_rate": 3.08e-07,
      "loss": 0.0226,
      "reward": 0.29857775568962097,
      "reward_std": 0.09227421134710312,
      "reward_std_all": 0.0870019719004631,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8959197402000427,
      "rewards/forensics_tempsamp_iou_reward": 0.29857775568962097,
      "step": 1039
    },
    {
      "completion_length": 37.828125,
      "epoch": 2.7669773635153128,
      "grad_norm": 5.0436787605285645,
      "kl": 0.5234375,
      "learning_rate": 3.0733333333333336e-07,
      "loss": 0.0209,
      "reward": 0.3245163708925247,
      "reward_std": 0.12719687074422836,
      "reward_std_all": 0.12125325202941895,
      "rewards-max/forensics_tempsamp_iou_reward": 0.981818825006485,
      "rewards/forensics_tempsamp_iou_reward": 0.3245163708925247,
      "step": 1040
    },
    {
      "completion_length": 37.75,
      "epoch": 2.7696404793608522,
      "grad_norm": 6.358025550842285,
      "kl": 0.4052734375,
      "learning_rate": 3.066666666666666e-07,
      "loss": 0.0162,
      "reward": 0.37672799825668335,
      "reward_std": 0.07289716973900795,
      "reward_std_all": 0.07378899864852428,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9929265081882477,
      "rewards/forensics_tempsamp_iou_reward": 0.37672799825668335,
      "step": 1041
    },
    {
      "completion_length": 40.15625,
      "epoch": 2.7723035952063917,
      "grad_norm": 6.643672466278076,
      "kl": 0.48046875,
      "learning_rate": 3.0599999999999996e-07,
      "loss": 0.0193,
      "reward": 0.39601215720176697,
      "reward_std": 0.08084907755255699,
      "reward_std_all": 0.0850462019443512,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9956677556037903,
      "rewards/forensics_tempsamp_iou_reward": 0.39601215720176697,
      "step": 1042
    },
    {
      "completion_length": 31.5,
      "epoch": 2.7749667110519307,
      "grad_norm": 8.109235763549805,
      "kl": 0.4619140625,
      "learning_rate": 3.053333333333333e-07,
      "loss": 0.0184,
      "reward": 0.2656524032354355,
      "reward_std": 0.04934192635118961,
      "reward_std_all": 0.04945080354809761,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9929812252521515,
      "rewards/forensics_tempsamp_iou_reward": 0.2656524032354355,
      "step": 1043
    },
    {
      "completion_length": 65.734375,
      "epoch": 2.77762982689747,
      "grad_norm": 9.20737361907959,
      "kl": 0.552734375,
      "learning_rate": 3.046666666666666e-07,
      "loss": 0.0221,
      "reward": 0.2574255019426346,
      "reward_std": 0.11716768145561218,
      "reward_std_all": 0.11705787107348442,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8417174220085144,
      "rewards/forensics_tempsamp_iou_reward": 0.2574255019426346,
      "step": 1044
    },
    {
      "completion_length": 59.015625,
      "epoch": 2.780292942743009,
      "grad_norm": 7.148258686065674,
      "kl": 0.5546875,
      "learning_rate": 3.0399999999999997e-07,
      "loss": 0.0222,
      "reward": 0.24115745723247528,
      "reward_std": 0.08581143617630005,
      "reward_std_all": 0.08602829277515411,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7902777791023254,
      "rewards/forensics_tempsamp_iou_reward": 0.24115745723247528,
      "step": 1045
    },
    {
      "completion_length": 45.625,
      "epoch": 2.7829560585885487,
      "grad_norm": 6.926023006439209,
      "kl": 0.546875,
      "learning_rate": 3.033333333333333e-07,
      "loss": 0.0219,
      "reward": 0.47816096246242523,
      "reward_std": 0.07160168513655663,
      "reward_std_all": 0.07703477889299393,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9954407215118408,
      "rewards/forensics_tempsamp_iou_reward": 0.47816096246242523,
      "step": 1046
    },
    {
      "completion_length": 52.890625,
      "epoch": 2.785619174434088,
      "grad_norm": 5.746521472930908,
      "kl": 0.5576171875,
      "learning_rate": 3.026666666666666e-07,
      "loss": 0.0222,
      "reward": 0.2944710701704025,
      "reward_std": 0.10348057001829147,
      "reward_std_all": 0.09830735251307487,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8830609917640686,
      "rewards/forensics_tempsamp_iou_reward": 0.2944710701704025,
      "step": 1047
    },
    {
      "completion_length": 47.734375,
      "epoch": 2.788282290279627,
      "grad_norm": 5.000545024871826,
      "kl": 0.45703125,
      "learning_rate": 3.02e-07,
      "loss": 0.0183,
      "reward": 0.34939926862716675,
      "reward_std": 0.06287837214767933,
      "reward_std_all": 0.06868061050772667,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9889040291309357,
      "rewards/forensics_tempsamp_iou_reward": 0.34939926862716675,
      "step": 1048
    },
    {
      "completion_length": 69.796875,
      "epoch": 2.790945406125166,
      "grad_norm": 5.510976791381836,
      "kl": 0.83984375,
      "learning_rate": 3.0133333333333333e-07,
      "loss": 0.0336,
      "reward": 0.19623346626758575,
      "reward_std": 0.06387020833790302,
      "reward_std_all": 0.06759023666381836,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5915214717388153,
      "rewards/forensics_tempsamp_iou_reward": 0.19623346626758575,
      "step": 1049
    },
    {
      "completion_length": 84.5,
      "epoch": 2.7936085219707056,
      "grad_norm": 7.2684831619262695,
      "kl": 0.654296875,
      "learning_rate": 3.006666666666667e-07,
      "loss": 0.0262,
      "reward": 0.31202878057956696,
      "reward_std": 0.09540961682796478,
      "reward_std_all": 0.12628819793462753,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9791666567325592,
      "rewards/forensics_tempsamp_iou_reward": 0.31202878057956696,
      "step": 1050
    },
    {
      "completion_length": 42.734375,
      "epoch": 2.796271637816245,
      "grad_norm": 7.58738374710083,
      "kl": 0.591796875,
      "learning_rate": 3e-07,
      "loss": 0.0236,
      "reward": 0.20602648705244064,
      "reward_std": 0.04553614743053913,
      "reward_std_all": 0.06059762090444565,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6914529949426651,
      "rewards/forensics_tempsamp_iou_reward": 0.20602648705244064,
      "step": 1051
    },
    {
      "completion_length": 64.296875,
      "epoch": 2.798934753661784,
      "grad_norm": 7.122767448425293,
      "kl": 0.591796875,
      "learning_rate": 2.9933333333333334e-07,
      "loss": 0.0236,
      "reward": 0.27085180580616,
      "reward_std": 0.09747452288866043,
      "reward_std_all": 0.0981258749961853,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8647669851779938,
      "rewards/forensics_tempsamp_iou_reward": 0.27085180580616,
      "step": 1052
    },
    {
      "completion_length": 32.09375,
      "epoch": 2.8015978695073236,
      "grad_norm": 6.604801654815674,
      "kl": 0.521484375,
      "learning_rate": 2.986666666666667e-07,
      "loss": 0.0209,
      "reward": 0.34493179619312286,
      "reward_std": 0.08206099085509777,
      "reward_std_all": 0.07748454995453358,
      "rewards-max/forensics_tempsamp_iou_reward": 0.979572206735611,
      "rewards/forensics_tempsamp_iou_reward": 0.34493179619312286,
      "step": 1053
    },
    {
      "completion_length": 61.390625,
      "epoch": 2.8042609853528626,
      "grad_norm": 7.858840465545654,
      "kl": 0.6875,
      "learning_rate": 2.98e-07,
      "loss": 0.0275,
      "reward": 0.32751229405403137,
      "reward_std": 0.1071629635989666,
      "reward_std_all": 0.09520638734102249,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9892640113830566,
      "rewards/forensics_tempsamp_iou_reward": 0.32751229405403137,
      "step": 1054
    },
    {
      "completion_length": 68.828125,
      "epoch": 2.806924101198402,
      "grad_norm": 6.591471195220947,
      "kl": 0.8671875,
      "learning_rate": 2.9733333333333335e-07,
      "loss": 0.0348,
      "reward": 0.431905135512352,
      "reward_std": 0.13395583629608154,
      "reward_std_all": 0.15156596153974533,
      "rewards-max/forensics_tempsamp_iou_reward": 0.997036874294281,
      "rewards/forensics_tempsamp_iou_reward": 0.431905135512352,
      "step": 1055
    },
    {
      "completion_length": 55.4375,
      "epoch": 2.8095872170439415,
      "grad_norm": 6.165871620178223,
      "kl": 0.49609375,
      "learning_rate": 2.966666666666667e-07,
      "loss": 0.0199,
      "reward": 0.34923718869686127,
      "reward_std": 0.06492171250283718,
      "reward_std_all": 0.07130233570933342,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9697336554527283,
      "rewards/forensics_tempsamp_iou_reward": 0.34923718869686127,
      "step": 1056
    },
    {
      "completion_length": 66.28125,
      "epoch": 2.8122503328894806,
      "grad_norm": 6.611512660980225,
      "kl": 0.4072265625,
      "learning_rate": 2.9599999999999995e-07,
      "loss": 0.0163,
      "reward": 0.25054532289505005,
      "reward_std": 0.05295851267874241,
      "reward_std_all": 0.08149079233407974,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9720084369182587,
      "rewards/forensics_tempsamp_iou_reward": 0.25054532289505005,
      "step": 1057
    },
    {
      "completion_length": 64.359375,
      "epoch": 2.81491344873502,
      "grad_norm": 6.3279500007629395,
      "kl": 0.85546875,
      "learning_rate": 2.953333333333333e-07,
      "loss": 0.0342,
      "reward": 0.32637983560562134,
      "reward_std": 0.05394993256777525,
      "reward_std_all": 0.05931493267416954,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9591662585735321,
      "rewards/forensics_tempsamp_iou_reward": 0.32637983560562134,
      "step": 1058
    },
    {
      "completion_length": 80.328125,
      "epoch": 2.817576564580559,
      "grad_norm": 7.068083763122559,
      "kl": 0.806640625,
      "learning_rate": 2.9466666666666666e-07,
      "loss": 0.0323,
      "reward": 0.17767918854951859,
      "reward_std": 0.0870941448956728,
      "reward_std_all": 0.0892864540219307,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5789937525987625,
      "rewards/forensics_tempsamp_iou_reward": 0.17767918854951859,
      "step": 1059
    },
    {
      "completion_length": 74.78125,
      "epoch": 2.8202396804260985,
      "grad_norm": 7.593339443206787,
      "kl": 0.45703125,
      "learning_rate": 2.9399999999999996e-07,
      "loss": 0.0183,
      "reward": 0.15526321530342102,
      "reward_std": 0.0629921481013298,
      "reward_std_all": 0.06539848074316978,
      "rewards-max/forensics_tempsamp_iou_reward": 0.36385948956012726,
      "rewards/forensics_tempsamp_iou_reward": 0.15526321530342102,
      "step": 1060
    },
    {
      "completion_length": 66.96875,
      "epoch": 2.822902796271638,
      "grad_norm": 6.212695598602295,
      "kl": 0.533203125,
      "learning_rate": 2.933333333333333e-07,
      "loss": 0.0213,
      "reward": 0.32368072867393494,
      "reward_std": 0.09956098347902298,
      "reward_std_all": 0.11589617282152176,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9835506975650787,
      "rewards/forensics_tempsamp_iou_reward": 0.32368072867393494,
      "step": 1061
    },
    {
      "completion_length": 51.59375,
      "epoch": 2.825565912117177,
      "grad_norm": 6.107614040374756,
      "kl": 0.328125,
      "learning_rate": 2.9266666666666667e-07,
      "loss": 0.0131,
      "reward": 0.41849811375141144,
      "reward_std": 0.07888336293399334,
      "reward_std_all": 0.07084718346595764,
      "rewards-max/forensics_tempsamp_iou_reward": 0.983106791973114,
      "rewards/forensics_tempsamp_iou_reward": 0.41849811375141144,
      "step": 1062
    },
    {
      "completion_length": 40.65625,
      "epoch": 2.8282290279627165,
      "grad_norm": 5.7705535888671875,
      "kl": 0.529296875,
      "learning_rate": 2.9199999999999997e-07,
      "loss": 0.0211,
      "reward": 0.46483975648880005,
      "reward_std": 0.07419557496905327,
      "reward_std_all": 0.0921827182173729,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9916666746139526,
      "rewards/forensics_tempsamp_iou_reward": 0.46483975648880005,
      "step": 1063
    },
    {
      "completion_length": 64.703125,
      "epoch": 2.8308921438082555,
      "grad_norm": 7.0026726722717285,
      "kl": 0.61328125,
      "learning_rate": 2.913333333333333e-07,
      "loss": 0.0245,
      "reward": 0.2067582681775093,
      "reward_std": 0.07043091580271721,
      "reward_std_all": 0.07658602111041546,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5829741880297661,
      "rewards/forensics_tempsamp_iou_reward": 0.2067582681775093,
      "step": 1064
    },
    {
      "completion_length": 41.96875,
      "epoch": 2.833555259653795,
      "grad_norm": 6.781606197357178,
      "kl": 0.556640625,
      "learning_rate": 2.906666666666667e-07,
      "loss": 0.0222,
      "reward": 0.32181333750486374,
      "reward_std": 0.09889182448387146,
      "reward_std_all": 0.0919688381254673,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9832765460014343,
      "rewards/forensics_tempsamp_iou_reward": 0.32181333750486374,
      "step": 1065
    },
    {
      "completion_length": 50.6875,
      "epoch": 2.8362183754993344,
      "grad_norm": 7.234423637390137,
      "kl": 0.5693359375,
      "learning_rate": 2.9e-07,
      "loss": 0.0228,
      "reward": 0.20033558458089828,
      "reward_std": 0.08693886920809746,
      "reward_std_all": 0.09259989112615585,
      "rewards-max/forensics_tempsamp_iou_reward": 0.71122807264328,
      "rewards/forensics_tempsamp_iou_reward": 0.20033558458089828,
      "step": 1066
    },
    {
      "completion_length": 48.03125,
      "epoch": 2.8388814913448734,
      "grad_norm": 5.429534912109375,
      "kl": 0.8642578125,
      "learning_rate": 2.8933333333333333e-07,
      "loss": 0.0345,
      "reward": 0.2510717362165451,
      "reward_std": 0.05964003503322601,
      "reward_std_all": 0.07039706036448479,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7318569123744965,
      "rewards/forensics_tempsamp_iou_reward": 0.2510717362165451,
      "step": 1067
    },
    {
      "completion_length": 82.953125,
      "epoch": 2.841544607190413,
      "grad_norm": 6.940547466278076,
      "kl": 0.4501953125,
      "learning_rate": 2.886666666666667e-07,
      "loss": 0.0181,
      "reward": 0.3074023723602295,
      "reward_std": 0.07501059770584106,
      "reward_std_all": 0.07996342331171036,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9898585379123688,
      "rewards/forensics_tempsamp_iou_reward": 0.3074023723602295,
      "step": 1068
    },
    {
      "completion_length": 72.765625,
      "epoch": 2.844207723035952,
      "grad_norm": 5.228595733642578,
      "kl": 0.6640625,
      "learning_rate": 2.88e-07,
      "loss": 0.0266,
      "reward": 0.3425074368715286,
      "reward_std": 0.10799907520413399,
      "reward_std_all": 0.13348764181137085,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9578212201595306,
      "rewards/forensics_tempsamp_iou_reward": 0.3425074368715286,
      "step": 1069
    },
    {
      "completion_length": 41.234375,
      "epoch": 2.8468708388814914,
      "grad_norm": 5.9166131019592285,
      "kl": 0.564453125,
      "learning_rate": 2.8733333333333334e-07,
      "loss": 0.0226,
      "reward": 0.3901197165250778,
      "reward_std": 0.08775247633457184,
      "reward_std_all": 0.08870378509163857,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9976079165935516,
      "rewards/forensics_tempsamp_iou_reward": 0.3901197165250778,
      "step": 1070
    },
    {
      "completion_length": 63.875,
      "epoch": 2.849533954727031,
      "grad_norm": 7.574952125549316,
      "kl": 0.591796875,
      "learning_rate": 2.866666666666667e-07,
      "loss": 0.0236,
      "reward": 0.27252211421728134,
      "reward_std": 0.06086721830070019,
      "reward_std_all": 0.08413565158843994,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8205292820930481,
      "rewards/forensics_tempsamp_iou_reward": 0.27252211421728134,
      "step": 1071
    },
    {
      "completion_length": 50.546875,
      "epoch": 2.85219707057257,
      "grad_norm": 6.093450546264648,
      "kl": 0.5791015625,
      "learning_rate": 2.8599999999999994e-07,
      "loss": 0.0232,
      "reward": 0.3543921709060669,
      "reward_std": 0.08152801357209682,
      "reward_std_all": 0.07989959046244621,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9548346698284149,
      "rewards/forensics_tempsamp_iou_reward": 0.3543921709060669,
      "step": 1072
    },
    {
      "completion_length": 71.96875,
      "epoch": 2.8548601864181093,
      "grad_norm": 6.409403324127197,
      "kl": 0.68359375,
      "learning_rate": 2.853333333333333e-07,
      "loss": 0.0273,
      "reward": 0.34946149587631226,
      "reward_std": 0.10193114541471004,
      "reward_std_all": 0.0973423570394516,
      "rewards-max/forensics_tempsamp_iou_reward": 0.991840124130249,
      "rewards/forensics_tempsamp_iou_reward": 0.34946149587631226,
      "step": 1073
    },
    {
      "completion_length": 63.609375,
      "epoch": 2.8575233022636484,
      "grad_norm": 7.107748031616211,
      "kl": 0.619140625,
      "learning_rate": 2.8466666666666665e-07,
      "loss": 0.0248,
      "reward": 0.27141544222831726,
      "reward_std": 0.09516017884016037,
      "reward_std_all": 0.09712813794612885,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9701455533504486,
      "rewards/forensics_tempsamp_iou_reward": 0.27141544222831726,
      "step": 1074
    },
    {
      "completion_length": 56.390625,
      "epoch": 2.860186418109188,
      "grad_norm": 6.5939555168151855,
      "kl": 0.4716796875,
      "learning_rate": 2.8399999999999995e-07,
      "loss": 0.0189,
      "reward": 0.44942888617515564,
      "reward_std": 0.10512315854430199,
      "reward_std_all": 0.1401805356144905,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9955173432826996,
      "rewards/forensics_tempsamp_iou_reward": 0.44942888617515564,
      "step": 1075
    },
    {
      "completion_length": 23.234375,
      "epoch": 2.8628495339547273,
      "grad_norm": 6.579409599304199,
      "kl": 0.4560546875,
      "learning_rate": 2.833333333333333e-07,
      "loss": 0.0182,
      "reward": 0.31655077636241913,
      "reward_std": 0.06238500401377678,
      "reward_std_all": 0.055228917859494686,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9961690008640289,
      "rewards/forensics_tempsamp_iou_reward": 0.31655077636241913,
      "step": 1076
    },
    {
      "completion_length": 76.03125,
      "epoch": 2.8655126498002663,
      "grad_norm": 7.724194049835205,
      "kl": 0.4501953125,
      "learning_rate": 2.8266666666666666e-07,
      "loss": 0.018,
      "reward": 0.2795167937874794,
      "reward_std": 0.08049260452389717,
      "reward_std_all": 0.08471247553825378,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9015122056007385,
      "rewards/forensics_tempsamp_iou_reward": 0.2795167937874794,
      "step": 1077
    },
    {
      "completion_length": 34.78125,
      "epoch": 2.8681757656458053,
      "grad_norm": 5.259130954742432,
      "kl": 0.4736328125,
      "learning_rate": 2.8199999999999996e-07,
      "loss": 0.019,
      "reward": 0.2552929222583771,
      "reward_std": 0.0731465108692646,
      "reward_std_all": 0.0643383301794529,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7257734090089798,
      "rewards/forensics_tempsamp_iou_reward": 0.2552929222583771,
      "step": 1078
    },
    {
      "completion_length": 51.84375,
      "epoch": 2.870838881491345,
      "grad_norm": 7.191152572631836,
      "kl": 0.783203125,
      "learning_rate": 2.813333333333333e-07,
      "loss": 0.0313,
      "reward": 0.4104444682598114,
      "reward_std": 0.07829748094081879,
      "reward_std_all": 0.06794390268623829,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9963056147098541,
      "rewards/forensics_tempsamp_iou_reward": 0.4104444682598114,
      "step": 1079
    },
    {
      "completion_length": 67.5,
      "epoch": 2.8735019973368843,
      "grad_norm": 6.498458385467529,
      "kl": 0.5322265625,
      "learning_rate": 2.8066666666666667e-07,
      "loss": 0.0213,
      "reward": 0.22194057703018188,
      "reward_std": 0.05351768061518669,
      "reward_std_all": 0.05188184976577759,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9761915504932404,
      "rewards/forensics_tempsamp_iou_reward": 0.22194057703018188,
      "step": 1080
    },
    {
      "completion_length": 36.484375,
      "epoch": 2.8761651131824233,
      "grad_norm": 7.565027713775635,
      "kl": 0.408203125,
      "learning_rate": 2.8e-07,
      "loss": 0.0163,
      "reward": 0.2552115470170975,
      "reward_std": 0.052421145141124725,
      "reward_std_all": 0.05566263943910599,
      "rewards-max/forensics_tempsamp_iou_reward": 0.938584953546524,
      "rewards/forensics_tempsamp_iou_reward": 0.2552115470170975,
      "step": 1081
    },
    {
      "completion_length": 69.125,
      "epoch": 2.8788282290279628,
      "grad_norm": 6.213925361633301,
      "kl": 0.591796875,
      "learning_rate": 2.793333333333333e-07,
      "loss": 0.0236,
      "reward": 0.1701706498861313,
      "reward_std": 0.07075244188308716,
      "reward_std_all": 0.07637863606214523,
      "rewards-max/forensics_tempsamp_iou_reward": 0.48695652186870575,
      "rewards/forensics_tempsamp_iou_reward": 0.1701706498861313,
      "step": 1082
    },
    {
      "completion_length": 68.578125,
      "epoch": 2.881491344873502,
      "grad_norm": 7.246175289154053,
      "kl": 0.859375,
      "learning_rate": 2.786666666666667e-07,
      "loss": 0.0343,
      "reward": 0.26775333285331726,
      "reward_std": 0.11596263200044632,
      "reward_std_all": 0.11035709083080292,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8926900327205658,
      "rewards/forensics_tempsamp_iou_reward": 0.26775333285331726,
      "step": 1083
    },
    {
      "completion_length": 51.15625,
      "epoch": 2.8841544607190412,
      "grad_norm": 8.14823055267334,
      "kl": 0.787109375,
      "learning_rate": 2.7800000000000003e-07,
      "loss": 0.0315,
      "reward": 0.27589239925146103,
      "reward_std": 0.07082692719995975,
      "reward_std_all": 0.07635113038122654,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9168227016925812,
      "rewards/forensics_tempsamp_iou_reward": 0.27589239925146103,
      "step": 1084
    },
    {
      "completion_length": 63.625,
      "epoch": 2.8868175765645807,
      "grad_norm": 6.447929382324219,
      "kl": 0.466796875,
      "learning_rate": 2.7733333333333333e-07,
      "loss": 0.0187,
      "reward": 0.20276660472154617,
      "reward_std": 0.10118043422698975,
      "reward_std_all": 0.09390301629900932,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7935605943202972,
      "rewards/forensics_tempsamp_iou_reward": 0.20276660472154617,
      "step": 1085
    },
    {
      "completion_length": 54.578125,
      "epoch": 2.8894806924101197,
      "grad_norm": 5.933666229248047,
      "kl": 0.462890625,
      "learning_rate": 2.766666666666667e-07,
      "loss": 0.0185,
      "reward": 0.4323224425315857,
      "reward_std": 0.09871237352490425,
      "reward_std_all": 0.12760257720947266,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9836647808551788,
      "rewards/forensics_tempsamp_iou_reward": 0.4323224425315857,
      "step": 1086
    },
    {
      "completion_length": 82.953125,
      "epoch": 2.892143808255659,
      "grad_norm": 5.898902893066406,
      "kl": 0.654296875,
      "learning_rate": 2.7600000000000004e-07,
      "loss": 0.0262,
      "reward": 0.3616512343287468,
      "reward_std": 0.18296605348587036,
      "reward_std_all": 0.1772991567850113,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9682460725307465,
      "rewards/forensics_tempsamp_iou_reward": 0.3616512343287468,
      "step": 1087
    },
    {
      "completion_length": 105.15625,
      "epoch": 2.894806924101198,
      "grad_norm": 4.173271656036377,
      "kl": 0.525390625,
      "learning_rate": 2.753333333333333e-07,
      "loss": 0.0211,
      "reward": 0.27802425622940063,
      "reward_std": 0.11846484243869781,
      "reward_std_all": 0.11352192610502243,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8396405875682831,
      "rewards/forensics_tempsamp_iou_reward": 0.27802425622940063,
      "step": 1088
    },
    {
      "completion_length": 80.015625,
      "epoch": 2.8974700399467377,
      "grad_norm": 5.990029811859131,
      "kl": 0.560546875,
      "learning_rate": 2.7466666666666664e-07,
      "loss": 0.0224,
      "reward": 0.2007494494318962,
      "reward_std": 0.07624949514865875,
      "reward_std_all": 0.06903290003538132,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8414090871810913,
      "rewards/forensics_tempsamp_iou_reward": 0.2007494494318962,
      "step": 1089
    },
    {
      "completion_length": 117.453125,
      "epoch": 2.900133155792277,
      "grad_norm": 7.574563026428223,
      "kl": 0.564453125,
      "learning_rate": 2.74e-07,
      "loss": 0.0226,
      "reward": 0.25600565969944,
      "reward_std": 0.10807032883167267,
      "reward_std_all": 0.13879601657390594,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9673978984355927,
      "rewards/forensics_tempsamp_iou_reward": 0.25600565969944,
      "step": 1090
    },
    {
      "completion_length": 55.578125,
      "epoch": 2.902796271637816,
      "grad_norm": 8.107633590698242,
      "kl": 0.7109375,
      "learning_rate": 2.733333333333333e-07,
      "loss": 0.0284,
      "reward": 0.38336512446403503,
      "reward_std": 0.15804129093885422,
      "reward_std_all": 0.16840074956417084,
      "rewards-max/forensics_tempsamp_iou_reward": 1.0,
      "rewards/forensics_tempsamp_iou_reward": 0.38336512446403503,
      "step": 1091
    },
    {
      "completion_length": 105.9375,
      "epoch": 2.9054593874833556,
      "grad_norm": 4.546135902404785,
      "kl": 0.64453125,
      "learning_rate": 2.7266666666666665e-07,
      "loss": 0.0257,
      "reward": 0.23524556308984756,
      "reward_std": 0.15574952214956284,
      "reward_std_all": 0.150973878800869,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7641225159168243,
      "rewards/forensics_tempsamp_iou_reward": 0.23524556308984756,
      "step": 1092
    },
    {
      "completion_length": 55.90625,
      "epoch": 2.9081225033288947,
      "grad_norm": 4.310430526733398,
      "kl": 0.5244140625,
      "learning_rate": 2.72e-07,
      "loss": 0.021,
      "reward": 0.3088548630475998,
      "reward_std": 0.08578852377831936,
      "reward_std_all": 0.08776062913239002,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8501190543174744,
      "rewards/forensics_tempsamp_iou_reward": 0.3088548630475998,
      "step": 1093
    },
    {
      "completion_length": 51.015625,
      "epoch": 2.910785619174434,
      "grad_norm": 8.961848258972168,
      "kl": 0.60546875,
      "learning_rate": 2.713333333333333e-07,
      "loss": 0.0242,
      "reward": 0.2780851125717163,
      "reward_std": 0.08833230659365654,
      "reward_std_all": 0.081366878002882,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8958109319210052,
      "rewards/forensics_tempsamp_iou_reward": 0.2780851125717163,
      "step": 1094
    },
    {
      "completion_length": 88.203125,
      "epoch": 2.9134487350199736,
      "grad_norm": 5.026880264282227,
      "kl": 0.48828125,
      "learning_rate": 2.7066666666666666e-07,
      "loss": 0.0196,
      "reward": 0.3083817660808563,
      "reward_std": 0.1600293591618538,
      "reward_std_all": 0.14649218320846558,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9213664531707764,
      "rewards/forensics_tempsamp_iou_reward": 0.3083817660808563,
      "step": 1095
    },
    {
      "completion_length": 63.4375,
      "epoch": 2.9161118508655126,
      "grad_norm": 6.48787784576416,
      "kl": 0.6640625,
      "learning_rate": 2.7e-07,
      "loss": 0.0265,
      "reward": 0.28197603672742844,
      "reward_std": 0.0882941335439682,
      "reward_std_all": 0.08279025927186012,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9982042014598846,
      "rewards/forensics_tempsamp_iou_reward": 0.28197603672742844,
      "step": 1096
    },
    {
      "completion_length": 111.625,
      "epoch": 2.918774966711052,
      "grad_norm": 5.43023157119751,
      "kl": 0.669921875,
      "learning_rate": 2.693333333333333e-07,
      "loss": 0.0268,
      "reward": 0.28751419484615326,
      "reward_std": 0.09756190329790115,
      "reward_std_all": 0.10220430046319962,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7019511759281158,
      "rewards/forensics_tempsamp_iou_reward": 0.28751419484615326,
      "step": 1097
    },
    {
      "completion_length": 58.796875,
      "epoch": 2.921438082556591,
      "grad_norm": 6.167535305023193,
      "kl": 0.6513671875,
      "learning_rate": 2.6866666666666666e-07,
      "loss": 0.0261,
      "reward": 0.25178124010562897,
      "reward_std": 0.1104726791381836,
      "reward_std_all": 0.10586981475353241,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9560036063194275,
      "rewards/forensics_tempsamp_iou_reward": 0.25178124010562897,
      "step": 1098
    },
    {
      "completion_length": 78.25,
      "epoch": 2.9241011984021306,
      "grad_norm": 5.8157429695129395,
      "kl": 0.603515625,
      "learning_rate": 2.68e-07,
      "loss": 0.0241,
      "reward": 0.23415987938642502,
      "reward_std": 0.10599619522690773,
      "reward_std_all": 0.10039955377578735,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9256756901741028,
      "rewards/forensics_tempsamp_iou_reward": 0.23415987938642502,
      "step": 1099
    },
    {
      "completion_length": 91.828125,
      "epoch": 2.92676431424767,
      "grad_norm": 6.372433185577393,
      "kl": 0.60546875,
      "learning_rate": 2.673333333333333e-07,
      "loss": 0.0242,
      "reward": 0.33545786142349243,
      "reward_std": 0.14917175471782684,
      "reward_std_all": 0.14301611855626106,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9718064665794373,
      "rewards/forensics_tempsamp_iou_reward": 0.33545786142349243,
      "step": 1100
    },
    {
      "completion_length": 98.9375,
      "epoch": 2.929427430093209,
      "grad_norm": 6.131017208099365,
      "kl": 0.5322265625,
      "learning_rate": 2.6666666666666667e-07,
      "loss": 0.0213,
      "reward": 0.22088836133480072,
      "reward_std": 0.09781123325228691,
      "reward_std_all": 0.10069501027464867,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7182731628417969,
      "rewards/forensics_tempsamp_iou_reward": 0.22088836133480072,
      "step": 1101
    },
    {
      "completion_length": 67.90625,
      "epoch": 2.932090545938748,
      "grad_norm": 6.030054569244385,
      "kl": 0.482421875,
      "learning_rate": 2.66e-07,
      "loss": 0.0193,
      "reward": 0.3689675033092499,
      "reward_std": 0.13600099086761475,
      "reward_std_all": 0.12773935124278069,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9824561476707458,
      "rewards/forensics_tempsamp_iou_reward": 0.3689675033092499,
      "step": 1102
    },
    {
      "completion_length": 103.984375,
      "epoch": 2.9347536617842875,
      "grad_norm": 7.086054801940918,
      "kl": 0.4599609375,
      "learning_rate": 2.653333333333333e-07,
      "loss": 0.0184,
      "reward": 0.28455789387226105,
      "reward_std": 0.089919563382864,
      "reward_std_all": 0.08150500804185867,
      "rewards-max/forensics_tempsamp_iou_reward": 0.977125734090805,
      "rewards/forensics_tempsamp_iou_reward": 0.28455789387226105,
      "step": 1103
    },
    {
      "completion_length": 111.9375,
      "epoch": 2.937416777629827,
      "grad_norm": 5.159308910369873,
      "kl": 0.5556640625,
      "learning_rate": 2.6466666666666663e-07,
      "loss": 0.0223,
      "reward": 0.21784979850053787,
      "reward_std": 0.1563883200287819,
      "reward_std_all": 0.1420081965625286,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7038461416959763,
      "rewards/forensics_tempsamp_iou_reward": 0.21784979850053787,
      "step": 1104
    },
    {
      "completion_length": 29.796875,
      "epoch": 2.940079893475366,
      "grad_norm": 8.570893287658691,
      "kl": 0.408203125,
      "learning_rate": 2.64e-07,
      "loss": 0.0163,
      "reward": 0.3358606547117233,
      "reward_std": 0.07639472559094429,
      "reward_std_all": 0.0750860795378685,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9946711957454681,
      "rewards/forensics_tempsamp_iou_reward": 0.3358606547117233,
      "step": 1105
    },
    {
      "completion_length": 41.90625,
      "epoch": 2.9427430093209055,
      "grad_norm": 7.045068264007568,
      "kl": 0.4072265625,
      "learning_rate": 2.633333333333333e-07,
      "loss": 0.0163,
      "reward": 0.178690105676651,
      "reward_std": 0.0342480530962348,
      "reward_std_all": 0.05033205822110176,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8308333456516266,
      "rewards/forensics_tempsamp_iou_reward": 0.178690105676651,
      "step": 1106
    },
    {
      "completion_length": 98.953125,
      "epoch": 2.9454061251664445,
      "grad_norm": 6.458528518676758,
      "kl": 0.7734375,
      "learning_rate": 2.6266666666666664e-07,
      "loss": 0.0309,
      "reward": 0.3080301880836487,
      "reward_std": 0.16781644523143768,
      "reward_std_all": 0.18303174525499344,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9518840909004211,
      "rewards/forensics_tempsamp_iou_reward": 0.3080301880836487,
      "step": 1107
    },
    {
      "completion_length": 38.53125,
      "epoch": 2.948069241011984,
      "grad_norm": 7.488367080688477,
      "kl": 0.4716796875,
      "learning_rate": 2.62e-07,
      "loss": 0.0189,
      "reward": 0.3793780505657196,
      "reward_std": 0.1420685425400734,
      "reward_std_all": 0.14821895956993103,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9832430481910706,
      "rewards/forensics_tempsamp_iou_reward": 0.3793780505657196,
      "step": 1108
    },
    {
      "completion_length": 84.328125,
      "epoch": 2.9507323568575234,
      "grad_norm": 7.6134867668151855,
      "kl": 0.595703125,
      "learning_rate": 2.613333333333333e-07,
      "loss": 0.0238,
      "reward": 0.16000109910964966,
      "reward_std": 0.10621730796992779,
      "reward_std_all": 0.10007059201598167,
      "rewards-max/forensics_tempsamp_iou_reward": 0.554246336221695,
      "rewards/forensics_tempsamp_iou_reward": 0.16000109910964966,
      "step": 1109
    },
    {
      "completion_length": 65.65625,
      "epoch": 2.9533954727030625,
      "grad_norm": 4.819548606872559,
      "kl": 0.5546875,
      "learning_rate": 2.6066666666666664e-07,
      "loss": 0.0221,
      "reward": 0.36661654710769653,
      "reward_std": 0.09945392608642578,
      "reward_std_all": 0.11913066729903221,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9860495030879974,
      "rewards/forensics_tempsamp_iou_reward": 0.36661654710769653,
      "step": 1110
    },
    {
      "completion_length": 58.640625,
      "epoch": 2.956058588548602,
      "grad_norm": 8.12216854095459,
      "kl": 0.6103515625,
      "learning_rate": 2.6e-07,
      "loss": 0.0244,
      "reward": 0.3974362313747406,
      "reward_std": 0.15541329234838486,
      "reward_std_all": 0.1521133929491043,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9730353653430939,
      "rewards/forensics_tempsamp_iou_reward": 0.3974362313747406,
      "step": 1111
    },
    {
      "completion_length": 70.640625,
      "epoch": 2.958721704394141,
      "grad_norm": 5.769015312194824,
      "kl": 0.4208984375,
      "learning_rate": 2.5933333333333335e-07,
      "loss": 0.0168,
      "reward": 0.38845549523830414,
      "reward_std": 0.10509976372122765,
      "reward_std_all": 0.10685261338949203,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9210993051528931,
      "rewards/forensics_tempsamp_iou_reward": 0.38845549523830414,
      "step": 1112
    },
    {
      "completion_length": 57.484375,
      "epoch": 2.9613848202396804,
      "grad_norm": 8.437132835388184,
      "kl": 0.4921875,
      "learning_rate": 2.5866666666666665e-07,
      "loss": 0.0197,
      "reward": 0.3488670587539673,
      "reward_std": 0.072559779509902,
      "reward_std_all": 0.07321479730308056,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9859230816364288,
      "rewards/forensics_tempsamp_iou_reward": 0.3488670587539673,
      "step": 1113
    },
    {
      "completion_length": 69.0625,
      "epoch": 2.96404793608522,
      "grad_norm": 6.308451175689697,
      "kl": 0.537109375,
      "learning_rate": 2.58e-07,
      "loss": 0.0215,
      "reward": 0.3509721979498863,
      "reward_std": 0.10285273939371109,
      "reward_std_all": 0.1277218759059906,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8254793584346771,
      "rewards/forensics_tempsamp_iou_reward": 0.3509721979498863,
      "step": 1114
    },
    {
      "completion_length": 67.21875,
      "epoch": 2.966711051930759,
      "grad_norm": 5.571609973907471,
      "kl": 0.5400390625,
      "learning_rate": 2.5733333333333336e-07,
      "loss": 0.0216,
      "reward": 0.24551808834075928,
      "reward_std": 0.07843727059662342,
      "reward_std_all": 0.12027677148580551,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9760149121284485,
      "rewards/forensics_tempsamp_iou_reward": 0.24551808834075928,
      "step": 1115
    },
    {
      "completion_length": 84.796875,
      "epoch": 2.9693741677762984,
      "grad_norm": 5.623020172119141,
      "kl": 0.697265625,
      "learning_rate": 2.5666666666666666e-07,
      "loss": 0.0279,
      "reward": 0.25999727845191956,
      "reward_std": 0.09573816880583763,
      "reward_std_all": 0.11809901520609856,
      "rewards-max/forensics_tempsamp_iou_reward": 0.952777773141861,
      "rewards/forensics_tempsamp_iou_reward": 0.25999727845191956,
      "step": 1116
    },
    {
      "completion_length": 152.546875,
      "epoch": 2.9720372836218374,
      "grad_norm": 5.636292934417725,
      "kl": 0.88671875,
      "learning_rate": 2.56e-07,
      "loss": 0.0355,
      "reward": 0.1954292356967926,
      "reward_std": 0.13677320256829262,
      "reward_std_all": 0.14103256911039352,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6832781881093979,
      "rewards/forensics_tempsamp_iou_reward": 0.1954292356967926,
      "step": 1117
    },
    {
      "completion_length": 74.59375,
      "epoch": 2.974700399467377,
      "grad_norm": 4.808640003204346,
      "kl": 0.5673828125,
      "learning_rate": 2.5533333333333337e-07,
      "loss": 0.0227,
      "reward": 0.30434735119342804,
      "reward_std": 0.09215487539768219,
      "reward_std_all": 0.11033802852034569,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9676669538021088,
      "rewards/forensics_tempsamp_iou_reward": 0.30434735119342804,
      "step": 1118
    },
    {
      "completion_length": 71.140625,
      "epoch": 2.9773635153129163,
      "grad_norm": 5.970154762268066,
      "kl": 0.6748046875,
      "learning_rate": 2.546666666666666e-07,
      "loss": 0.027,
      "reward": 0.39468422532081604,
      "reward_std": 0.1612311452627182,
      "reward_std_all": 0.14805227518081665,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9709001779556274,
      "rewards/forensics_tempsamp_iou_reward": 0.39468422532081604,
      "step": 1119
    },
    {
      "completion_length": 55.421875,
      "epoch": 2.9800266311584553,
      "grad_norm": 5.981649398803711,
      "kl": 0.443359375,
      "learning_rate": 2.5399999999999997e-07,
      "loss": 0.0178,
      "reward": 0.18797162920236588,
      "reward_std": 0.09383733570575714,
      "reward_std_all": 0.0890425480902195,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6713855266571045,
      "rewards/forensics_tempsamp_iou_reward": 0.18797162920236588,
      "step": 1120
    },
    {
      "completion_length": 62.4375,
      "epoch": 2.982689747003995,
      "grad_norm": 6.593993186950684,
      "kl": 0.53125,
      "learning_rate": 2.533333333333333e-07,
      "loss": 0.0213,
      "reward": 0.3331603705883026,
      "reward_std": 0.11117560416460037,
      "reward_std_all": 0.1079804003238678,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9512899518013,
      "rewards/forensics_tempsamp_iou_reward": 0.3331603705883026,
      "step": 1121
    },
    {
      "completion_length": 77.859375,
      "epoch": 2.985352862849534,
      "grad_norm": 6.195252895355225,
      "kl": 0.4814453125,
      "learning_rate": 2.526666666666666e-07,
      "loss": 0.0192,
      "reward": 0.21819883584976196,
      "reward_std": 0.06869159266352654,
      "reward_std_all": 0.08056220598518848,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7990041971206665,
      "rewards/forensics_tempsamp_iou_reward": 0.21819883584976196,
      "step": 1122
    },
    {
      "completion_length": 76.484375,
      "epoch": 2.9880159786950733,
      "grad_norm": 9.066922187805176,
      "kl": 0.548828125,
      "learning_rate": 2.52e-07,
      "loss": 0.0219,
      "reward": 0.3117143511772156,
      "reward_std": 0.11662581190466881,
      "reward_std_all": 0.11490239202976227,
      "rewards-max/forensics_tempsamp_iou_reward": 0.973563939332962,
      "rewards/forensics_tempsamp_iou_reward": 0.3117143511772156,
      "step": 1123
    },
    {
      "completion_length": 82.890625,
      "epoch": 2.9906790945406128,
      "grad_norm": 6.154927730560303,
      "kl": 0.576171875,
      "learning_rate": 2.5133333333333333e-07,
      "loss": 0.0231,
      "reward": 0.2955896705389023,
      "reward_std": 0.08799413964152336,
      "reward_std_all": 0.09679621830582619,
      "rewards-max/forensics_tempsamp_iou_reward": 0.982142835855484,
      "rewards/forensics_tempsamp_iou_reward": 0.2955896705389023,
      "step": 1124
    },
    {
      "completion_length": 99.0625,
      "epoch": 2.993342210386152,
      "grad_norm": 7.54512882232666,
      "kl": 0.7890625,
      "learning_rate": 2.5066666666666663e-07,
      "loss": 0.0315,
      "reward": 0.38480694591999054,
      "reward_std": 0.173265740275383,
      "reward_std_all": 0.17629363387823105,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9740006923675537,
      "rewards/forensics_tempsamp_iou_reward": 0.38480694591999054,
      "step": 1125
    },
    {
      "completion_length": 68.078125,
      "epoch": 2.996005326231691,
      "grad_norm": 5.620226860046387,
      "kl": 0.53515625,
      "learning_rate": 2.5e-07,
      "loss": 0.0214,
      "reward": 0.2852519825100899,
      "reward_std": 0.04055996425449848,
      "reward_std_all": 0.04717816971242428,
      "rewards-max/forensics_tempsamp_iou_reward": 0.840429961681366,
      "rewards/forensics_tempsamp_iou_reward": 0.2852519825100899,
      "step": 1126
    },
    {
      "completion_length": 70.65625,
      "epoch": 2.9986684420772303,
      "grad_norm": 6.742706298828125,
      "kl": 0.650390625,
      "learning_rate": 2.493333333333333e-07,
      "loss": 0.0259,
      "reward": 0.47471582889556885,
      "reward_std": 0.153200201690197,
      "reward_std_all": 0.16544049233198166,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9974874258041382,
      "rewards/forensics_tempsamp_iou_reward": 0.47471582889556885,
      "step": 1127
    },
    {
      "completion_length": 14.0,
      "epoch": 3.0,
      "grad_norm": 4.0295634269714355,
      "kl": 0.498046875,
      "learning_rate": 2.4866666666666664e-07,
      "loss": 0.0066,
      "reward": 0.0,
      "reward_std": 0.03313463181257248,
      "reward_std_all": 0.0449785552918911,
      "rewards-max/forensics_tempsamp_iou_reward": 0.0,
      "rewards/forensics_tempsamp_iou_reward": 0.0,
      "step": 1128
    },
    {
      "completion_length": 66.78125,
      "epoch": 3.0026631158455395,
      "grad_norm": 6.085719585418701,
      "kl": 0.734375,
      "learning_rate": 2.48e-07,
      "loss": 0.0294,
      "reward": 0.20840881764888763,
      "reward_std": 0.1219136044383049,
      "reward_std_all": 0.11996643245220184,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8253302872180939,
      "rewards/forensics_tempsamp_iou_reward": 0.20840881764888763,
      "step": 1129
    },
    {
      "completion_length": 107.8125,
      "epoch": 3.0053262316910785,
      "grad_norm": 6.139739513397217,
      "kl": 0.65625,
      "learning_rate": 2.473333333333333e-07,
      "loss": 0.0263,
      "reward": 0.30418022722005844,
      "reward_std": 0.08715824037790298,
      "reward_std_all": 0.1198689229786396,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9818309247493744,
      "rewards/forensics_tempsamp_iou_reward": 0.30418022722005844,
      "step": 1130
    },
    {
      "completion_length": 79.328125,
      "epoch": 3.007989347536618,
      "grad_norm": 6.309010982513428,
      "kl": 0.623046875,
      "learning_rate": 2.4666666666666665e-07,
      "loss": 0.0249,
      "reward": 0.20992550253868103,
      "reward_std": 0.09333476424217224,
      "reward_std_all": 0.09234318882226944,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8576227426528931,
      "rewards/forensics_tempsamp_iou_reward": 0.20992550253868103,
      "step": 1131
    },
    {
      "completion_length": 61.359375,
      "epoch": 3.010652463382157,
      "grad_norm": 6.054462432861328,
      "kl": 0.3798828125,
      "learning_rate": 2.46e-07,
      "loss": 0.0152,
      "reward": 0.40278926491737366,
      "reward_std": 0.04940206930041313,
      "reward_std_all": 0.06972567364573479,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9949922263622284,
      "rewards/forensics_tempsamp_iou_reward": 0.40278926491737366,
      "step": 1132
    },
    {
      "completion_length": 66.109375,
      "epoch": 3.0133155792276964,
      "grad_norm": 5.997733116149902,
      "kl": 0.5771484375,
      "learning_rate": 2.453333333333333e-07,
      "loss": 0.0231,
      "reward": 0.3334674835205078,
      "reward_std": 0.10633722692728043,
      "reward_std_all": 0.10683688893914223,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9166783392429352,
      "rewards/forensics_tempsamp_iou_reward": 0.3334674835205078,
      "step": 1133
    },
    {
      "completion_length": 98.3125,
      "epoch": 3.015978695073236,
      "grad_norm": 5.385681629180908,
      "kl": 0.625,
      "learning_rate": 2.4466666666666666e-07,
      "loss": 0.025,
      "reward": 0.39091770350933075,
      "reward_std": 0.15492480248212814,
      "reward_std_all": 0.1524108722805977,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9946543872356415,
      "rewards/forensics_tempsamp_iou_reward": 0.39091770350933075,
      "step": 1134
    },
    {
      "completion_length": 91.921875,
      "epoch": 3.018641810918775,
      "grad_norm": 4.738413333892822,
      "kl": 0.626953125,
      "learning_rate": 2.4399999999999996e-07,
      "loss": 0.025,
      "reward": 0.2446717768907547,
      "reward_std": 0.1061597615480423,
      "reward_std_all": 0.11516335234045982,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8247601389884949,
      "rewards/forensics_tempsamp_iou_reward": 0.2446717768907547,
      "step": 1135
    },
    {
      "completion_length": 86.25,
      "epoch": 3.0213049267643144,
      "grad_norm": 6.349506855010986,
      "kl": 0.560546875,
      "learning_rate": 2.433333333333333e-07,
      "loss": 0.0224,
      "reward": 0.2841155230998993,
      "reward_std": 0.0884226094931364,
      "reward_std_all": 0.09056948870420456,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9805594086647034,
      "rewards/forensics_tempsamp_iou_reward": 0.2841155230998993,
      "step": 1136
    },
    {
      "completion_length": 119.765625,
      "epoch": 3.0239680426098534,
      "grad_norm": 6.687239646911621,
      "kl": 0.849609375,
      "learning_rate": 2.4266666666666667e-07,
      "loss": 0.034,
      "reward": 0.34877246618270874,
      "reward_std": 0.1401616558432579,
      "reward_std_all": 0.15612883865833282,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9457203447818756,
      "rewards/forensics_tempsamp_iou_reward": 0.34877246618270874,
      "step": 1137
    },
    {
      "completion_length": 69.875,
      "epoch": 3.026631158455393,
      "grad_norm": 6.3103227615356445,
      "kl": 0.8828125,
      "learning_rate": 2.4199999999999997e-07,
      "loss": 0.0352,
      "reward": 0.2373705804347992,
      "reward_std": 0.12257687002420425,
      "reward_std_all": 0.12888099253177643,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6745900213718414,
      "rewards/forensics_tempsamp_iou_reward": 0.2373705804347992,
      "step": 1138
    },
    {
      "completion_length": 46.578125,
      "epoch": 3.029294274300932,
      "grad_norm": 6.066370964050293,
      "kl": 0.6337890625,
      "learning_rate": 2.413333333333333e-07,
      "loss": 0.0253,
      "reward": 0.2847466543316841,
      "reward_std": 0.11508836224675179,
      "reward_std_all": 0.102185919880867,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9965371787548065,
      "rewards/forensics_tempsamp_iou_reward": 0.2847466543316841,
      "step": 1139
    },
    {
      "completion_length": 126.21875,
      "epoch": 3.0319573901464714,
      "grad_norm": 5.693447589874268,
      "kl": 0.578125,
      "learning_rate": 2.406666666666667e-07,
      "loss": 0.0231,
      "reward": 0.272081658244133,
      "reward_std": 0.09650543332099915,
      "reward_std_all": 0.11513088643550873,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9815532565116882,
      "rewards/forensics_tempsamp_iou_reward": 0.272081658244133,
      "step": 1140
    },
    {
      "completion_length": 106.875,
      "epoch": 3.034620505992011,
      "grad_norm": 6.539723873138428,
      "kl": 0.75,
      "learning_rate": 2.4e-07,
      "loss": 0.03,
      "reward": 0.2643757686018944,
      "reward_std": 0.1146576851606369,
      "reward_std_all": 0.11038945987820625,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9373786747455597,
      "rewards/forensics_tempsamp_iou_reward": 0.2643757686018944,
      "step": 1141
    },
    {
      "completion_length": 64.390625,
      "epoch": 3.03728362183755,
      "grad_norm": 5.231326103210449,
      "kl": 0.5263671875,
      "learning_rate": 2.3933333333333333e-07,
      "loss": 0.0211,
      "reward": 0.3601095378398895,
      "reward_std": 0.08724414557218552,
      "reward_std_all": 0.09536590427160263,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7822661399841309,
      "rewards/forensics_tempsamp_iou_reward": 0.3601095378398895,
      "step": 1142
    },
    {
      "completion_length": 104.328125,
      "epoch": 3.0399467376830893,
      "grad_norm": 7.273261070251465,
      "kl": 0.90234375,
      "learning_rate": 2.3866666666666663e-07,
      "loss": 0.036,
      "reward": 0.33980149030685425,
      "reward_std": 0.13536065071821213,
      "reward_std_all": 0.15044352039694786,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9931188821792603,
      "rewards/forensics_tempsamp_iou_reward": 0.33980149030685425,
      "step": 1143
    },
    {
      "completion_length": 92.0,
      "epoch": 3.0426098535286283,
      "grad_norm": 6.625125408172607,
      "kl": 0.791015625,
      "learning_rate": 2.38e-07,
      "loss": 0.0317,
      "reward": 0.2555737644433975,
      "reward_std": 0.07534820213913918,
      "reward_std_all": 0.07756281644105911,
      "rewards-max/forensics_tempsamp_iou_reward": 0.769696980714798,
      "rewards/forensics_tempsamp_iou_reward": 0.2555737644433975,
      "step": 1144
    },
    {
      "completion_length": 88.078125,
      "epoch": 3.045272969374168,
      "grad_norm": 6.546300888061523,
      "kl": 0.708984375,
      "learning_rate": 2.3733333333333334e-07,
      "loss": 0.0283,
      "reward": 0.21592473983764648,
      "reward_std": 0.08867931738495827,
      "reward_std_all": 0.09035325795412064,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9622685015201569,
      "rewards/forensics_tempsamp_iou_reward": 0.21592473983764648,
      "step": 1145
    },
    {
      "completion_length": 47.65625,
      "epoch": 3.0479360852197073,
      "grad_norm": 6.836873531341553,
      "kl": 0.4052734375,
      "learning_rate": 2.3666666666666664e-07,
      "loss": 0.0162,
      "reward": 0.2472710907459259,
      "reward_std": 0.06287017837166786,
      "reward_std_all": 0.0915961042046547,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8597943484783173,
      "rewards/forensics_tempsamp_iou_reward": 0.2472710907459259,
      "step": 1146
    },
    {
      "completion_length": 167.546875,
      "epoch": 3.0505992010652463,
      "grad_norm": 6.179069519042969,
      "kl": 0.80859375,
      "learning_rate": 2.3599999999999997e-07,
      "loss": 0.0323,
      "reward": 0.158024363219738,
      "reward_std": 0.06677292287349701,
      "reward_std_all": 0.07805789448320866,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5199999958276749,
      "rewards/forensics_tempsamp_iou_reward": 0.158024363219738,
      "step": 1147
    },
    {
      "completion_length": 128.9375,
      "epoch": 3.0532623169107858,
      "grad_norm": 8.219809532165527,
      "kl": 1.06640625,
      "learning_rate": 2.3533333333333332e-07,
      "loss": 0.0427,
      "reward": 0.10839350149035454,
      "reward_std": 0.051319751888513565,
      "reward_std_all": 0.05305122584104538,
      "rewards-max/forensics_tempsamp_iou_reward": 0.3922398090362549,
      "rewards/forensics_tempsamp_iou_reward": 0.10839350149035454,
      "step": 1148
    },
    {
      "completion_length": 84.671875,
      "epoch": 3.0559254327563248,
      "grad_norm": 5.421518325805664,
      "kl": 0.779296875,
      "learning_rate": 2.3466666666666665e-07,
      "loss": 0.0312,
      "reward": 0.22967709600925446,
      "reward_std": 0.053716765716671944,
      "reward_std_all": 0.05083256959915161,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9883987009525299,
      "rewards/forensics_tempsamp_iou_reward": 0.22967709600925446,
      "step": 1149
    },
    {
      "completion_length": 111.140625,
      "epoch": 3.0585885486018642,
      "grad_norm": 6.2333984375,
      "kl": 0.609375,
      "learning_rate": 2.34e-07,
      "loss": 0.0244,
      "reward": 0.17776089906692505,
      "reward_std": 0.07926357164978981,
      "reward_std_all": 0.08461557328701019,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5238789319992065,
      "rewards/forensics_tempsamp_iou_reward": 0.17776089906692505,
      "step": 1150
    },
    {
      "completion_length": 64.796875,
      "epoch": 3.0612516644474033,
      "grad_norm": 8.463163375854492,
      "kl": 0.63671875,
      "learning_rate": 2.3333333333333333e-07,
      "loss": 0.0255,
      "reward": 0.21604561805725098,
      "reward_std": 0.08467106893658638,
      "reward_std_all": 0.09806481003761292,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9852305054664612,
      "rewards/forensics_tempsamp_iou_reward": 0.21604561805725098,
      "step": 1151
    },
    {
      "completion_length": 115.890625,
      "epoch": 3.0639147802929427,
      "grad_norm": 6.844715595245361,
      "kl": 0.875,
      "learning_rate": 2.3266666666666666e-07,
      "loss": 0.035,
      "reward": 0.22310905158519745,
      "reward_std": 0.14345800131559372,
      "reward_std_all": 0.13141192495822906,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9155875444412231,
      "rewards/forensics_tempsamp_iou_reward": 0.22310905158519745,
      "step": 1152
    },
    {
      "completion_length": 102.34375,
      "epoch": 3.066577896138482,
      "grad_norm": 8.624492645263672,
      "kl": 0.85546875,
      "learning_rate": 2.32e-07,
      "loss": 0.0343,
      "reward": 0.1735125631093979,
      "reward_std": 0.10084949061274529,
      "reward_std_all": 0.09629471227526665,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6793661117553711,
      "rewards/forensics_tempsamp_iou_reward": 0.1735125631093979,
      "step": 1153
    },
    {
      "completion_length": 152.703125,
      "epoch": 3.069241011984021,
      "grad_norm": 7.305624008178711,
      "kl": 0.96484375,
      "learning_rate": 2.3133333333333331e-07,
      "loss": 0.0387,
      "reward": 0.30210262537002563,
      "reward_std": 0.14086302369832993,
      "reward_std_all": 0.14395495504140854,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9788510203361511,
      "rewards/forensics_tempsamp_iou_reward": 0.30210262537002563,
      "step": 1154
    },
    {
      "completion_length": 129.640625,
      "epoch": 3.0719041278295607,
      "grad_norm": 8.618168830871582,
      "kl": 1.0390625,
      "learning_rate": 2.3066666666666664e-07,
      "loss": 0.0416,
      "reward": 0.16763191670179367,
      "reward_std": 0.0465072188526392,
      "reward_std_all": 0.06477435864508152,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6207852065563202,
      "rewards/forensics_tempsamp_iou_reward": 0.16763191670179367,
      "step": 1155
    },
    {
      "completion_length": 72.265625,
      "epoch": 3.0745672436750997,
      "grad_norm": 7.168065547943115,
      "kl": 0.7890625,
      "learning_rate": 2.3e-07,
      "loss": 0.0316,
      "reward": 0.20566942542791367,
      "reward_std": 0.11692563444375992,
      "reward_std_all": 0.12419955432415009,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7861111164093018,
      "rewards/forensics_tempsamp_iou_reward": 0.20566942542791367,
      "step": 1156
    },
    {
      "completion_length": 67.5,
      "epoch": 3.077230359520639,
      "grad_norm": 6.158328056335449,
      "kl": 0.791015625,
      "learning_rate": 2.2933333333333332e-07,
      "loss": 0.0316,
      "reward": 0.31916213035583496,
      "reward_std": 0.07578055374324322,
      "reward_std_all": 0.07104069739580154,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9247180819511414,
      "rewards/forensics_tempsamp_iou_reward": 0.31916213035583496,
      "step": 1157
    },
    {
      "completion_length": 74.46875,
      "epoch": 3.0798934753661786,
      "grad_norm": 7.455006122589111,
      "kl": 0.66796875,
      "learning_rate": 2.2866666666666665e-07,
      "loss": 0.0267,
      "reward": 0.3135867863893509,
      "reward_std": 0.06980529427528381,
      "reward_std_all": 0.0908934697508812,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8637198209762573,
      "rewards/forensics_tempsamp_iou_reward": 0.3135867863893509,
      "step": 1158
    },
    {
      "completion_length": 61.171875,
      "epoch": 3.0825565912117177,
      "grad_norm": 11.410551071166992,
      "kl": 0.748046875,
      "learning_rate": 2.28e-07,
      "loss": 0.0299,
      "reward": 0.26305676251649857,
      "reward_std": 0.11106658726930618,
      "reward_std_all": 0.11152025312185287,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7825296223163605,
      "rewards/forensics_tempsamp_iou_reward": 0.26305676251649857,
      "step": 1159
    },
    {
      "completion_length": 51.4375,
      "epoch": 3.085219707057257,
      "grad_norm": 5.53942346572876,
      "kl": 0.73828125,
      "learning_rate": 2.2733333333333333e-07,
      "loss": 0.0295,
      "reward": 0.28067542612552643,
      "reward_std": 0.09541692212224007,
      "reward_std_all": 0.10586437210440636,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9938498735427856,
      "rewards/forensics_tempsamp_iou_reward": 0.28067542612552643,
      "step": 1160
    },
    {
      "completion_length": 114.6875,
      "epoch": 3.087882822902796,
      "grad_norm": 5.832479476928711,
      "kl": 0.671875,
      "learning_rate": 2.2666666666666663e-07,
      "loss": 0.0269,
      "reward": 0.2672603875398636,
      "reward_std": 0.13090533390641212,
      "reward_std_all": 0.132486704736948,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9523809552192688,
      "rewards/forensics_tempsamp_iou_reward": 0.2672603875398636,
      "step": 1161
    },
    {
      "completion_length": 40.15625,
      "epoch": 3.0905459387483356,
      "grad_norm": 7.193521022796631,
      "kl": 0.5986328125,
      "learning_rate": 2.2599999999999999e-07,
      "loss": 0.024,
      "reward": 0.25379496812820435,
      "reward_std": 0.06224989891052246,
      "reward_std_all": 0.07711951807141304,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7727646827697754,
      "rewards/forensics_tempsamp_iou_reward": 0.25379496812820435,
      "step": 1162
    },
    {
      "completion_length": 46.640625,
      "epoch": 3.0932090545938746,
      "grad_norm": 6.329503059387207,
      "kl": 0.611328125,
      "learning_rate": 2.253333333333333e-07,
      "loss": 0.0244,
      "reward": 0.29864954203367233,
      "reward_std": 0.09392189048230648,
      "reward_std_all": 0.10138364881277084,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8910406231880188,
      "rewards/forensics_tempsamp_iou_reward": 0.29864954203367233,
      "step": 1163
    },
    {
      "completion_length": 96.140625,
      "epoch": 3.095872170439414,
      "grad_norm": 7.84957218170166,
      "kl": 0.826171875,
      "learning_rate": 2.2466666666666664e-07,
      "loss": 0.033,
      "reward": 0.22447333484888077,
      "reward_std": 0.0987577959895134,
      "reward_std_all": 0.10771520063281059,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6985617429018021,
      "rewards/forensics_tempsamp_iou_reward": 0.22447333484888077,
      "step": 1164
    },
    {
      "completion_length": 102.6875,
      "epoch": 3.0985352862849536,
      "grad_norm": 7.6202521324157715,
      "kl": 1.060546875,
      "learning_rate": 2.24e-07,
      "loss": 0.0424,
      "reward": 0.23237751424312592,
      "reward_std": 0.08469769358634949,
      "reward_std_all": 0.10170454531908035,
      "rewards-max/forensics_tempsamp_iou_reward": 0.681517094373703,
      "rewards/forensics_tempsamp_iou_reward": 0.23237751424312592,
      "step": 1165
    },
    {
      "completion_length": 56.453125,
      "epoch": 3.1011984021304926,
      "grad_norm": 6.774279594421387,
      "kl": 0.55859375,
      "learning_rate": 2.2333333333333332e-07,
      "loss": 0.0223,
      "reward": 0.25765934586524963,
      "reward_std": 0.10296446830034256,
      "reward_std_all": 0.10769256576895714,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7568542659282684,
      "rewards/forensics_tempsamp_iou_reward": 0.25765934586524963,
      "step": 1166
    },
    {
      "completion_length": 98.078125,
      "epoch": 3.103861517976032,
      "grad_norm": 6.667318344116211,
      "kl": 0.732421875,
      "learning_rate": 2.2266666666666668e-07,
      "loss": 0.0293,
      "reward": 0.23270874470472336,
      "reward_std": 0.09667917340993881,
      "reward_std_all": 0.09445767849683762,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8496119976043701,
      "rewards/forensics_tempsamp_iou_reward": 0.23270874470472336,
      "step": 1167
    },
    {
      "completion_length": 63.53125,
      "epoch": 3.106524633821571,
      "grad_norm": 5.621517658233643,
      "kl": 0.732421875,
      "learning_rate": 2.22e-07,
      "loss": 0.0293,
      "reward": 0.26917197555303574,
      "reward_std": 0.09106821566820145,
      "reward_std_all": 0.09439460560679436,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8447267711162567,
      "rewards/forensics_tempsamp_iou_reward": 0.26917197555303574,
      "step": 1168
    },
    {
      "completion_length": 72.765625,
      "epoch": 3.1091877496671105,
      "grad_norm": 4.9334492683410645,
      "kl": 0.666015625,
      "learning_rate": 2.213333333333333e-07,
      "loss": 0.0267,
      "reward": 0.2199421301484108,
      "reward_std": 0.08154455199837685,
      "reward_std_all": 0.11468193680047989,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9763890206813812,
      "rewards/forensics_tempsamp_iou_reward": 0.2199421301484108,
      "step": 1169
    },
    {
      "completion_length": 97.984375,
      "epoch": 3.11185086551265,
      "grad_norm": 5.388285160064697,
      "kl": 0.67578125,
      "learning_rate": 2.2066666666666666e-07,
      "loss": 0.0271,
      "reward": 0.23666474223136902,
      "reward_std": 0.1101742535829544,
      "reward_std_all": 0.1188633032143116,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7765665352344513,
      "rewards/forensics_tempsamp_iou_reward": 0.23666474223136902,
      "step": 1170
    },
    {
      "completion_length": 65.109375,
      "epoch": 3.114513981358189,
      "grad_norm": 6.464552879333496,
      "kl": 0.59375,
      "learning_rate": 2.1999999999999998e-07,
      "loss": 0.0237,
      "reward": 0.20485220849514008,
      "reward_std": 0.08669929206371307,
      "reward_std_all": 0.09800456836819649,
      "rewards-max/forensics_tempsamp_iou_reward": 0.813923716545105,
      "rewards/forensics_tempsamp_iou_reward": 0.20485220849514008,
      "step": 1171
    },
    {
      "completion_length": 67.6875,
      "epoch": 3.1171770972037285,
      "grad_norm": 7.993134498596191,
      "kl": 0.78515625,
      "learning_rate": 2.193333333333333e-07,
      "loss": 0.0314,
      "reward": 0.24590960144996643,
      "reward_std": 0.07097167149186134,
      "reward_std_all": 0.12169954553246498,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8436939120292664,
      "rewards/forensics_tempsamp_iou_reward": 0.24590960144996643,
      "step": 1172
    },
    {
      "completion_length": 115.125,
      "epoch": 3.1198402130492675,
      "grad_norm": 5.532992839813232,
      "kl": 0.826171875,
      "learning_rate": 2.1866666666666667e-07,
      "loss": 0.033,
      "reward": 0.214205764234066,
      "reward_std": 0.11712571978569031,
      "reward_std_all": 0.12061946094036102,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6530701816082001,
      "rewards/forensics_tempsamp_iou_reward": 0.214205764234066,
      "step": 1173
    },
    {
      "completion_length": 110.265625,
      "epoch": 3.122503328894807,
      "grad_norm": 6.466548442840576,
      "kl": 0.64453125,
      "learning_rate": 2.18e-07,
      "loss": 0.0258,
      "reward": 0.15471692383289337,
      "reward_std": 0.09997487813234329,
      "reward_std_all": 0.1262662075459957,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6652184724807739,
      "rewards/forensics_tempsamp_iou_reward": 0.15471692383289337,
      "step": 1174
    },
    {
      "completion_length": 53.375,
      "epoch": 3.125166444740346,
      "grad_norm": 5.109834671020508,
      "kl": 0.470703125,
      "learning_rate": 2.1733333333333332e-07,
      "loss": 0.0188,
      "reward": 0.28153934329748154,
      "reward_std": 0.06046261079609394,
      "reward_std_all": 0.082325030118227,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6852461844682693,
      "rewards/forensics_tempsamp_iou_reward": 0.28153934329748154,
      "step": 1175
    },
    {
      "completion_length": 64.375,
      "epoch": 3.1278295605858855,
      "grad_norm": 6.122613906860352,
      "kl": 0.595703125,
      "learning_rate": 2.1666666666666667e-07,
      "loss": 0.0238,
      "reward": 0.2640286535024643,
      "reward_std": 0.10072978585958481,
      "reward_std_all": 0.10520706698298454,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9459174871444702,
      "rewards/forensics_tempsamp_iou_reward": 0.2640286535024643,
      "step": 1176
    },
    {
      "completion_length": 53.265625,
      "epoch": 3.130492676431425,
      "grad_norm": 6.441740036010742,
      "kl": 0.65625,
      "learning_rate": 2.1599999999999998e-07,
      "loss": 0.0263,
      "reward": 0.19210045039653778,
      "reward_std": 0.10986869037151337,
      "reward_std_all": 0.11918258294463158,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8300865888595581,
      "rewards/forensics_tempsamp_iou_reward": 0.19210045039653778,
      "step": 1177
    },
    {
      "completion_length": 58.984375,
      "epoch": 3.133155792276964,
      "grad_norm": 6.7893967628479,
      "kl": 0.578125,
      "learning_rate": 2.153333333333333e-07,
      "loss": 0.0232,
      "reward": 0.3409285694360733,
      "reward_std": 0.11620952188968658,
      "reward_std_all": 0.10602667182683945,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9846101403236389,
      "rewards/forensics_tempsamp_iou_reward": 0.3409285694360733,
      "step": 1178
    },
    {
      "completion_length": 45.578125,
      "epoch": 3.1358189081225034,
      "grad_norm": 5.7709269523620605,
      "kl": 0.5625,
      "learning_rate": 2.1466666666666666e-07,
      "loss": 0.0225,
      "reward": 0.312813863158226,
      "reward_std": 0.062466371804475784,
      "reward_std_all": 0.0730874352157116,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7488662004470825,
      "rewards/forensics_tempsamp_iou_reward": 0.312813863158226,
      "step": 1179
    },
    {
      "completion_length": 104.09375,
      "epoch": 3.1384820239680424,
      "grad_norm": 6.627326011657715,
      "kl": 0.7685546875,
      "learning_rate": 2.1399999999999998e-07,
      "loss": 0.0308,
      "reward": 0.4605318605899811,
      "reward_std": 0.17481380328536034,
      "reward_std_all": 0.15858691185712814,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9868018329143524,
      "rewards/forensics_tempsamp_iou_reward": 0.4605318605899811,
      "step": 1180
    },
    {
      "completion_length": 29.953125,
      "epoch": 3.141145139813582,
      "grad_norm": 9.130753517150879,
      "kl": 0.443359375,
      "learning_rate": 2.1333333333333334e-07,
      "loss": 0.0177,
      "reward": 0.3920207917690277,
      "reward_std": 0.07842610776424408,
      "reward_std_all": 0.08596914261579514,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9588868618011475,
      "rewards/forensics_tempsamp_iou_reward": 0.3920207917690277,
      "step": 1181
    },
    {
      "completion_length": 47.890625,
      "epoch": 3.1438082556591214,
      "grad_norm": 7.095854759216309,
      "kl": 0.83984375,
      "learning_rate": 2.1266666666666667e-07,
      "loss": 0.0336,
      "reward": 0.3131624013185501,
      "reward_std": 0.06258915364742279,
      "reward_std_all": 0.08735198900103569,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9693849384784698,
      "rewards/forensics_tempsamp_iou_reward": 0.3131624013185501,
      "step": 1182
    },
    {
      "completion_length": 42.171875,
      "epoch": 3.1464713715046604,
      "grad_norm": 7.061532497406006,
      "kl": 0.73046875,
      "learning_rate": 2.12e-07,
      "loss": 0.0292,
      "reward": 0.18978329002857208,
      "reward_std": 0.06563069485127926,
      "reward_std_all": 0.06899517215788364,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5847108587622643,
      "rewards/forensics_tempsamp_iou_reward": 0.18978329002857208,
      "step": 1183
    },
    {
      "completion_length": 53.28125,
      "epoch": 3.1491344873502,
      "grad_norm": 6.993721008300781,
      "kl": 0.599609375,
      "learning_rate": 2.1133333333333335e-07,
      "loss": 0.024,
      "reward": 0.31459367275238037,
      "reward_std": 0.0515576908364892,
      "reward_std_all": 0.0784112885594368,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9381968379020691,
      "rewards/forensics_tempsamp_iou_reward": 0.31459367275238037,
      "step": 1184
    },
    {
      "completion_length": 58.515625,
      "epoch": 3.151797603195739,
      "grad_norm": 8.096020698547363,
      "kl": 0.921875,
      "learning_rate": 2.1066666666666665e-07,
      "loss": 0.0368,
      "reward": 0.27127815783023834,
      "reward_std": 0.07237866520881653,
      "reward_std_all": 0.08078246936202049,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9600454270839691,
      "rewards/forensics_tempsamp_iou_reward": 0.27127815783023834,
      "step": 1185
    },
    {
      "completion_length": 65.984375,
      "epoch": 3.1544607190412783,
      "grad_norm": 6.890444278717041,
      "kl": 0.6748046875,
      "learning_rate": 2.0999999999999997e-07,
      "loss": 0.027,
      "reward": 0.34514106810092926,
      "reward_std": 0.10002769529819489,
      "reward_std_all": 0.13138297200202942,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9590723812580109,
      "rewards/forensics_tempsamp_iou_reward": 0.34514106810092926,
      "step": 1186
    },
    {
      "completion_length": 78.3125,
      "epoch": 3.157123834886818,
      "grad_norm": 7.004696369171143,
      "kl": 0.552734375,
      "learning_rate": 2.0933333333333333e-07,
      "loss": 0.0221,
      "reward": 0.307667538523674,
      "reward_std": 0.12439508736133575,
      "reward_std_all": 0.11332891136407852,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8664780855178833,
      "rewards/forensics_tempsamp_iou_reward": 0.307667538523674,
      "step": 1187
    },
    {
      "completion_length": 51.1875,
      "epoch": 3.159786950732357,
      "grad_norm": 6.950068950653076,
      "kl": 0.59375,
      "learning_rate": 2.0866666666666666e-07,
      "loss": 0.0237,
      "reward": 0.3381143808364868,
      "reward_std": 0.09312484785914421,
      "reward_std_all": 0.12139841541647911,
      "rewards-max/forensics_tempsamp_iou_reward": 0.939042866230011,
      "rewards/forensics_tempsamp_iou_reward": 0.3381143808364868,
      "step": 1188
    },
    {
      "completion_length": 94.921875,
      "epoch": 3.1624500665778963,
      "grad_norm": 8.692487716674805,
      "kl": 0.763671875,
      "learning_rate": 2.0799999999999998e-07,
      "loss": 0.0305,
      "reward": 0.3733155429363251,
      "reward_std": 0.06530355662107468,
      "reward_std_all": 0.06696042791008949,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9765926897525787,
      "rewards/forensics_tempsamp_iou_reward": 0.3733155429363251,
      "step": 1189
    },
    {
      "completion_length": 39.453125,
      "epoch": 3.1651131824234353,
      "grad_norm": 6.959124565124512,
      "kl": 0.4560546875,
      "learning_rate": 2.0733333333333334e-07,
      "loss": 0.0182,
      "reward": 0.3977026343345642,
      "reward_std": 0.08975305035710335,
      "reward_std_all": 0.09918109327554703,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9962891936302185,
      "rewards/forensics_tempsamp_iou_reward": 0.3977026343345642,
      "step": 1190
    },
    {
      "completion_length": 115.171875,
      "epoch": 3.1677762982689748,
      "grad_norm": 5.348171234130859,
      "kl": 0.791015625,
      "learning_rate": 2.0666666666666666e-07,
      "loss": 0.0317,
      "reward": 0.285156786441803,
      "reward_std": 0.11057813093066216,
      "reward_std_all": 0.10380540415644646,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9754655659198761,
      "rewards/forensics_tempsamp_iou_reward": 0.285156786441803,
      "step": 1191
    },
    {
      "completion_length": 25.828125,
      "epoch": 3.170439414114514,
      "grad_norm": 6.590680122375488,
      "kl": 0.4990234375,
      "learning_rate": 2.06e-07,
      "loss": 0.0199,
      "reward": 0.2758740931749344,
      "reward_std": 0.07824222557246685,
      "reward_std_all": 0.0737411342561245,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7602272629737854,
      "rewards/forensics_tempsamp_iou_reward": 0.2758740931749344,
      "step": 1192
    },
    {
      "completion_length": 44.03125,
      "epoch": 3.1731025299600533,
      "grad_norm": 8.665843963623047,
      "kl": 0.53125,
      "learning_rate": 2.0533333333333332e-07,
      "loss": 0.0212,
      "reward": 0.35170894861221313,
      "reward_std": 0.10134449228644371,
      "reward_std_all": 0.1256188005208969,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9764672219753265,
      "rewards/forensics_tempsamp_iou_reward": 0.35170894861221313,
      "step": 1193
    },
    {
      "completion_length": 56.390625,
      "epoch": 3.1757656458055927,
      "grad_norm": 7.36550760269165,
      "kl": 0.599609375,
      "learning_rate": 2.0466666666666665e-07,
      "loss": 0.024,
      "reward": 0.28053395450115204,
      "reward_std": 0.07242665626108646,
      "reward_std_all": 0.10255450010299683,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9097951054573059,
      "rewards/forensics_tempsamp_iou_reward": 0.28053395450115204,
      "step": 1194
    },
    {
      "completion_length": 67.75,
      "epoch": 3.1784287616511318,
      "grad_norm": 7.659611701965332,
      "kl": 0.953125,
      "learning_rate": 2.0399999999999997e-07,
      "loss": 0.0382,
      "reward": 0.3533894121646881,
      "reward_std": 0.08949054777622223,
      "reward_std_all": 0.09157858788967133,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9867724776268005,
      "rewards/forensics_tempsamp_iou_reward": 0.3533894121646881,
      "step": 1195
    },
    {
      "completion_length": 60.390625,
      "epoch": 3.181091877496671,
      "grad_norm": 7.269970417022705,
      "kl": 0.6015625,
      "learning_rate": 2.0333333333333333e-07,
      "loss": 0.0241,
      "reward": 0.195932075381279,
      "reward_std": 0.05755043588578701,
      "reward_std_all": 0.06570864468812943,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8722979724407196,
      "rewards/forensics_tempsamp_iou_reward": 0.195932075381279,
      "step": 1196
    },
    {
      "completion_length": 79.390625,
      "epoch": 3.1837549933422102,
      "grad_norm": 7.369988441467285,
      "kl": 0.7001953125,
      "learning_rate": 2.0266666666666666e-07,
      "loss": 0.028,
      "reward": 0.3268597722053528,
      "reward_std": 0.10784890502691269,
      "reward_std_all": 0.11368977651000023,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9572205245494843,
      "rewards/forensics_tempsamp_iou_reward": 0.3268597722053528,
      "step": 1197
    },
    {
      "completion_length": 91.015625,
      "epoch": 3.1864181091877497,
      "grad_norm": 6.749199867248535,
      "kl": 0.642578125,
      "learning_rate": 2.02e-07,
      "loss": 0.0257,
      "reward": 0.3975220024585724,
      "reward_std": 0.10347338393330574,
      "reward_std_all": 0.09541097283363342,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8355233073234558,
      "rewards/forensics_tempsamp_iou_reward": 0.3975220024585724,
      "step": 1198
    },
    {
      "completion_length": 67.703125,
      "epoch": 3.1890812250332887,
      "grad_norm": 7.6276469230651855,
      "kl": 0.716796875,
      "learning_rate": 2.0133333333333334e-07,
      "loss": 0.0287,
      "reward": 0.32552459836006165,
      "reward_std": 0.10054447874426842,
      "reward_std_all": 0.10746768489480019,
      "rewards-max/forensics_tempsamp_iou_reward": 0.895932137966156,
      "rewards/forensics_tempsamp_iou_reward": 0.32552459836006165,
      "step": 1199
    },
    {
      "completion_length": 49.453125,
      "epoch": 3.191744340878828,
      "grad_norm": 6.57889986038208,
      "kl": 0.8203125,
      "learning_rate": 2.0066666666666666e-07,
      "loss": 0.0328,
      "reward": 0.284493625164032,
      "reward_std": 0.10796669498085976,
      "reward_std_all": 0.12252397835254669,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8634147346019745,
      "rewards/forensics_tempsamp_iou_reward": 0.284493625164032,
      "step": 1200
    },
    {
      "completion_length": 90.171875,
      "epoch": 3.1944074567243677,
      "grad_norm": 6.79689359664917,
      "kl": 0.876953125,
      "learning_rate": 2e-07,
      "loss": 0.035,
      "reward": 0.23322013765573502,
      "reward_std": 0.10158748179674149,
      "reward_std_all": 0.10397923737764359,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8550642132759094,
      "rewards/forensics_tempsamp_iou_reward": 0.23322013765573502,
      "step": 1201
    },
    {
      "completion_length": 77.203125,
      "epoch": 3.1970705725699067,
      "grad_norm": 8.199649810791016,
      "kl": 0.74609375,
      "learning_rate": 1.9933333333333332e-07,
      "loss": 0.0298,
      "reward": 0.3330286964774132,
      "reward_std": 0.13410767167806625,
      "reward_std_all": 0.14053868502378464,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9685443341732025,
      "rewards/forensics_tempsamp_iou_reward": 0.3330286964774132,
      "step": 1202
    },
    {
      "completion_length": 54.46875,
      "epoch": 3.199733688415446,
      "grad_norm": 6.8982648849487305,
      "kl": 0.580078125,
      "learning_rate": 1.9866666666666665e-07,
      "loss": 0.0232,
      "reward": 0.3080194890499115,
      "reward_std": 0.13106068968772888,
      "reward_std_all": 0.11577067896723747,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8962681293487549,
      "rewards/forensics_tempsamp_iou_reward": 0.3080194890499115,
      "step": 1203
    },
    {
      "completion_length": 86.734375,
      "epoch": 3.202396804260985,
      "grad_norm": 6.825769901275635,
      "kl": 1.052734375,
      "learning_rate": 1.98e-07,
      "loss": 0.0421,
      "reward": 0.22010599076747894,
      "reward_std": 0.10655384138226509,
      "reward_std_all": 0.0977262370288372,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7853810787200928,
      "rewards/forensics_tempsamp_iou_reward": 0.22010599076747894,
      "step": 1204
    },
    {
      "completion_length": 81.328125,
      "epoch": 3.2050599201065246,
      "grad_norm": 7.085434436798096,
      "kl": 0.908203125,
      "learning_rate": 1.9733333333333333e-07,
      "loss": 0.0364,
      "reward": 0.280817911028862,
      "reward_std": 0.18351805955171585,
      "reward_std_all": 0.1735103316605091,
      "rewards-max/forensics_tempsamp_iou_reward": 0.857584148645401,
      "rewards/forensics_tempsamp_iou_reward": 0.280817911028862,
      "step": 1205
    },
    {
      "completion_length": 99.328125,
      "epoch": 3.207723035952064,
      "grad_norm": 7.037524223327637,
      "kl": 0.94921875,
      "learning_rate": 1.9666666666666665e-07,
      "loss": 0.0381,
      "reward": 0.2350759208202362,
      "reward_std": 0.1160447783768177,
      "reward_std_all": 0.1044556200504303,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9364444613456726,
      "rewards/forensics_tempsamp_iou_reward": 0.2350759208202362,
      "step": 1206
    },
    {
      "completion_length": 81.265625,
      "epoch": 3.210386151797603,
      "grad_norm": 5.354842185974121,
      "kl": 0.9453125,
      "learning_rate": 1.96e-07,
      "loss": 0.0378,
      "reward": 0.31579579412937164,
      "reward_std": 0.0701935812830925,
      "reward_std_all": 0.07522740587592125,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8014759123325348,
      "rewards/forensics_tempsamp_iou_reward": 0.31579579412937164,
      "step": 1207
    },
    {
      "completion_length": 95.03125,
      "epoch": 3.2130492676431426,
      "grad_norm": 8.742352485656738,
      "kl": 0.765625,
      "learning_rate": 1.953333333333333e-07,
      "loss": 0.0307,
      "reward": 0.18964910134673119,
      "reward_std": 0.08705414459109306,
      "reward_std_all": 0.10487401112914085,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4890776574611664,
      "rewards/forensics_tempsamp_iou_reward": 0.18964910134673119,
      "step": 1208
    },
    {
      "completion_length": 72.890625,
      "epoch": 3.2157123834886816,
      "grad_norm": 6.178894996643066,
      "kl": 0.779296875,
      "learning_rate": 1.9466666666666664e-07,
      "loss": 0.0312,
      "reward": 0.137306597083807,
      "reward_std": 0.08472953364253044,
      "reward_std_all": 0.08073698729276657,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6781132221221924,
      "rewards/forensics_tempsamp_iou_reward": 0.137306597083807,
      "step": 1209
    },
    {
      "completion_length": 95.640625,
      "epoch": 3.218375499334221,
      "grad_norm": 8.404915809631348,
      "kl": 1.072265625,
      "learning_rate": 1.94e-07,
      "loss": 0.0429,
      "reward": 0.22769935429096222,
      "reward_std": 0.12082964181900024,
      "reward_std_all": 0.1176779493689537,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7851361632347107,
      "rewards/forensics_tempsamp_iou_reward": 0.22769935429096222,
      "step": 1210
    },
    {
      "completion_length": 72.75,
      "epoch": 3.2210386151797605,
      "grad_norm": 6.4533867835998535,
      "kl": 0.65625,
      "learning_rate": 1.9333333333333332e-07,
      "loss": 0.0262,
      "reward": 0.2727743983268738,
      "reward_std": 0.094838697463274,
      "reward_std_all": 0.11168236285448074,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8449897170066833,
      "rewards/forensics_tempsamp_iou_reward": 0.2727743983268738,
      "step": 1211
    },
    {
      "completion_length": 78.90625,
      "epoch": 3.2237017310252996,
      "grad_norm": 7.779126167297363,
      "kl": 0.779296875,
      "learning_rate": 1.9266666666666667e-07,
      "loss": 0.0312,
      "reward": 0.2722291648387909,
      "reward_std": 0.09981723129749298,
      "reward_std_all": 0.098772332072258,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7558539807796478,
      "rewards/forensics_tempsamp_iou_reward": 0.2722291648387909,
      "step": 1212
    },
    {
      "completion_length": 52.9375,
      "epoch": 3.226364846870839,
      "grad_norm": 6.916533946990967,
      "kl": 0.482421875,
      "learning_rate": 1.92e-07,
      "loss": 0.0193,
      "reward": 0.26230648159980774,
      "reward_std": 0.08262824825942516,
      "reward_std_all": 0.072920773178339,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8794518113136292,
      "rewards/forensics_tempsamp_iou_reward": 0.26230648159980774,
      "step": 1213
    },
    {
      "completion_length": 91.53125,
      "epoch": 3.229027962716378,
      "grad_norm": 6.634772300720215,
      "kl": 0.591796875,
      "learning_rate": 1.9133333333333333e-07,
      "loss": 0.0237,
      "reward": 0.26317451894283295,
      "reward_std": 0.10313047841191292,
      "reward_std_all": 0.10342987626791,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9189271926879883,
      "rewards/forensics_tempsamp_iou_reward": 0.26317451894283295,
      "step": 1214
    },
    {
      "completion_length": 47.734375,
      "epoch": 3.2316910785619175,
      "grad_norm": 6.803066730499268,
      "kl": 0.51220703125,
      "learning_rate": 1.9066666666666668e-07,
      "loss": 0.0205,
      "reward": 0.4039500504732132,
      "reward_std": 0.07969009876251221,
      "reward_std_all": 0.1325049214065075,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9926102459430695,
      "rewards/forensics_tempsamp_iou_reward": 0.4039500504732132,
      "step": 1215
    },
    {
      "completion_length": 44.296875,
      "epoch": 3.2343541944074565,
      "grad_norm": 9.649839401245117,
      "kl": 0.51953125,
      "learning_rate": 1.8999999999999998e-07,
      "loss": 0.0207,
      "reward": 0.2164973020553589,
      "reward_std": 0.0890076756477356,
      "reward_std_all": 0.11996609717607498,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7451644837856293,
      "rewards/forensics_tempsamp_iou_reward": 0.2164973020553589,
      "step": 1216
    },
    {
      "completion_length": 94.0625,
      "epoch": 3.237017310252996,
      "grad_norm": 7.13119649887085,
      "kl": 0.716796875,
      "learning_rate": 1.893333333333333e-07,
      "loss": 0.0287,
      "reward": 0.3107936531305313,
      "reward_std": 0.12773104012012482,
      "reward_std_all": 0.1347476989030838,
      "rewards-max/forensics_tempsamp_iou_reward": 0.924782395362854,
      "rewards/forensics_tempsamp_iou_reward": 0.3107936531305313,
      "step": 1217
    },
    {
      "completion_length": 56.09375,
      "epoch": 3.2396804260985355,
      "grad_norm": 6.182520866394043,
      "kl": 0.6494140625,
      "learning_rate": 1.8866666666666666e-07,
      "loss": 0.026,
      "reward": 0.3124181926250458,
      "reward_std": 0.062023336067795753,
      "reward_std_all": 0.06370342150330544,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9835431277751923,
      "rewards/forensics_tempsamp_iou_reward": 0.3124181926250458,
      "step": 1218
    },
    {
      "completion_length": 120.515625,
      "epoch": 3.2423435419440745,
      "grad_norm": 7.922834396362305,
      "kl": 0.814453125,
      "learning_rate": 1.88e-07,
      "loss": 0.0326,
      "reward": 0.2804664149880409,
      "reward_std": 0.08848659321665764,
      "reward_std_all": 0.10076315701007843,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8553943634033203,
      "rewards/forensics_tempsamp_iou_reward": 0.2804664149880409,
      "step": 1219
    },
    {
      "completion_length": 40.5625,
      "epoch": 3.245006657789614,
      "grad_norm": 7.921935081481934,
      "kl": 0.638671875,
      "learning_rate": 1.8733333333333332e-07,
      "loss": 0.0256,
      "reward": 0.3371106907725334,
      "reward_std": 0.06170198507606983,
      "reward_std_all": 0.06007438525557518,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8896845281124115,
      "rewards/forensics_tempsamp_iou_reward": 0.3371106907725334,
      "step": 1220
    },
    {
      "completion_length": 79.28125,
      "epoch": 3.247669773635153,
      "grad_norm": 6.329800605773926,
      "kl": 0.54296875,
      "learning_rate": 1.8666666666666667e-07,
      "loss": 0.0217,
      "reward": 0.2147998809814453,
      "reward_std": 0.06772571429610252,
      "reward_std_all": 0.07471825182437897,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6905695199966431,
      "rewards/forensics_tempsamp_iou_reward": 0.2147998809814453,
      "step": 1221
    },
    {
      "completion_length": 91.6875,
      "epoch": 3.2503328894806924,
      "grad_norm": 5.968420028686523,
      "kl": 0.7265625,
      "learning_rate": 1.86e-07,
      "loss": 0.029,
      "reward": 0.26257409155368805,
      "reward_std": 0.07945161312818527,
      "reward_std_all": 0.07534663379192352,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7129756659269333,
      "rewards/forensics_tempsamp_iou_reward": 0.26257409155368805,
      "step": 1222
    },
    {
      "completion_length": 73.6875,
      "epoch": 3.2529960053262315,
      "grad_norm": 6.164752006530762,
      "kl": 0.783203125,
      "learning_rate": 1.8533333333333333e-07,
      "loss": 0.0314,
      "reward": 0.1505475640296936,
      "reward_std": 0.06462943740189075,
      "reward_std_all": 0.07365786656737328,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5070450156927109,
      "rewards/forensics_tempsamp_iou_reward": 0.1505475640296936,
      "step": 1223
    },
    {
      "completion_length": 85.625,
      "epoch": 3.255659121171771,
      "grad_norm": 6.227341651916504,
      "kl": 0.51171875,
      "learning_rate": 1.8466666666666665e-07,
      "loss": 0.0205,
      "reward": 0.304954931139946,
      "reward_std": 0.09776220843195915,
      "reward_std_all": 0.11948972940444946,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7931970357894897,
      "rewards/forensics_tempsamp_iou_reward": 0.304954931139946,
      "step": 1224
    },
    {
      "completion_length": 52.84375,
      "epoch": 3.2583222370173104,
      "grad_norm": 6.236211776733398,
      "kl": 0.66796875,
      "learning_rate": 1.8399999999999998e-07,
      "loss": 0.0267,
      "reward": 0.3301016688346863,
      "reward_std": 0.10373107343912125,
      "reward_std_all": 0.14863254129886627,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8867472112178802,
      "rewards/forensics_tempsamp_iou_reward": 0.3301016688346863,
      "step": 1225
    },
    {
      "completion_length": 104.1875,
      "epoch": 3.2609853528628494,
      "grad_norm": 8.753555297851562,
      "kl": 0.7265625,
      "learning_rate": 1.833333333333333e-07,
      "loss": 0.0291,
      "reward": 0.18727784603834152,
      "reward_std": 0.09909499064087868,
      "reward_std_all": 0.10401948168873787,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9312140941619873,
      "rewards/forensics_tempsamp_iou_reward": 0.18727784603834152,
      "step": 1226
    },
    {
      "completion_length": 101.515625,
      "epoch": 3.263648468708389,
      "grad_norm": 9.73619270324707,
      "kl": 0.900390625,
      "learning_rate": 1.8266666666666666e-07,
      "loss": 0.0361,
      "reward": 0.2172275334596634,
      "reward_std": 0.14717848598957062,
      "reward_std_all": 0.1379612684249878,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7866727709770203,
      "rewards/forensics_tempsamp_iou_reward": 0.2172275334596634,
      "step": 1227
    },
    {
      "completion_length": 77.609375,
      "epoch": 3.266311584553928,
      "grad_norm": 7.73088264465332,
      "kl": 0.837890625,
      "learning_rate": 1.82e-07,
      "loss": 0.0336,
      "reward": 0.3167189657688141,
      "reward_std": 0.06849664449691772,
      "reward_std_all": 0.10103784129023552,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9036461710929871,
      "rewards/forensics_tempsamp_iou_reward": 0.3167189657688141,
      "step": 1228
    },
    {
      "completion_length": 70.828125,
      "epoch": 3.2689747003994674,
      "grad_norm": 13.329157829284668,
      "kl": 0.837890625,
      "learning_rate": 1.8133333333333334e-07,
      "loss": 0.0335,
      "reward": 0.25140901654958725,
      "reward_std": 0.10886141657829285,
      "reward_std_all": 0.11463442072272301,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7917292416095734,
      "rewards/forensics_tempsamp_iou_reward": 0.25140901654958725,
      "step": 1229
    },
    {
      "completion_length": 47.234375,
      "epoch": 3.271637816245007,
      "grad_norm": 7.497112274169922,
      "kl": 0.982421875,
      "learning_rate": 1.8066666666666667e-07,
      "loss": 0.0393,
      "reward": 0.19886645674705505,
      "reward_std": 0.10364828258752823,
      "reward_std_all": 0.12392283976078033,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7117251455783844,
      "rewards/forensics_tempsamp_iou_reward": 0.19886645674705505,
      "step": 1230
    },
    {
      "completion_length": 81.34375,
      "epoch": 3.274300932090546,
      "grad_norm": 10.134871482849121,
      "kl": 1.1953125,
      "learning_rate": 1.8e-07,
      "loss": 0.0479,
      "reward": 0.31535008549690247,
      "reward_std": 0.1116204485297203,
      "reward_std_all": 0.10610290616750717,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9333932399749756,
      "rewards/forensics_tempsamp_iou_reward": 0.31535008549690247,
      "step": 1231
    },
    {
      "completion_length": 89.875,
      "epoch": 3.2769640479360853,
      "grad_norm": 9.06727409362793,
      "kl": 0.671875,
      "learning_rate": 1.7933333333333332e-07,
      "loss": 0.0269,
      "reward": 0.45191721618175507,
      "reward_std": 0.12165244668722153,
      "reward_std_all": 0.13307978212833405,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9729747772216797,
      "rewards/forensics_tempsamp_iou_reward": 0.45191721618175507,
      "step": 1232
    },
    {
      "completion_length": 110.640625,
      "epoch": 3.2796271637816243,
      "grad_norm": 6.5265936851501465,
      "kl": 0.9453125,
      "learning_rate": 1.7866666666666665e-07,
      "loss": 0.0377,
      "reward": 0.19061671197414398,
      "reward_std": 0.06608499586582184,
      "reward_std_all": 0.08135613426566124,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7886094748973846,
      "rewards/forensics_tempsamp_iou_reward": 0.19061671197414398,
      "step": 1233
    },
    {
      "completion_length": 84.796875,
      "epoch": 3.282290279627164,
      "grad_norm": 6.3928728103637695,
      "kl": 0.505859375,
      "learning_rate": 1.7799999999999998e-07,
      "loss": 0.0202,
      "reward": 0.3674357533454895,
      "reward_std": 0.08797838911414146,
      "reward_std_all": 0.09118389710783958,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8377652764320374,
      "rewards/forensics_tempsamp_iou_reward": 0.3674357533454895,
      "step": 1234
    },
    {
      "completion_length": 96.34375,
      "epoch": 3.2849533954727033,
      "grad_norm": 4.639517784118652,
      "kl": 0.666015625,
      "learning_rate": 1.7733333333333333e-07,
      "loss": 0.0266,
      "reward": 0.2043413743376732,
      "reward_std": 0.07960284315049648,
      "reward_std_all": 0.0849878117442131,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7789132297039032,
      "rewards/forensics_tempsamp_iou_reward": 0.2043413743376732,
      "step": 1235
    },
    {
      "completion_length": 98.296875,
      "epoch": 3.2876165113182423,
      "grad_norm": 7.690802097320557,
      "kl": 0.66015625,
      "learning_rate": 1.7666666666666666e-07,
      "loss": 0.0263,
      "reward": 0.25177479535341263,
      "reward_std": 0.09518799185752869,
      "reward_std_all": 0.09184396266937256,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8477124273777008,
      "rewards/forensics_tempsamp_iou_reward": 0.25177479535341263,
      "step": 1236
    },
    {
      "completion_length": 64.09375,
      "epoch": 3.2902796271637818,
      "grad_norm": 6.489963531494141,
      "kl": 0.7109375,
      "learning_rate": 1.76e-07,
      "loss": 0.0285,
      "reward": 0.2610856220126152,
      "reward_std": 0.13498622924089432,
      "reward_std_all": 0.13629699498414993,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9124448895454407,
      "rewards/forensics_tempsamp_iou_reward": 0.2610856220126152,
      "step": 1237
    },
    {
      "completion_length": 69.03125,
      "epoch": 3.2929427430093208,
      "grad_norm": 8.33909797668457,
      "kl": 0.841796875,
      "learning_rate": 1.7533333333333334e-07,
      "loss": 0.0337,
      "reward": 0.30135220289230347,
      "reward_std": 0.11156536266207695,
      "reward_std_all": 0.10829519852995872,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9321581125259399,
      "rewards/forensics_tempsamp_iou_reward": 0.30135220289230347,
      "step": 1238
    },
    {
      "completion_length": 80.328125,
      "epoch": 3.2956058588548602,
      "grad_norm": 8.916841506958008,
      "kl": 0.91796875,
      "learning_rate": 1.7466666666666667e-07,
      "loss": 0.0367,
      "reward": 0.14453956857323647,
      "reward_std": 0.07462940365076065,
      "reward_std_all": 0.07614706829190254,
      "rewards-max/forensics_tempsamp_iou_reward": 0.4475095719099045,
      "rewards/forensics_tempsamp_iou_reward": 0.14453956857323647,
      "step": 1239
    },
    {
      "completion_length": 105.546875,
      "epoch": 3.2982689747003997,
      "grad_norm": 7.78352165222168,
      "kl": 1.11328125,
      "learning_rate": 1.7399999999999997e-07,
      "loss": 0.0444,
      "reward": 0.3514484167098999,
      "reward_std": 0.15748226642608643,
      "reward_std_all": 0.14555467665195465,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9877633452415466,
      "rewards/forensics_tempsamp_iou_reward": 0.3514484167098999,
      "step": 1240
    },
    {
      "completion_length": 63.9375,
      "epoch": 3.3009320905459387,
      "grad_norm": 7.0938005447387695,
      "kl": 0.681640625,
      "learning_rate": 1.7333333333333332e-07,
      "loss": 0.0272,
      "reward": 0.3238740488886833,
      "reward_std": 0.12838418781757355,
      "reward_std_all": 0.12055249512195587,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8079868853092194,
      "rewards/forensics_tempsamp_iou_reward": 0.3238740488886833,
      "step": 1241
    },
    {
      "completion_length": 57.578125,
      "epoch": 3.303595206391478,
      "grad_norm": 7.109021186828613,
      "kl": 0.66015625,
      "learning_rate": 1.7266666666666665e-07,
      "loss": 0.0264,
      "reward": 0.2557827830314636,
      "reward_std": 0.09992325305938721,
      "reward_std_all": 0.10403145477175713,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8468768894672394,
      "rewards/forensics_tempsamp_iou_reward": 0.2557827830314636,
      "step": 1242
    },
    {
      "completion_length": 90.5,
      "epoch": 3.306258322237017,
      "grad_norm": 6.694886684417725,
      "kl": 0.859375,
      "learning_rate": 1.7199999999999998e-07,
      "loss": 0.0344,
      "reward": 0.2853540927171707,
      "reward_std": 0.11882907152175903,
      "reward_std_all": 0.11366017907857895,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8422947525978088,
      "rewards/forensics_tempsamp_iou_reward": 0.2853540927171707,
      "step": 1243
    },
    {
      "completion_length": 84.171875,
      "epoch": 3.3089214380825567,
      "grad_norm": 7.443741798400879,
      "kl": 0.83203125,
      "learning_rate": 1.7133333333333333e-07,
      "loss": 0.0332,
      "reward": 0.2580152750015259,
      "reward_std": 0.11450966820120811,
      "reward_std_all": 0.10593164339661598,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9236755073070526,
      "rewards/forensics_tempsamp_iou_reward": 0.2580152750015259,
      "step": 1244
    },
    {
      "completion_length": 66.28125,
      "epoch": 3.3115845539280957,
      "grad_norm": 9.799491882324219,
      "kl": 0.7578125,
      "learning_rate": 1.7066666666666666e-07,
      "loss": 0.0304,
      "reward": 0.2875992953777313,
      "reward_std": 0.1211678721010685,
      "reward_std_all": 0.11953407526016235,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8578438460826874,
      "rewards/forensics_tempsamp_iou_reward": 0.2875992953777313,
      "step": 1245
    },
    {
      "completion_length": 85.21875,
      "epoch": 3.314247669773635,
      "grad_norm": 7.7520270347595215,
      "kl": 0.91796875,
      "learning_rate": 1.7000000000000001e-07,
      "loss": 0.0367,
      "reward": 0.2281055971980095,
      "reward_std": 0.09532063081860542,
      "reward_std_all": 0.09047384932637215,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6822222024202347,
      "rewards/forensics_tempsamp_iou_reward": 0.2281055971980095,
      "step": 1246
    },
    {
      "completion_length": 47.21875,
      "epoch": 3.316910785619174,
      "grad_norm": 7.774498462677002,
      "kl": 0.685546875,
      "learning_rate": 1.6933333333333334e-07,
      "loss": 0.0274,
      "reward": 0.30641019344329834,
      "reward_std": 0.08440543711185455,
      "reward_std_all": 0.08976110443472862,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9651454091072083,
      "rewards/forensics_tempsamp_iou_reward": 0.30641019344329834,
      "step": 1247
    },
    {
      "completion_length": 65.53125,
      "epoch": 3.3195739014647137,
      "grad_norm": 5.936551094055176,
      "kl": 0.6640625,
      "learning_rate": 1.6866666666666664e-07,
      "loss": 0.0266,
      "reward": 0.36745092272758484,
      "reward_std": 0.060527253895998,
      "reward_std_all": 0.062069082632660866,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9989428818225861,
      "rewards/forensics_tempsamp_iou_reward": 0.36745092272758484,
      "step": 1248
    },
    {
      "completion_length": 90.3125,
      "epoch": 3.322237017310253,
      "grad_norm": 5.541586399078369,
      "kl": 0.81640625,
      "learning_rate": 1.68e-07,
      "loss": 0.0327,
      "reward": 0.30645161867141724,
      "reward_std": 0.05371293984353542,
      "reward_std_all": 0.07286123000085354,
      "rewards-max/forensics_tempsamp_iou_reward": 0.991213858127594,
      "rewards/forensics_tempsamp_iou_reward": 0.30645161867141724,
      "step": 1249
    },
    {
      "completion_length": 89.09375,
      "epoch": 3.324900133155792,
      "grad_norm": 7.436034202575684,
      "kl": 0.72265625,
      "learning_rate": 1.6733333333333332e-07,
      "loss": 0.0289,
      "reward": 0.2235053852200508,
      "reward_std": 0.09708945825695992,
      "reward_std_all": 0.09804615750908852,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7038426101207733,
      "rewards/forensics_tempsamp_iou_reward": 0.2235053852200508,
      "step": 1250
    },
    {
      "completion_length": 97.125,
      "epoch": 3.3275632490013316,
      "grad_norm": 7.924428939819336,
      "kl": 0.974609375,
      "learning_rate": 1.6666666666666665e-07,
      "loss": 0.039,
      "reward": 0.3638482540845871,
      "reward_std": 0.06033090315759182,
      "reward_std_all": 0.05964277870953083,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9921700358390808,
      "rewards/forensics_tempsamp_iou_reward": 0.3638482540845871,
      "step": 1251
    },
    {
      "completion_length": 52.59375,
      "epoch": 3.3302263648468706,
      "grad_norm": 6.794558525085449,
      "kl": 0.80078125,
      "learning_rate": 1.66e-07,
      "loss": 0.0321,
      "reward": 0.3292993754148483,
      "reward_std": 0.16719362884759903,
      "reward_std_all": 0.15796734392642975,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8704253435134888,
      "rewards/forensics_tempsamp_iou_reward": 0.3292993754148483,
      "step": 1252
    },
    {
      "completion_length": 71.921875,
      "epoch": 3.33288948069241,
      "grad_norm": 7.241439342498779,
      "kl": 0.875,
      "learning_rate": 1.6533333333333333e-07,
      "loss": 0.035,
      "reward": 0.2515905052423477,
      "reward_std": 0.12725257873535156,
      "reward_std_all": 0.13586955145001411,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9898836612701416,
      "rewards/forensics_tempsamp_iou_reward": 0.2515905052423477,
      "step": 1253
    },
    {
      "completion_length": 110.625,
      "epoch": 3.3355525965379496,
      "grad_norm": 7.065522193908691,
      "kl": 0.7109375,
      "learning_rate": 1.6466666666666666e-07,
      "loss": 0.0285,
      "reward": 0.2806006893515587,
      "reward_std": 0.11017508059740067,
      "reward_std_all": 0.11955806240439415,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9631915092468262,
      "rewards/forensics_tempsamp_iou_reward": 0.2806006893515587,
      "step": 1254
    },
    {
      "completion_length": 70.46875,
      "epoch": 3.3382157123834886,
      "grad_norm": 7.637517929077148,
      "kl": 0.796875,
      "learning_rate": 1.64e-07,
      "loss": 0.0319,
      "reward": 0.37788714468479156,
      "reward_std": 0.12854871898889542,
      "reward_std_all": 0.13435672968626022,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9533263444900513,
      "rewards/forensics_tempsamp_iou_reward": 0.37788714468479156,
      "step": 1255
    },
    {
      "completion_length": 51.9375,
      "epoch": 3.340878828229028,
      "grad_norm": 7.628133773803711,
      "kl": 0.6123046875,
      "learning_rate": 1.6333333333333331e-07,
      "loss": 0.0245,
      "reward": 0.29356252402067184,
      "reward_std": 0.08787409588694572,
      "reward_std_all": 0.08885189145803452,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8331678211688995,
      "rewards/forensics_tempsamp_iou_reward": 0.29356252402067184,
      "step": 1256
    },
    {
      "completion_length": 65.625,
      "epoch": 3.343541944074567,
      "grad_norm": 7.673868656158447,
      "kl": 1.05078125,
      "learning_rate": 1.6266666666666664e-07,
      "loss": 0.0419,
      "reward": 0.33511099219322205,
      "reward_std": 0.08962149731814861,
      "reward_std_all": 0.13385408371686935,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9935097992420197,
      "rewards/forensics_tempsamp_iou_reward": 0.33511099219322205,
      "step": 1257
    },
    {
      "completion_length": 73.484375,
      "epoch": 3.3462050599201065,
      "grad_norm": 8.40373420715332,
      "kl": 0.912109375,
      "learning_rate": 1.62e-07,
      "loss": 0.0365,
      "reward": 0.212518110871315,
      "reward_std": 0.08834139257669449,
      "reward_std_all": 0.10037051886320114,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6364718675613403,
      "rewards/forensics_tempsamp_iou_reward": 0.212518110871315,
      "step": 1258
    },
    {
      "completion_length": 126.125,
      "epoch": 3.348868175765646,
      "grad_norm": 7.112109184265137,
      "kl": 0.796875,
      "learning_rate": 1.6133333333333332e-07,
      "loss": 0.0319,
      "reward": 0.21377387642860413,
      "reward_std": 0.12814389169216156,
      "reward_std_all": 0.11893847212195396,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9536982774734497,
      "rewards/forensics_tempsamp_iou_reward": 0.21377387642860413,
      "step": 1259
    },
    {
      "completion_length": 66.015625,
      "epoch": 3.351531291611185,
      "grad_norm": 8.498636245727539,
      "kl": 0.916015625,
      "learning_rate": 1.6066666666666668e-07,
      "loss": 0.0367,
      "reward": 0.2325425073504448,
      "reward_std": 0.1282385215163231,
      "reward_std_all": 0.12522584944963455,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7419823408126831,
      "rewards/forensics_tempsamp_iou_reward": 0.2325425073504448,
      "step": 1260
    },
    {
      "completion_length": 77.9375,
      "epoch": 3.3541944074567245,
      "grad_norm": 6.622129440307617,
      "kl": 0.884765625,
      "learning_rate": 1.6e-07,
      "loss": 0.0354,
      "reward": 0.31420332193374634,
      "reward_std": 0.12173480540513992,
      "reward_std_all": 0.1112046092748642,
      "rewards-max/forensics_tempsamp_iou_reward": 0.906694769859314,
      "rewards/forensics_tempsamp_iou_reward": 0.31420332193374634,
      "step": 1261
    },
    {
      "completion_length": 89.609375,
      "epoch": 3.3568575233022635,
      "grad_norm": 8.548338890075684,
      "kl": 0.80078125,
      "learning_rate": 1.5933333333333333e-07,
      "loss": 0.0321,
      "reward": 0.256439171731472,
      "reward_std": 0.09835086017847061,
      "reward_std_all": 0.10008477419614792,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8632510006427765,
      "rewards/forensics_tempsamp_iou_reward": 0.256439171731472,
      "step": 1262
    },
    {
      "completion_length": 60.953125,
      "epoch": 3.359520639147803,
      "grad_norm": 6.037802696228027,
      "kl": 0.701171875,
      "learning_rate": 1.5866666666666666e-07,
      "loss": 0.028,
      "reward": 0.31817641854286194,
      "reward_std": 0.15521075576543808,
      "reward_std_all": 0.15121563524007797,
      "rewards-max/forensics_tempsamp_iou_reward": 0.911077231168747,
      "rewards/forensics_tempsamp_iou_reward": 0.31817641854286194,
      "step": 1263
    },
    {
      "completion_length": 50.0,
      "epoch": 3.3621837549933424,
      "grad_norm": 7.873925685882568,
      "kl": 0.736328125,
      "learning_rate": 1.5799999999999999e-07,
      "loss": 0.0294,
      "reward": 0.24337458610534668,
      "reward_std": 0.11801907792687416,
      "reward_std_all": 0.1053500548005104,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8609006702899933,
      "rewards/forensics_tempsamp_iou_reward": 0.24337458610534668,
      "step": 1264
    },
    {
      "completion_length": 87.59375,
      "epoch": 3.3648468708388815,
      "grad_norm": 6.866157054901123,
      "kl": 0.71484375,
      "learning_rate": 1.573333333333333e-07,
      "loss": 0.0286,
      "reward": 0.2586664482951164,
      "reward_std": 0.11090590432286263,
      "reward_std_all": 0.11000753939151764,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8622449040412903,
      "rewards/forensics_tempsamp_iou_reward": 0.2586664482951164,
      "step": 1265
    },
    {
      "completion_length": 91.6875,
      "epoch": 3.367509986684421,
      "grad_norm": 6.113330841064453,
      "kl": 0.80078125,
      "learning_rate": 1.5666666666666667e-07,
      "loss": 0.032,
      "reward": 0.21946076676249504,
      "reward_std": 0.12008372321724892,
      "reward_std_all": 0.1255035065114498,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6573496609926224,
      "rewards/forensics_tempsamp_iou_reward": 0.21946076676249504,
      "step": 1266
    },
    {
      "completion_length": 53.71875,
      "epoch": 3.37017310252996,
      "grad_norm": 6.2533464431762695,
      "kl": 0.6044921875,
      "learning_rate": 1.56e-07,
      "loss": 0.0242,
      "reward": 0.2645096331834793,
      "reward_std": 0.0846768356859684,
      "reward_std_all": 0.09473946690559387,
      "rewards-max/forensics_tempsamp_iou_reward": 0.820836067199707,
      "rewards/forensics_tempsamp_iou_reward": 0.2645096331834793,
      "step": 1267
    },
    {
      "completion_length": 79.875,
      "epoch": 3.3728362183754994,
      "grad_norm": 10.109810829162598,
      "kl": 0.767578125,
      "learning_rate": 1.5533333333333332e-07,
      "loss": 0.0307,
      "reward": 0.28479818999767303,
      "reward_std": 0.08314028009772301,
      "reward_std_all": 0.09051605686545372,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9971043169498444,
      "rewards/forensics_tempsamp_iou_reward": 0.28479818999767303,
      "step": 1268
    },
    {
      "completion_length": 52.109375,
      "epoch": 3.3754993342210384,
      "grad_norm": 5.720717906951904,
      "kl": 0.677734375,
      "learning_rate": 1.5466666666666668e-07,
      "loss": 0.0271,
      "reward": 0.36638782918453217,
      "reward_std": 0.08864573016762733,
      "reward_std_all": 0.08012506365776062,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9759660363197327,
      "rewards/forensics_tempsamp_iou_reward": 0.36638782918453217,
      "step": 1269
    },
    {
      "completion_length": 47.0625,
      "epoch": 3.378162450066578,
      "grad_norm": 7.694883823394775,
      "kl": 0.7646484375,
      "learning_rate": 1.54e-07,
      "loss": 0.0306,
      "reward": 0.31039924919605255,
      "reward_std": 0.04636826738715172,
      "reward_std_all": 0.07649841904640198,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9308226704597473,
      "rewards/forensics_tempsamp_iou_reward": 0.31039924919605255,
      "step": 1270
    },
    {
      "completion_length": 103.65625,
      "epoch": 3.3808255659121174,
      "grad_norm": 7.292233467102051,
      "kl": 0.8671875,
      "learning_rate": 1.533333333333333e-07,
      "loss": 0.0347,
      "reward": 0.1964259222149849,
      "reward_std": 0.04273061268031597,
      "reward_std_all": 0.043063120916485786,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9536105096340179,
      "rewards/forensics_tempsamp_iou_reward": 0.1964259222149849,
      "step": 1271
    },
    {
      "completion_length": 62.0,
      "epoch": 3.3834886817576564,
      "grad_norm": 5.969642162322998,
      "kl": 0.568359375,
      "learning_rate": 1.5266666666666666e-07,
      "loss": 0.0227,
      "reward": 0.32552149146795273,
      "reward_std": 0.11947114765644073,
      "reward_std_all": 0.11012502759695053,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8658243417739868,
      "rewards/forensics_tempsamp_iou_reward": 0.32552149146795273,
      "step": 1272
    },
    {
      "completion_length": 93.28125,
      "epoch": 3.386151797603196,
      "grad_norm": 7.6904449462890625,
      "kl": 0.98046875,
      "learning_rate": 1.5199999999999998e-07,
      "loss": 0.0392,
      "reward": 0.2949373573064804,
      "reward_std": 0.11989337205886841,
      "reward_std_all": 0.1185198649764061,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8864759504795074,
      "rewards/forensics_tempsamp_iou_reward": 0.2949373573064804,
      "step": 1273
    },
    {
      "completion_length": 90.078125,
      "epoch": 3.388814913448735,
      "grad_norm": 7.049698352813721,
      "kl": 0.603515625,
      "learning_rate": 1.513333333333333e-07,
      "loss": 0.0242,
      "reward": 0.36200010776519775,
      "reward_std": 0.12810205295681953,
      "reward_std_all": 0.1318897120654583,
      "rewards-max/forensics_tempsamp_iou_reward": 0.967074990272522,
      "rewards/forensics_tempsamp_iou_reward": 0.36200010776519775,
      "step": 1274
    },
    {
      "completion_length": 80.578125,
      "epoch": 3.3914780292942743,
      "grad_norm": 5.925412178039551,
      "kl": 0.5029296875,
      "learning_rate": 1.5066666666666667e-07,
      "loss": 0.0201,
      "reward": 0.22262609004974365,
      "reward_std": 0.11161864548921585,
      "reward_std_all": 0.11617409437894821,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8054899871349335,
      "rewards/forensics_tempsamp_iou_reward": 0.22262609004974365,
      "step": 1275
    },
    {
      "completion_length": 82.765625,
      "epoch": 3.3941411451398134,
      "grad_norm": 6.374367713928223,
      "kl": 0.552734375,
      "learning_rate": 1.5e-07,
      "loss": 0.0221,
      "reward": 0.3450249582529068,
      "reward_std": 0.18803804367780685,
      "reward_std_all": 0.16544652730226517,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9664983749389648,
      "rewards/forensics_tempsamp_iou_reward": 0.3450249582529068,
      "step": 1276
    },
    {
      "completion_length": 102.75,
      "epoch": 3.396804260985353,
      "grad_norm": 7.852759838104248,
      "kl": 0.73046875,
      "learning_rate": 1.4933333333333335e-07,
      "loss": 0.0292,
      "reward": 0.282540887594223,
      "reward_std": 0.12086999416351318,
      "reward_std_all": 0.1103816106915474,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9882886409759521,
      "rewards/forensics_tempsamp_iou_reward": 0.282540887594223,
      "step": 1277
    },
    {
      "completion_length": 45.3125,
      "epoch": 3.3994673768308923,
      "grad_norm": 9.617256164550781,
      "kl": 0.625,
      "learning_rate": 1.4866666666666667e-07,
      "loss": 0.025,
      "reward": 0.3707738518714905,
      "reward_std": 0.11110371351242065,
      "reward_std_all": 0.10755520686507225,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8436302840709686,
      "rewards/forensics_tempsamp_iou_reward": 0.3707738518714905,
      "step": 1278
    },
    {
      "completion_length": 60.59375,
      "epoch": 3.4021304926764313,
      "grad_norm": 5.887720584869385,
      "kl": 0.697265625,
      "learning_rate": 1.4799999999999998e-07,
      "loss": 0.0279,
      "reward": 0.30386340618133545,
      "reward_std": 0.08944490551948547,
      "reward_std_all": 0.09506761655211449,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9783475697040558,
      "rewards/forensics_tempsamp_iou_reward": 0.30386340618133545,
      "step": 1279
    },
    {
      "completion_length": 90.53125,
      "epoch": 3.4047936085219708,
      "grad_norm": 6.749434947967529,
      "kl": 0.7890625,
      "learning_rate": 1.4733333333333333e-07,
      "loss": 0.0316,
      "reward": 0.23519879579544067,
      "reward_std": 0.052560631185770035,
      "reward_std_all": 0.06661640480160713,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9375419616699219,
      "rewards/forensics_tempsamp_iou_reward": 0.23519879579544067,
      "step": 1280
    },
    {
      "completion_length": 92.59375,
      "epoch": 3.40745672436751,
      "grad_norm": 8.190831184387207,
      "kl": 0.9375,
      "learning_rate": 1.4666666666666666e-07,
      "loss": 0.0375,
      "reward": 0.17718026041984558,
      "reward_std": 0.09784922003746033,
      "reward_std_all": 0.10491282492876053,
      "rewards-max/forensics_tempsamp_iou_reward": 0.630501002073288,
      "rewards/forensics_tempsamp_iou_reward": 0.17718026041984558,
      "step": 1281
    },
    {
      "completion_length": 74.46875,
      "epoch": 3.4101198402130493,
      "grad_norm": 7.806985378265381,
      "kl": 0.4609375,
      "learning_rate": 1.4599999999999998e-07,
      "loss": 0.0184,
      "reward": 0.3606025129556656,
      "reward_std": 0.07299491763114929,
      "reward_std_all": 0.09268514439463615,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9551064670085907,
      "rewards/forensics_tempsamp_iou_reward": 0.3606025129556656,
      "step": 1282
    },
    {
      "completion_length": 67.3125,
      "epoch": 3.4127829560585887,
      "grad_norm": 9.525540351867676,
      "kl": 0.873046875,
      "learning_rate": 1.4533333333333334e-07,
      "loss": 0.035,
      "reward": 0.3034074753522873,
      "reward_std": 0.11942693218588829,
      "reward_std_all": 0.1188059151172638,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9475216865539551,
      "rewards/forensics_tempsamp_iou_reward": 0.3034074753522873,
      "step": 1283
    },
    {
      "completion_length": 66.96875,
      "epoch": 3.4154460719041277,
      "grad_norm": 5.934661388397217,
      "kl": 0.517578125,
      "learning_rate": 1.4466666666666667e-07,
      "loss": 0.0207,
      "reward": 0.26307062804698944,
      "reward_std": 0.05446258746087551,
      "reward_std_all": 0.07519498467445374,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8472906351089478,
      "rewards/forensics_tempsamp_iou_reward": 0.26307062804698944,
      "step": 1284
    },
    {
      "completion_length": 81.296875,
      "epoch": 3.418109187749667,
      "grad_norm": 5.434176921844482,
      "kl": 0.693359375,
      "learning_rate": 1.44e-07,
      "loss": 0.0277,
      "reward": 0.2936773747205734,
      "reward_std": 0.08763068541884422,
      "reward_std_all": 0.10558049008250237,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9964874982833862,
      "rewards/forensics_tempsamp_iou_reward": 0.2936773747205734,
      "step": 1285
    },
    {
      "completion_length": 74.546875,
      "epoch": 3.4207723035952062,
      "grad_norm": 8.480940818786621,
      "kl": 0.728515625,
      "learning_rate": 1.4333333333333335e-07,
      "loss": 0.0291,
      "reward": 0.21070323139429092,
      "reward_std": 0.08677364140748978,
      "reward_std_all": 0.09514261409640312,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6219838708639145,
      "rewards/forensics_tempsamp_iou_reward": 0.21070323139429092,
      "step": 1286
    },
    {
      "completion_length": 75.234375,
      "epoch": 3.4234354194407457,
      "grad_norm": 6.1159186363220215,
      "kl": 0.80078125,
      "learning_rate": 1.4266666666666665e-07,
      "loss": 0.0321,
      "reward": 0.22602930665016174,
      "reward_std": 0.10547506809234619,
      "reward_std_all": 0.1042790561914444,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6501006036996841,
      "rewards/forensics_tempsamp_iou_reward": 0.22602930665016174,
      "step": 1287
    },
    {
      "completion_length": 55.640625,
      "epoch": 3.426098535286285,
      "grad_norm": 6.284422397613525,
      "kl": 0.693359375,
      "learning_rate": 1.4199999999999997e-07,
      "loss": 0.0277,
      "reward": 0.24256376177072525,
      "reward_std": 0.08053722605109215,
      "reward_std_all": 0.0792975164949894,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9279479384422302,
      "rewards/forensics_tempsamp_iou_reward": 0.24256376177072525,
      "step": 1288
    },
    {
      "completion_length": 100.90625,
      "epoch": 3.428761651131824,
      "grad_norm": 9.088752746582031,
      "kl": 0.791015625,
      "learning_rate": 1.4133333333333333e-07,
      "loss": 0.0316,
      "reward": 0.25931060314178467,
      "reward_std": 0.06397867761552334,
      "reward_std_all": 0.061831435188651085,
      "rewards-max/forensics_tempsamp_iou_reward": 0.961246520280838,
      "rewards/forensics_tempsamp_iou_reward": 0.25931060314178467,
      "step": 1289
    },
    {
      "completion_length": 73.3125,
      "epoch": 3.4314247669773636,
      "grad_norm": 5.736721992492676,
      "kl": 0.734375,
      "learning_rate": 1.4066666666666666e-07,
      "loss": 0.0293,
      "reward": 0.3434506803750992,
      "reward_std": 0.09157699346542358,
      "reward_std_all": 0.11986258998513222,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9648519456386566,
      "rewards/forensics_tempsamp_iou_reward": 0.3434506803750992,
      "step": 1290
    },
    {
      "completion_length": 74.375,
      "epoch": 3.4340878828229027,
      "grad_norm": 41.84638977050781,
      "kl": 1.74609375,
      "learning_rate": 1.4e-07,
      "loss": 0.0696,
      "reward": 0.2541433051228523,
      "reward_std": 0.09533826261758804,
      "reward_std_all": 0.08930306881666183,
      "rewards-max/forensics_tempsamp_iou_reward": 0.984319418668747,
      "rewards/forensics_tempsamp_iou_reward": 0.2541433051228523,
      "step": 1291
    },
    {
      "completion_length": 61.28125,
      "epoch": 3.436750998668442,
      "grad_norm": 6.855373859405518,
      "kl": 0.6318359375,
      "learning_rate": 1.3933333333333334e-07,
      "loss": 0.0253,
      "reward": 0.3278149515390396,
      "reward_std": 0.11712972447276115,
      "reward_std_all": 0.1047104001045227,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9823279678821564,
      "rewards/forensics_tempsamp_iou_reward": 0.3278149515390396,
      "step": 1292
    },
    {
      "completion_length": 64.0625,
      "epoch": 3.4394141145139816,
      "grad_norm": 8.121284484863281,
      "kl": 0.66015625,
      "learning_rate": 1.3866666666666666e-07,
      "loss": 0.0264,
      "reward": 0.26894859969615936,
      "reward_std": 0.08152713999152184,
      "reward_std_all": 0.12482305616140366,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8374367952346802,
      "rewards/forensics_tempsamp_iou_reward": 0.26894859969615936,
      "step": 1293
    },
    {
      "completion_length": 69.75,
      "epoch": 3.4420772303595206,
      "grad_norm": 5.738999366760254,
      "kl": 0.646484375,
      "learning_rate": 1.3800000000000002e-07,
      "loss": 0.0259,
      "reward": 0.23587973415851593,
      "reward_std": 0.12244468182325363,
      "reward_std_all": 0.11902818828821182,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9933386147022247,
      "rewards/forensics_tempsamp_iou_reward": 0.23587973415851593,
      "step": 1294
    },
    {
      "completion_length": 32.921875,
      "epoch": 3.44474034620506,
      "grad_norm": 7.191456317901611,
      "kl": 0.5419921875,
      "learning_rate": 1.3733333333333332e-07,
      "loss": 0.0216,
      "reward": 0.29039886593818665,
      "reward_std": 0.07996411621570587,
      "reward_std_all": 0.08544912189245224,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7187698185443878,
      "rewards/forensics_tempsamp_iou_reward": 0.29039886593818665,
      "step": 1295
    },
    {
      "completion_length": 42.578125,
      "epoch": 3.447403462050599,
      "grad_norm": 8.965425491333008,
      "kl": 0.693359375,
      "learning_rate": 1.3666666666666665e-07,
      "loss": 0.0278,
      "reward": 0.3779078871011734,
      "reward_std": 0.09849446266889572,
      "reward_std_all": 0.1267850585281849,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9311352372169495,
      "rewards/forensics_tempsamp_iou_reward": 0.3779078871011734,
      "step": 1296
    },
    {
      "completion_length": 121.671875,
      "epoch": 3.4500665778961386,
      "grad_norm": 6.416714191436768,
      "kl": 0.66015625,
      "learning_rate": 1.36e-07,
      "loss": 0.0264,
      "reward": 0.26203762739896774,
      "reward_std": 0.11640831083059311,
      "reward_std_all": 0.14274045452475548,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9972972571849823,
      "rewards/forensics_tempsamp_iou_reward": 0.26203762739896774,
      "step": 1297
    },
    {
      "completion_length": 93.453125,
      "epoch": 3.4527296937416776,
      "grad_norm": 6.696905612945557,
      "kl": 0.814453125,
      "learning_rate": 1.3533333333333333e-07,
      "loss": 0.0326,
      "reward": 0.30437466502189636,
      "reward_std": 0.12767839059233665,
      "reward_std_all": 0.12750936299562454,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9136363565921783,
      "rewards/forensics_tempsamp_iou_reward": 0.30437466502189636,
      "step": 1298
    },
    {
      "completion_length": 94.984375,
      "epoch": 3.455392809587217,
      "grad_norm": 6.248603820800781,
      "kl": 0.80859375,
      "learning_rate": 1.3466666666666665e-07,
      "loss": 0.0323,
      "reward": 0.12311264127492905,
      "reward_std": 0.06606292352080345,
      "reward_std_all": 0.0682640727609396,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5486111044883728,
      "rewards/forensics_tempsamp_iou_reward": 0.12311264127492905,
      "step": 1299
    },
    {
      "completion_length": 90.59375,
      "epoch": 3.458055925432756,
      "grad_norm": 7.608530044555664,
      "kl": 0.6171875,
      "learning_rate": 1.34e-07,
      "loss": 0.0246,
      "reward": 0.3034742623567581,
      "reward_std": 0.08599521219730377,
      "reward_std_all": 0.10456831008195877,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9972730278968811,
      "rewards/forensics_tempsamp_iou_reward": 0.3034742623567581,
      "step": 1300
    },
    {
      "completion_length": 109.375,
      "epoch": 3.4607190412782955,
      "grad_norm": 8.606790542602539,
      "kl": 1.173828125,
      "learning_rate": 1.3333333333333334e-07,
      "loss": 0.0469,
      "reward": 0.17321263998746872,
      "reward_std": 0.10574930906295776,
      "reward_std_all": 0.1103423498570919,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5973684191703796,
      "rewards/forensics_tempsamp_iou_reward": 0.17321263998746872,
      "step": 1301
    },
    {
      "completion_length": 60.96875,
      "epoch": 3.463382157123835,
      "grad_norm": 6.764240741729736,
      "kl": 0.841796875,
      "learning_rate": 1.3266666666666664e-07,
      "loss": 0.0336,
      "reward": 0.2541128844022751,
      "reward_std": 0.09276970475912094,
      "reward_std_all": 0.08782188594341278,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9905638992786407,
      "rewards/forensics_tempsamp_iou_reward": 0.2541128844022751,
      "step": 1302
    },
    {
      "completion_length": 80.0,
      "epoch": 3.466045272969374,
      "grad_norm": 10.496405601501465,
      "kl": 1.0380859375,
      "learning_rate": 1.32e-07,
      "loss": 0.0415,
      "reward": 0.25380705296993256,
      "reward_std": 0.0644807368516922,
      "reward_std_all": 0.06275189481675625,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7097296416759491,
      "rewards/forensics_tempsamp_iou_reward": 0.25380705296993256,
      "step": 1303
    },
    {
      "completion_length": 96.953125,
      "epoch": 3.4687083888149135,
      "grad_norm": 8.165837287902832,
      "kl": 0.888671875,
      "learning_rate": 1.3133333333333332e-07,
      "loss": 0.0356,
      "reward": 0.23936061561107635,
      "reward_std": 0.08127810806035995,
      "reward_std_all": 0.09299192205071449,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8066666722297668,
      "rewards/forensics_tempsamp_iou_reward": 0.23936061561107635,
      "step": 1304
    },
    {
      "completion_length": 80.796875,
      "epoch": 3.4713715046604525,
      "grad_norm": 9.403022766113281,
      "kl": 0.912109375,
      "learning_rate": 1.3066666666666665e-07,
      "loss": 0.0365,
      "reward": 0.35338185727596283,
      "reward_std": 0.12676146626472473,
      "reward_std_all": 0.11811638250946999,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7681397497653961,
      "rewards/forensics_tempsamp_iou_reward": 0.35338185727596283,
      "step": 1305
    },
    {
      "completion_length": 42.609375,
      "epoch": 3.474034620505992,
      "grad_norm": 16.234872817993164,
      "kl": 1.16796875,
      "learning_rate": 1.3e-07,
      "loss": 0.0468,
      "reward": 0.2744322419166565,
      "reward_std": 0.09289354085922241,
      "reward_std_all": 0.09326902776956558,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9793396890163422,
      "rewards/forensics_tempsamp_iou_reward": 0.2744322419166565,
      "step": 1306
    },
    {
      "completion_length": 63.53125,
      "epoch": 3.4766977363515315,
      "grad_norm": 6.317926406860352,
      "kl": 0.654296875,
      "learning_rate": 1.2933333333333333e-07,
      "loss": 0.0261,
      "reward": 0.2870938628911972,
      "reward_std": 0.1081898994743824,
      "reward_std_all": 0.1087254136800766,
      "rewards-max/forensics_tempsamp_iou_reward": 0.931461364030838,
      "rewards/forensics_tempsamp_iou_reward": 0.2870938628911972,
      "step": 1307
    },
    {
      "completion_length": 94.328125,
      "epoch": 3.4793608521970705,
      "grad_norm": 7.352465629577637,
      "kl": 0.791015625,
      "learning_rate": 1.2866666666666668e-07,
      "loss": 0.0316,
      "reward": 0.1508246660232544,
      "reward_std": 0.09612922742962837,
      "reward_std_all": 0.09714759141206741,
      "rewards-max/forensics_tempsamp_iou_reward": 0.95222008228302,
      "rewards/forensics_tempsamp_iou_reward": 0.1508246660232544,
      "step": 1308
    },
    {
      "completion_length": 45.46875,
      "epoch": 3.48202396804261,
      "grad_norm": 8.180109977722168,
      "kl": 0.6806640625,
      "learning_rate": 1.28e-07,
      "loss": 0.0273,
      "reward": 0.4078506976366043,
      "reward_std": 0.11220230907201767,
      "reward_std_all": 0.10047727450728416,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8566317856311798,
      "rewards/forensics_tempsamp_iou_reward": 0.4078506976366043,
      "step": 1309
    },
    {
      "completion_length": 135.75,
      "epoch": 3.484687083888149,
      "grad_norm": 8.242186546325684,
      "kl": 0.9140625,
      "learning_rate": 1.273333333333333e-07,
      "loss": 0.0365,
      "reward": 0.2938735783100128,
      "reward_std": 0.13073544204235077,
      "reward_std_all": 0.13779879733920097,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7786195278167725,
      "rewards/forensics_tempsamp_iou_reward": 0.2938735783100128,
      "step": 1310
    },
    {
      "completion_length": 83.0,
      "epoch": 3.4873501997336884,
      "grad_norm": 7.343684673309326,
      "kl": 0.630859375,
      "learning_rate": 1.2666666666666666e-07,
      "loss": 0.0252,
      "reward": 0.22847063094377518,
      "reward_std": 0.10123861581087112,
      "reward_std_all": 0.09698142856359482,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7976280152797699,
      "rewards/forensics_tempsamp_iou_reward": 0.22847063094377518,
      "step": 1311
    },
    {
      "completion_length": 103.65625,
      "epoch": 3.490013315579228,
      "grad_norm": 8.77862548828125,
      "kl": 1.1328125,
      "learning_rate": 1.26e-07,
      "loss": 0.0452,
      "reward": 0.29157277941703796,
      "reward_std": 0.09117333963513374,
      "reward_std_all": 0.1052551195025444,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8063423037528992,
      "rewards/forensics_tempsamp_iou_reward": 0.29157277941703796,
      "step": 1312
    },
    {
      "completion_length": 78.6875,
      "epoch": 3.492676431424767,
      "grad_norm": 7.873570919036865,
      "kl": 1.01953125,
      "learning_rate": 1.2533333333333332e-07,
      "loss": 0.0408,
      "reward": 0.3431662321090698,
      "reward_std": 0.10921333730220795,
      "reward_std_all": 0.12765420973300934,
      "rewards-max/forensics_tempsamp_iou_reward": 0.995866060256958,
      "rewards/forensics_tempsamp_iou_reward": 0.3431662321090698,
      "step": 1313
    },
    {
      "completion_length": 44.0625,
      "epoch": 3.4953395472703064,
      "grad_norm": 9.229249954223633,
      "kl": 0.90234375,
      "learning_rate": 1.2466666666666664e-07,
      "loss": 0.0362,
      "reward": 0.33946891129016876,
      "reward_std": 0.13078325241804123,
      "reward_std_all": 0.12489969655871391,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9790891706943512,
      "rewards/forensics_tempsamp_iou_reward": 0.33946891129016876,
      "step": 1314
    },
    {
      "completion_length": 111.140625,
      "epoch": 3.4980026631158454,
      "grad_norm": 7.815958499908447,
      "kl": 0.697265625,
      "learning_rate": 1.24e-07,
      "loss": 0.028,
      "reward": 0.19801585376262665,
      "reward_std": 0.12866169214248657,
      "reward_std_all": 0.12366434186697006,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8154460191726685,
      "rewards/forensics_tempsamp_iou_reward": 0.19801585376262665,
      "step": 1315
    },
    {
      "completion_length": 43.859375,
      "epoch": 3.500665778961385,
      "grad_norm": 6.725625991821289,
      "kl": 0.673828125,
      "learning_rate": 1.2333333333333333e-07,
      "loss": 0.0269,
      "reward": 0.31721608340740204,
      "reward_std": 0.07831311225891113,
      "reward_std_all": 0.091094721108675,
      "rewards-max/forensics_tempsamp_iou_reward": 0.983840823173523,
      "rewards/forensics_tempsamp_iou_reward": 0.31721608340740204,
      "step": 1316
    },
    {
      "completion_length": 100.328125,
      "epoch": 3.5033288948069243,
      "grad_norm": 7.697299957275391,
      "kl": 0.681640625,
      "learning_rate": 1.2266666666666665e-07,
      "loss": 0.0273,
      "reward": 0.27777285873889923,
      "reward_std": 0.10094791650772095,
      "reward_std_all": 0.11741773411631584,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7046216428279877,
      "rewards/forensics_tempsamp_iou_reward": 0.27777285873889923,
      "step": 1317
    },
    {
      "completion_length": 66.71875,
      "epoch": 3.5059920106524634,
      "grad_norm": 7.964829444885254,
      "kl": 0.869140625,
      "learning_rate": 1.2199999999999998e-07,
      "loss": 0.0348,
      "reward": 0.23045238107442856,
      "reward_std": 0.09069156087934971,
      "reward_std_all": 0.09498932026326656,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9439746141433716,
      "rewards/forensics_tempsamp_iou_reward": 0.23045238107442856,
      "step": 1318
    },
    {
      "completion_length": 45.953125,
      "epoch": 3.508655126498003,
      "grad_norm": 6.986118793487549,
      "kl": 0.6435546875,
      "learning_rate": 1.2133333333333333e-07,
      "loss": 0.0257,
      "reward": 0.17625071853399277,
      "reward_std": 0.06991540640592575,
      "reward_std_all": 0.06882248446345329,
      "rewards-max/forensics_tempsamp_iou_reward": 0.41928862035274506,
      "rewards/forensics_tempsamp_iou_reward": 0.17625071853399277,
      "step": 1319
    },
    {
      "completion_length": 120.4375,
      "epoch": 3.511318242343542,
      "grad_norm": 6.3230390548706055,
      "kl": 0.677734375,
      "learning_rate": 1.2066666666666666e-07,
      "loss": 0.0271,
      "reward": 0.27595867216587067,
      "reward_std": 0.14015264809131622,
      "reward_std_all": 0.13077142089605331,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8546566367149353,
      "rewards/forensics_tempsamp_iou_reward": 0.27595867216587067,
      "step": 1320
    },
    {
      "completion_length": 42.765625,
      "epoch": 3.5139813581890813,
      "grad_norm": 6.361483573913574,
      "kl": 0.6953125,
      "learning_rate": 1.2e-07,
      "loss": 0.0278,
      "reward": 0.31092285364866257,
      "reward_std": 0.12524167448282242,
      "reward_std_all": 0.12898309156298637,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9084562361240387,
      "rewards/forensics_tempsamp_iou_reward": 0.31092285364866257,
      "step": 1321
    },
    {
      "completion_length": 54.609375,
      "epoch": 3.5166444740346208,
      "grad_norm": 8.642081260681152,
      "kl": 0.78515625,
      "learning_rate": 1.1933333333333332e-07,
      "loss": 0.0314,
      "reward": 0.22017865628004074,
      "reward_std": 0.08093864098191261,
      "reward_std_all": 0.09610449895262718,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8152908086776733,
      "rewards/forensics_tempsamp_iou_reward": 0.22017865628004074,
      "step": 1322
    },
    {
      "completion_length": 82.1875,
      "epoch": 3.51930758988016,
      "grad_norm": 8.582603454589844,
      "kl": 0.7314453125,
      "learning_rate": 1.1866666666666667e-07,
      "loss": 0.0292,
      "reward": 0.21615515649318695,
      "reward_std": 0.08153605088591576,
      "reward_std_all": 0.08217078447341919,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9291174411773682,
      "rewards/forensics_tempsamp_iou_reward": 0.21615515649318695,
      "step": 1323
    },
    {
      "completion_length": 39.78125,
      "epoch": 3.521970705725699,
      "grad_norm": 8.219220161437988,
      "kl": 1.037109375,
      "learning_rate": 1.1799999999999998e-07,
      "loss": 0.0415,
      "reward": 0.2097192481160164,
      "reward_std": 0.11861815303564072,
      "reward_std_all": 0.12254097312688828,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6357142776250839,
      "rewards/forensics_tempsamp_iou_reward": 0.2097192481160164,
      "step": 1324
    },
    {
      "completion_length": 74.15625,
      "epoch": 3.5246338215712383,
      "grad_norm": 7.793252944946289,
      "kl": 0.517578125,
      "learning_rate": 1.1733333333333333e-07,
      "loss": 0.0207,
      "reward": 0.3198886066675186,
      "reward_std": 0.07082006521522999,
      "reward_std_all": 0.09938723593950272,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8480519652366638,
      "rewards/forensics_tempsamp_iou_reward": 0.3198886066675186,
      "step": 1325
    },
    {
      "completion_length": 71.5625,
      "epoch": 3.5272969374167777,
      "grad_norm": 7.948963642120361,
      "kl": 0.708984375,
      "learning_rate": 1.1666666666666667e-07,
      "loss": 0.0284,
      "reward": 0.16241128742694855,
      "reward_std": 0.09436837956309319,
      "reward_std_all": 0.09744227677583694,
      "rewards-max/forensics_tempsamp_iou_reward": 0.47323794662952423,
      "rewards/forensics_tempsamp_iou_reward": 0.16241128742694855,
      "step": 1326
    },
    {
      "completion_length": 65.109375,
      "epoch": 3.5299600532623168,
      "grad_norm": 8.651603698730469,
      "kl": 0.818359375,
      "learning_rate": 1.16e-07,
      "loss": 0.0328,
      "reward": 0.2726309224963188,
      "reward_std": 0.07814621925354004,
      "reward_std_all": 0.08615278080105782,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7355967164039612,
      "rewards/forensics_tempsamp_iou_reward": 0.2726309224963188,
      "step": 1327
    },
    {
      "completion_length": 91.84375,
      "epoch": 3.5326231691078562,
      "grad_norm": 8.168451309204102,
      "kl": 0.921875,
      "learning_rate": 1.1533333333333332e-07,
      "loss": 0.0369,
      "reward": 0.24039973318576813,
      "reward_std": 0.0633737025782466,
      "reward_std_all": 0.07939804345369339,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8954554200172424,
      "rewards/forensics_tempsamp_iou_reward": 0.24039973318576813,
      "step": 1328
    },
    {
      "completion_length": 48.890625,
      "epoch": 3.5352862849533953,
      "grad_norm": 6.707285404205322,
      "kl": 0.4765625,
      "learning_rate": 1.1466666666666666e-07,
      "loss": 0.0191,
      "reward": 0.23652960360050201,
      "reward_std": 0.07958490774035454,
      "reward_std_all": 0.07692503556609154,
      "rewards-max/forensics_tempsamp_iou_reward": 0.869143545627594,
      "rewards/forensics_tempsamp_iou_reward": 0.23652960360050201,
      "step": 1329
    },
    {
      "completion_length": 52.328125,
      "epoch": 3.5379494007989347,
      "grad_norm": 7.730066776275635,
      "kl": 0.5517578125,
      "learning_rate": 1.14e-07,
      "loss": 0.0221,
      "reward": 0.4778399169445038,
      "reward_std": 0.11519316956400871,
      "reward_std_all": 0.12424469739198685,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9922856986522675,
      "rewards/forensics_tempsamp_iou_reward": 0.4778399169445038,
      "step": 1330
    },
    {
      "completion_length": 40.453125,
      "epoch": 3.540612516644474,
      "grad_norm": 7.181894779205322,
      "kl": 0.765625,
      "learning_rate": 1.1333333333333332e-07,
      "loss": 0.0306,
      "reward": 0.3320945054292679,
      "reward_std": 0.1320127621293068,
      "reward_std_all": 0.13548093289136887,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8748931586742401,
      "rewards/forensics_tempsamp_iou_reward": 0.3320945054292679,
      "step": 1331
    },
    {
      "completion_length": 66.75,
      "epoch": 3.543275632490013,
      "grad_norm": 9.213545799255371,
      "kl": 0.619140625,
      "learning_rate": 1.1266666666666666e-07,
      "loss": 0.0248,
      "reward": 0.29249654710292816,
      "reward_std": 0.1003364771604538,
      "reward_std_all": 0.09865913167595863,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8522191941738129,
      "rewards/forensics_tempsamp_iou_reward": 0.29249654710292816,
      "step": 1332
    },
    {
      "completion_length": 128.015625,
      "epoch": 3.5459387483355527,
      "grad_norm": 7.1827826499938965,
      "kl": 0.849609375,
      "learning_rate": 1.12e-07,
      "loss": 0.0339,
      "reward": 0.16642824932932854,
      "reward_std": 0.05728480406105518,
      "reward_std_all": 0.08047118782997131,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7272970229387283,
      "rewards/forensics_tempsamp_iou_reward": 0.16642824932932854,
      "step": 1333
    },
    {
      "completion_length": 73.984375,
      "epoch": 3.5486018641810917,
      "grad_norm": 8.479418754577637,
      "kl": 0.724609375,
      "learning_rate": 1.1133333333333334e-07,
      "loss": 0.029,
      "reward": 0.13817555829882622,
      "reward_std": 0.08651726506650448,
      "reward_std_all": 0.08492754772305489,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5369491726160049,
      "rewards/forensics_tempsamp_iou_reward": 0.13817555829882622,
      "step": 1334
    },
    {
      "completion_length": 87.0,
      "epoch": 3.551264980026631,
      "grad_norm": 6.413187026977539,
      "kl": 0.6416015625,
      "learning_rate": 1.1066666666666665e-07,
      "loss": 0.0257,
      "reward": 0.20181342214345932,
      "reward_std": 0.11340351402759552,
      "reward_std_all": 0.11461244896054268,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6347396522760391,
      "rewards/forensics_tempsamp_iou_reward": 0.20181342214345932,
      "step": 1335
    },
    {
      "completion_length": 90.671875,
      "epoch": 3.5539280958721706,
      "grad_norm": 8.414183616638184,
      "kl": 0.6484375,
      "learning_rate": 1.0999999999999999e-07,
      "loss": 0.026,
      "reward": 0.18121296912431717,
      "reward_std": 0.08188913390040398,
      "reward_std_all": 0.08721772208809853,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5149400383234024,
      "rewards/forensics_tempsamp_iou_reward": 0.18121296912431717,
      "step": 1336
    },
    {
      "completion_length": 62.4375,
      "epoch": 3.5565912117177096,
      "grad_norm": 8.344260215759277,
      "kl": 0.515625,
      "learning_rate": 1.0933333333333333e-07,
      "loss": 0.0206,
      "reward": 0.37059859931468964,
      "reward_std": 0.10017327964305878,
      "reward_std_all": 0.12126250565052032,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9027340114116669,
      "rewards/forensics_tempsamp_iou_reward": 0.37059859931468964,
      "step": 1337
    },
    {
      "completion_length": 116.328125,
      "epoch": 3.559254327563249,
      "grad_norm": 5.14065408706665,
      "kl": 0.619140625,
      "learning_rate": 1.0866666666666666e-07,
      "loss": 0.0248,
      "reward": 0.19098110124468803,
      "reward_std": 0.08694043383002281,
      "reward_std_all": 0.10456594452261925,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8635088801383972,
      "rewards/forensics_tempsamp_iou_reward": 0.19098110124468803,
      "step": 1338
    },
    {
      "completion_length": 49.375,
      "epoch": 3.561917443408788,
      "grad_norm": 6.2541608810424805,
      "kl": 0.4619140625,
      "learning_rate": 1.0799999999999999e-07,
      "loss": 0.0184,
      "reward": 0.30141736567020416,
      "reward_std": 0.11265099048614502,
      "reward_std_all": 0.1295337751507759,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7418025135993958,
      "rewards/forensics_tempsamp_iou_reward": 0.30141736567020416,
      "step": 1339
    },
    {
      "completion_length": 77.265625,
      "epoch": 3.5645805592543276,
      "grad_norm": 7.370177268981934,
      "kl": 0.52734375,
      "learning_rate": 1.0733333333333333e-07,
      "loss": 0.0211,
      "reward": 0.20185451209545135,
      "reward_std": 0.06951400637626648,
      "reward_std_all": 0.07369459420442581,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9777361154556274,
      "rewards/forensics_tempsamp_iou_reward": 0.20185451209545135,
      "step": 1340
    },
    {
      "completion_length": 41.796875,
      "epoch": 3.567243675099867,
      "grad_norm": 7.9556803703308105,
      "kl": 0.51953125,
      "learning_rate": 1.0666666666666667e-07,
      "loss": 0.0208,
      "reward": 0.32406194508075714,
      "reward_std": 0.08810224384069443,
      "reward_std_all": 0.08224286884069443,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9540002346038818,
      "rewards/forensics_tempsamp_iou_reward": 0.32406194508075714,
      "step": 1341
    },
    {
      "completion_length": 74.6875,
      "epoch": 3.569906790945406,
      "grad_norm": 6.431295871734619,
      "kl": 0.513671875,
      "learning_rate": 1.06e-07,
      "loss": 0.0205,
      "reward": 0.2903384119272232,
      "reward_std": 0.12275248393416405,
      "reward_std_all": 0.10889869183301926,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9874204993247986,
      "rewards/forensics_tempsamp_iou_reward": 0.2903384119272232,
      "step": 1342
    },
    {
      "completion_length": 50.515625,
      "epoch": 3.5725699067909455,
      "grad_norm": 7.608808517456055,
      "kl": 0.640625,
      "learning_rate": 1.0533333333333332e-07,
      "loss": 0.0256,
      "reward": 0.3255097270011902,
      "reward_std": 0.10231473669409752,
      "reward_std_all": 0.11519955843687057,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9780152142047882,
      "rewards/forensics_tempsamp_iou_reward": 0.3255097270011902,
      "step": 1343
    },
    {
      "completion_length": 94.90625,
      "epoch": 3.5752330226364846,
      "grad_norm": 8.01926326751709,
      "kl": 0.734375,
      "learning_rate": 1.0466666666666666e-07,
      "loss": 0.0293,
      "reward": 0.20645450055599213,
      "reward_std": 0.10992483049631119,
      "reward_std_all": 0.11816775053739548,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7749342024326324,
      "rewards/forensics_tempsamp_iou_reward": 0.20645450055599213,
      "step": 1344
    },
    {
      "completion_length": 66.375,
      "epoch": 3.577896138482024,
      "grad_norm": 36.16609191894531,
      "kl": 1.4140625,
      "learning_rate": 1.0399999999999999e-07,
      "loss": 0.0568,
      "reward": 0.275847926735878,
      "reward_std": 0.07558550871908665,
      "reward_std_all": 0.07640020176768303,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8150611221790314,
      "rewards/forensics_tempsamp_iou_reward": 0.275847926735878,
      "step": 1345
    },
    {
      "completion_length": 83.5,
      "epoch": 3.5805592543275635,
      "grad_norm": 7.41264533996582,
      "kl": 0.802734375,
      "learning_rate": 1.0333333333333333e-07,
      "loss": 0.0322,
      "reward": 0.25569935888051987,
      "reward_std": 0.1008378341794014,
      "reward_std_all": 0.10675804316997528,
      "rewards-max/forensics_tempsamp_iou_reward": 0.96161288022995,
      "rewards/forensics_tempsamp_iou_reward": 0.25569935888051987,
      "step": 1346
    },
    {
      "completion_length": 52.53125,
      "epoch": 3.5832223701731025,
      "grad_norm": 9.225470542907715,
      "kl": 0.59375,
      "learning_rate": 1.0266666666666666e-07,
      "loss": 0.0238,
      "reward": 0.24257957190275192,
      "reward_std": 0.10124441236257553,
      "reward_std_all": 0.10275403037667274,
      "rewards-max/forensics_tempsamp_iou_reward": 0.809010922908783,
      "rewards/forensics_tempsamp_iou_reward": 0.24257957190275192,
      "step": 1347
    },
    {
      "completion_length": 81.46875,
      "epoch": 3.5858854860186415,
      "grad_norm": 6.865204811096191,
      "kl": 0.873046875,
      "learning_rate": 1.0199999999999999e-07,
      "loss": 0.035,
      "reward": 0.1900021657347679,
      "reward_std": 0.05099297687411308,
      "reward_std_all": 0.06537709012627602,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7106659710407257,
      "rewards/forensics_tempsamp_iou_reward": 0.1900021657347679,
      "step": 1348
    },
    {
      "completion_length": 73.59375,
      "epoch": 3.588548601864181,
      "grad_norm": 7.496218204498291,
      "kl": 0.685546875,
      "learning_rate": 1.0133333333333333e-07,
      "loss": 0.0274,
      "reward": 0.34986722469329834,
      "reward_std": 0.10361068695783615,
      "reward_std_all": 0.14143969491124153,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9909382462501526,
      "rewards/forensics_tempsamp_iou_reward": 0.34986722469329834,
      "step": 1349
    },
    {
      "completion_length": 50.03125,
      "epoch": 3.5912117177097205,
      "grad_norm": 8.118083953857422,
      "kl": 0.64453125,
      "learning_rate": 1.0066666666666667e-07,
      "loss": 0.0258,
      "reward": 0.33321626484394073,
      "reward_std": 0.13374566659331322,
      "reward_std_all": 0.1569751240313053,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8521952033042908,
      "rewards/forensics_tempsamp_iou_reward": 0.33321626484394073,
      "step": 1350
    },
    {
      "completion_length": 103.59375,
      "epoch": 3.5938748335552595,
      "grad_norm": 8.97614860534668,
      "kl": 0.787109375,
      "learning_rate": 1e-07,
      "loss": 0.0315,
      "reward": 0.21855439245700836,
      "reward_std": 0.09741482138633728,
      "reward_std_all": 0.11450507864356041,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6033613532781601,
      "rewards/forensics_tempsamp_iou_reward": 0.21855439245700836,
      "step": 1351
    },
    {
      "completion_length": 68.546875,
      "epoch": 3.596537949400799,
      "grad_norm": 6.177867889404297,
      "kl": 0.5703125,
      "learning_rate": 9.933333333333332e-08,
      "loss": 0.0228,
      "reward": 0.2032327950000763,
      "reward_std": 0.05430947616696358,
      "reward_std_all": 0.05809543468058109,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5905359983444214,
      "rewards/forensics_tempsamp_iou_reward": 0.2032327950000763,
      "step": 1352
    },
    {
      "completion_length": 55.84375,
      "epoch": 3.599201065246338,
      "grad_norm": 7.283995151519775,
      "kl": 0.59375,
      "learning_rate": 9.866666666666666e-08,
      "loss": 0.0238,
      "reward": 0.30600886046886444,
      "reward_std": 0.08866710960865021,
      "reward_std_all": 0.09938887320458889,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9925792813301086,
      "rewards/forensics_tempsamp_iou_reward": 0.30600886046886444,
      "step": 1353
    },
    {
      "completion_length": 66.90625,
      "epoch": 3.6018641810918774,
      "grad_norm": 21.502363204956055,
      "kl": 0.8671875,
      "learning_rate": 9.8e-08,
      "loss": 0.0346,
      "reward": 0.19855272769927979,
      "reward_std": 0.07513890787959099,
      "reward_std_all": 0.1013302281498909,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5897786021232605,
      "rewards/forensics_tempsamp_iou_reward": 0.19855272769927979,
      "step": 1354
    },
    {
      "completion_length": 85.5625,
      "epoch": 3.604527296937417,
      "grad_norm": 6.20465612411499,
      "kl": 0.822265625,
      "learning_rate": 9.733333333333332e-08,
      "loss": 0.0329,
      "reward": 0.24279901385307312,
      "reward_std": 0.07164918258786201,
      "reward_std_all": 0.08822982385754585,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7506774961948395,
      "rewards/forensics_tempsamp_iou_reward": 0.24279901385307312,
      "step": 1355
    },
    {
      "completion_length": 62.71875,
      "epoch": 3.607190412782956,
      "grad_norm": 9.210076332092285,
      "kl": 0.703125,
      "learning_rate": 9.666666666666666e-08,
      "loss": 0.0281,
      "reward": 0.2741064727306366,
      "reward_std": 0.09467235952615738,
      "reward_std_all": 0.11776531860232353,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9017927348613739,
      "rewards/forensics_tempsamp_iou_reward": 0.2741064727306366,
      "step": 1356
    },
    {
      "completion_length": 83.640625,
      "epoch": 3.6098535286284954,
      "grad_norm": 6.326016426086426,
      "kl": 0.607421875,
      "learning_rate": 9.6e-08,
      "loss": 0.0243,
      "reward": 0.27604835480451584,
      "reward_std": 0.10150536894798279,
      "reward_std_all": 0.09774038940668106,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8271313309669495,
      "rewards/forensics_tempsamp_iou_reward": 0.27604835480451584,
      "step": 1357
    },
    {
      "completion_length": 75.5,
      "epoch": 3.6125166444740344,
      "grad_norm": 5.703260898590088,
      "kl": 0.83203125,
      "learning_rate": 9.533333333333334e-08,
      "loss": 0.0332,
      "reward": 0.2536266818642616,
      "reward_std": 0.058805596083402634,
      "reward_std_all": 0.07577456161379814,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7452064752578735,
      "rewards/forensics_tempsamp_iou_reward": 0.2536266818642616,
      "step": 1358
    },
    {
      "completion_length": 58.484375,
      "epoch": 3.615179760319574,
      "grad_norm": 9.469295501708984,
      "kl": 0.9765625,
      "learning_rate": 9.466666666666665e-08,
      "loss": 0.039,
      "reward": 0.42074304819107056,
      "reward_std": 0.10632530599832535,
      "reward_std_all": 0.10998575016856194,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9823385775089264,
      "rewards/forensics_tempsamp_iou_reward": 0.42074304819107056,
      "step": 1359
    },
    {
      "completion_length": 79.578125,
      "epoch": 3.6178428761651134,
      "grad_norm": 8.439776420593262,
      "kl": 0.71484375,
      "learning_rate": 9.4e-08,
      "loss": 0.0286,
      "reward": 0.2753216624259949,
      "reward_std": 0.0797462910413742,
      "reward_std_all": 0.09877178817987442,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7252941131591797,
      "rewards/forensics_tempsamp_iou_reward": 0.2753216624259949,
      "step": 1360
    },
    {
      "completion_length": 72.484375,
      "epoch": 3.6205059920106524,
      "grad_norm": 6.644494533538818,
      "kl": 0.5224609375,
      "learning_rate": 9.333333333333334e-08,
      "loss": 0.021,
      "reward": 0.2860092744231224,
      "reward_std": 0.12832185253500938,
      "reward_std_all": 0.15343088656663895,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9932495951652527,
      "rewards/forensics_tempsamp_iou_reward": 0.2860092744231224,
      "step": 1361
    },
    {
      "completion_length": 148.25,
      "epoch": 3.623169107856192,
      "grad_norm": 6.340518951416016,
      "kl": 0.669921875,
      "learning_rate": 9.266666666666666e-08,
      "loss": 0.0267,
      "reward": 0.2083103358745575,
      "reward_std": 0.10022951662540436,
      "reward_std_all": 0.10446858778595924,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6547657996416092,
      "rewards/forensics_tempsamp_iou_reward": 0.2083103358745575,
      "step": 1362
    },
    {
      "completion_length": 93.265625,
      "epoch": 3.625832223701731,
      "grad_norm": 5.83003568649292,
      "kl": 0.724609375,
      "learning_rate": 9.199999999999999e-08,
      "loss": 0.029,
      "reward": 0.2370523065328598,
      "reward_std": 0.11399869620800018,
      "reward_std_all": 0.1084674671292305,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9727766215801239,
      "rewards/forensics_tempsamp_iou_reward": 0.2370523065328598,
      "step": 1363
    },
    {
      "completion_length": 83.65625,
      "epoch": 3.6284953395472703,
      "grad_norm": 9.71653938293457,
      "kl": 0.779296875,
      "learning_rate": 9.133333333333333e-08,
      "loss": 0.0312,
      "reward": 0.23170115798711777,
      "reward_std": 0.1287159100174904,
      "reward_std_all": 0.1287151835858822,
      "rewards-max/forensics_tempsamp_iou_reward": 0.789962112903595,
      "rewards/forensics_tempsamp_iou_reward": 0.23170115798711777,
      "step": 1364
    },
    {
      "completion_length": 62.671875,
      "epoch": 3.63115845539281,
      "grad_norm": 8.822818756103516,
      "kl": 0.798828125,
      "learning_rate": 9.066666666666667e-08,
      "loss": 0.032,
      "reward": 0.2485719844698906,
      "reward_std": 0.10809770599007607,
      "reward_std_all": 0.10338503867387772,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7701721489429474,
      "rewards/forensics_tempsamp_iou_reward": 0.2485719844698906,
      "step": 1365
    },
    {
      "completion_length": 98.75,
      "epoch": 3.633821571238349,
      "grad_norm": 10.512511253356934,
      "kl": 1.42578125,
      "learning_rate": 9e-08,
      "loss": 0.057,
      "reward": 0.2268773391842842,
      "reward_std": 0.08409934863448143,
      "reward_std_all": 0.08794623240828514,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8180743157863617,
      "rewards/forensics_tempsamp_iou_reward": 0.2268773391842842,
      "step": 1366
    },
    {
      "completion_length": 55.953125,
      "epoch": 3.6364846870838883,
      "grad_norm": 6.453632831573486,
      "kl": 0.654296875,
      "learning_rate": 8.933333333333333e-08,
      "loss": 0.0261,
      "reward": 0.3130349665880203,
      "reward_std": 0.0677583497017622,
      "reward_std_all": 0.08450796827673912,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9429309070110321,
      "rewards/forensics_tempsamp_iou_reward": 0.3130349665880203,
      "step": 1367
    },
    {
      "completion_length": 59.96875,
      "epoch": 3.6391478029294273,
      "grad_norm": 9.683226585388184,
      "kl": 0.7109375,
      "learning_rate": 8.866666666666667e-08,
      "loss": 0.0284,
      "reward": 0.20657147467136383,
      "reward_std": 0.0817405916750431,
      "reward_std_all": 0.08500572293996811,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6711632460355759,
      "rewards/forensics_tempsamp_iou_reward": 0.20657147467136383,
      "step": 1368
    },
    {
      "completion_length": 98.375,
      "epoch": 3.6418109187749668,
      "grad_norm": 5.171467304229736,
      "kl": 0.6318359375,
      "learning_rate": 8.8e-08,
      "loss": 0.0253,
      "reward": 0.19301363825798035,
      "reward_std": 0.09956397861242294,
      "reward_std_all": 0.10743656009435654,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6514797508716583,
      "rewards/forensics_tempsamp_iou_reward": 0.19301363825798035,
      "step": 1369
    },
    {
      "completion_length": 72.078125,
      "epoch": 3.6444740346205062,
      "grad_norm": 6.555440902709961,
      "kl": 0.611328125,
      "learning_rate": 8.733333333333333e-08,
      "loss": 0.0245,
      "reward": 0.22298169136047363,
      "reward_std": 0.10678927972912788,
      "reward_std_all": 0.10357094928622246,
      "rewards-max/forensics_tempsamp_iou_reward": 0.710187166929245,
      "rewards/forensics_tempsamp_iou_reward": 0.22298169136047363,
      "step": 1370
    },
    {
      "completion_length": 66.75,
      "epoch": 3.6471371504660453,
      "grad_norm": 6.657570838928223,
      "kl": 0.623046875,
      "learning_rate": 8.666666666666666e-08,
      "loss": 0.0249,
      "reward": 0.26928292214870453,
      "reward_std": 0.13439291343092918,
      "reward_std_all": 0.1295541748404503,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7358768284320831,
      "rewards/forensics_tempsamp_iou_reward": 0.26928292214870453,
      "step": 1371
    },
    {
      "completion_length": 78.765625,
      "epoch": 3.6498002663115847,
      "grad_norm": 7.2371110916137695,
      "kl": 0.779296875,
      "learning_rate": 8.599999999999999e-08,
      "loss": 0.0312,
      "reward": 0.25418074429035187,
      "reward_std": 0.11339772492647171,
      "reward_std_all": 0.11249963939189911,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7532258331775665,
      "rewards/forensics_tempsamp_iou_reward": 0.25418074429035187,
      "step": 1372
    },
    {
      "completion_length": 64.890625,
      "epoch": 3.6524633821571237,
      "grad_norm": 7.51252555847168,
      "kl": 0.69140625,
      "learning_rate": 8.533333333333333e-08,
      "loss": 0.0276,
      "reward": 0.2040424793958664,
      "reward_std": 0.09273020550608635,
      "reward_std_all": 0.09137862175703049,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7069119811058044,
      "rewards/forensics_tempsamp_iou_reward": 0.2040424793958664,
      "step": 1373
    },
    {
      "completion_length": 59.390625,
      "epoch": 3.655126498002663,
      "grad_norm": 7.901826858520508,
      "kl": 0.708984375,
      "learning_rate": 8.466666666666667e-08,
      "loss": 0.0283,
      "reward": 0.29795128107070923,
      "reward_std": 0.09185018390417099,
      "reward_std_all": 0.08937744423747063,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9332840144634247,
      "rewards/forensics_tempsamp_iou_reward": 0.29795128107070923,
      "step": 1374
    },
    {
      "completion_length": 52.3125,
      "epoch": 3.6577896138482027,
      "grad_norm": 6.127125263214111,
      "kl": 0.59375,
      "learning_rate": 8.4e-08,
      "loss": 0.0238,
      "reward": 0.26174991577863693,
      "reward_std": 0.08163468539714813,
      "reward_std_all": 0.07830369099974632,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7480127513408661,
      "rewards/forensics_tempsamp_iou_reward": 0.26174991577863693,
      "step": 1375
    },
    {
      "completion_length": 74.6875,
      "epoch": 3.6604527296937417,
      "grad_norm": 7.109124183654785,
      "kl": 0.6484375,
      "learning_rate": 8.333333333333333e-08,
      "loss": 0.026,
      "reward": 0.1856754943728447,
      "reward_std": 0.07279416918754578,
      "reward_std_all": 0.08635369315743446,
      "rewards-max/forensics_tempsamp_iou_reward": 0.49937865138053894,
      "rewards/forensics_tempsamp_iou_reward": 0.1856754943728447,
      "step": 1376
    },
    {
      "completion_length": 51.03125,
      "epoch": 3.6631158455392807,
      "grad_norm": 6.916115760803223,
      "kl": 0.55078125,
      "learning_rate": 8.266666666666667e-08,
      "loss": 0.0221,
      "reward": 0.21155624091625214,
      "reward_std": 0.055636437609791756,
      "reward_std_all": 0.062365178018808365,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7652743458747864,
      "rewards/forensics_tempsamp_iou_reward": 0.21155624091625214,
      "step": 1377
    },
    {
      "completion_length": 104.53125,
      "epoch": 3.66577896138482,
      "grad_norm": 7.296109676361084,
      "kl": 0.7177734375,
      "learning_rate": 8.2e-08,
      "loss": 0.0287,
      "reward": 0.2523684799671173,
      "reward_std": 0.07792669534683228,
      "reward_std_all": 0.09613566845655441,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9010575115680695,
      "rewards/forensics_tempsamp_iou_reward": 0.2523684799671173,
      "step": 1378
    },
    {
      "completion_length": 58.9375,
      "epoch": 3.6684420772303596,
      "grad_norm": 8.70927619934082,
      "kl": 0.58203125,
      "learning_rate": 8.133333333333332e-08,
      "loss": 0.0233,
      "reward": 0.3099089711904526,
      "reward_std": 0.07248367182910442,
      "reward_std_all": 0.09433358535170555,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7322718799114227,
      "rewards/forensics_tempsamp_iou_reward": 0.3099089711904526,
      "step": 1379
    },
    {
      "completion_length": 65.859375,
      "epoch": 3.6711051930758987,
      "grad_norm": 8.506265640258789,
      "kl": 0.53125,
      "learning_rate": 8.066666666666666e-08,
      "loss": 0.0213,
      "reward": 0.21358591318130493,
      "reward_std": 0.10815012082457542,
      "reward_std_all": 0.1186959333717823,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6796279400587082,
      "rewards/forensics_tempsamp_iou_reward": 0.21358591318130493,
      "step": 1380
    },
    {
      "completion_length": 58.53125,
      "epoch": 3.673768308921438,
      "grad_norm": 9.86050033569336,
      "kl": 0.4814453125,
      "learning_rate": 8e-08,
      "loss": 0.0192,
      "reward": 0.1918066143989563,
      "reward_std": 0.14418941736221313,
      "reward_std_all": 0.1341446191072464,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7994949221611023,
      "rewards/forensics_tempsamp_iou_reward": 0.1918066143989563,
      "step": 1381
    },
    {
      "completion_length": 64.65625,
      "epoch": 3.676431424766977,
      "grad_norm": 7.6409502029418945,
      "kl": 0.59375,
      "learning_rate": 7.933333333333333e-08,
      "loss": 0.0237,
      "reward": 0.1592867225408554,
      "reward_std": 0.06478794291615486,
      "reward_std_all": 0.08893200382590294,
      "rewards-max/forensics_tempsamp_iou_reward": 0.46415816247463226,
      "rewards/forensics_tempsamp_iou_reward": 0.1592867225408554,
      "step": 1382
    },
    {
      "completion_length": 50.359375,
      "epoch": 3.6790945406125166,
      "grad_norm": 6.578601360321045,
      "kl": 0.494140625,
      "learning_rate": 7.866666666666666e-08,
      "loss": 0.0197,
      "reward": 0.3311129957437515,
      "reward_std": 0.07793275266885757,
      "reward_std_all": 0.08338301628828049,
      "rewards-max/forensics_tempsamp_iou_reward": 0.959756463766098,
      "rewards/forensics_tempsamp_iou_reward": 0.3311129957437515,
      "step": 1383
    },
    {
      "completion_length": 82.203125,
      "epoch": 3.681757656458056,
      "grad_norm": 7.175759315490723,
      "kl": 0.65234375,
      "learning_rate": 7.8e-08,
      "loss": 0.0261,
      "reward": 0.18569833040237427,
      "reward_std": 0.06742661818861961,
      "reward_std_all": 0.08135607838630676,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6530934423208237,
      "rewards/forensics_tempsamp_iou_reward": 0.18569833040237427,
      "step": 1384
    },
    {
      "completion_length": 82.296875,
      "epoch": 3.684420772303595,
      "grad_norm": 7.2503838539123535,
      "kl": 0.818359375,
      "learning_rate": 7.733333333333334e-08,
      "loss": 0.0328,
      "reward": 0.2599392831325531,
      "reward_std": 0.11783238127827644,
      "reward_std_all": 0.13725081458687782,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9486111104488373,
      "rewards/forensics_tempsamp_iou_reward": 0.2599392831325531,
      "step": 1385
    },
    {
      "completion_length": 90.1875,
      "epoch": 3.6870838881491346,
      "grad_norm": 10.442767143249512,
      "kl": 0.6240234375,
      "learning_rate": 7.666666666666665e-08,
      "loss": 0.025,
      "reward": 0.15049375593662262,
      "reward_std": 0.07329430058598518,
      "reward_std_all": 0.09677610546350479,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5262152850627899,
      "rewards/forensics_tempsamp_iou_reward": 0.15049375593662262,
      "step": 1386
    },
    {
      "completion_length": 37.140625,
      "epoch": 3.6897470039946736,
      "grad_norm": 6.363940238952637,
      "kl": 0.46875,
      "learning_rate": 7.599999999999999e-08,
      "loss": 0.0187,
      "reward": 0.25714029371738434,
      "reward_std": 0.08372295647859573,
      "reward_std_all": 0.08863892033696175,
      "rewards-max/forensics_tempsamp_iou_reward": 0.802815318107605,
      "rewards/forensics_tempsamp_iou_reward": 0.25714029371738434,
      "step": 1387
    },
    {
      "completion_length": 83.421875,
      "epoch": 3.692410119840213,
      "grad_norm": 7.896991729736328,
      "kl": 0.677734375,
      "learning_rate": 7.533333333333333e-08,
      "loss": 0.0272,
      "reward": 0.2740170955657959,
      "reward_std": 0.1114010363817215,
      "reward_std_all": 0.11555814370512962,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9642326235771179,
      "rewards/forensics_tempsamp_iou_reward": 0.2740170955657959,
      "step": 1388
    },
    {
      "completion_length": 51.4375,
      "epoch": 3.6950732356857525,
      "grad_norm": 7.9171624183654785,
      "kl": 0.732421875,
      "learning_rate": 7.466666666666667e-08,
      "loss": 0.0293,
      "reward": 0.25675392150878906,
      "reward_std": 0.060736946761608124,
      "reward_std_all": 0.06707599386572838,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9756596982479095,
      "rewards/forensics_tempsamp_iou_reward": 0.25675392150878906,
      "step": 1389
    },
    {
      "completion_length": 71.078125,
      "epoch": 3.6977363515312915,
      "grad_norm": 6.967497825622559,
      "kl": 0.595703125,
      "learning_rate": 7.399999999999999e-08,
      "loss": 0.0238,
      "reward": 0.23363736271858215,
      "reward_std": 0.11799874529242516,
      "reward_std_all": 0.13041505217552185,
      "rewards-max/forensics_tempsamp_iou_reward": 0.903124988079071,
      "rewards/forensics_tempsamp_iou_reward": 0.23363736271858215,
      "step": 1390
    },
    {
      "completion_length": 47.296875,
      "epoch": 3.700399467376831,
      "grad_norm": 6.608709335327148,
      "kl": 0.4609375,
      "learning_rate": 7.333333333333333e-08,
      "loss": 0.0184,
      "reward": 0.3199664056301117,
      "reward_std": 0.10124045237898827,
      "reward_std_all": 0.11907793581485748,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9574333131313324,
      "rewards/forensics_tempsamp_iou_reward": 0.3199664056301117,
      "step": 1391
    },
    {
      "completion_length": 173.484375,
      "epoch": 3.70306258322237,
      "grad_norm": 7.216620922088623,
      "kl": 0.974609375,
      "learning_rate": 7.266666666666667e-08,
      "loss": 0.039,
      "reward": 0.223733838647604,
      "reward_std": 0.07813091948628426,
      "reward_std_all": 0.0781632512807846,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7069558203220367,
      "rewards/forensics_tempsamp_iou_reward": 0.223733838647604,
      "step": 1392
    },
    {
      "completion_length": 74.140625,
      "epoch": 3.7057256990679095,
      "grad_norm": 6.564273357391357,
      "kl": 0.41796875,
      "learning_rate": 7.2e-08,
      "loss": 0.0167,
      "reward": 0.3654390871524811,
      "reward_std": 0.09433335065841675,
      "reward_std_all": 0.08733730018138885,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8958133459091187,
      "rewards/forensics_tempsamp_iou_reward": 0.3654390871524811,
      "step": 1393
    },
    {
      "completion_length": 89.3125,
      "epoch": 3.708388814913449,
      "grad_norm": 9.63992691040039,
      "kl": 0.728515625,
      "learning_rate": 7.133333333333332e-08,
      "loss": 0.0292,
      "reward": 0.2625349462032318,
      "reward_std": 0.14764832705259323,
      "reward_std_all": 0.13751386851072311,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8386939167976379,
      "rewards/forensics_tempsamp_iou_reward": 0.2625349462032318,
      "step": 1394
    },
    {
      "completion_length": 116.21875,
      "epoch": 3.711051930758988,
      "grad_norm": 7.76474142074585,
      "kl": 0.751953125,
      "learning_rate": 7.066666666666666e-08,
      "loss": 0.0301,
      "reward": 0.11726776510477066,
      "reward_std": 0.0730358213186264,
      "reward_std_all": 0.08147366717457771,
      "rewards-max/forensics_tempsamp_iou_reward": 0.39351852238178253,
      "rewards/forensics_tempsamp_iou_reward": 0.11726776510477066,
      "step": 1395
    },
    {
      "completion_length": 94.234375,
      "epoch": 3.7137150466045274,
      "grad_norm": 6.559638023376465,
      "kl": 0.705078125,
      "learning_rate": 7e-08,
      "loss": 0.0282,
      "reward": 0.2772260904312134,
      "reward_std": 0.08464209921658039,
      "reward_std_all": 0.09147552028298378,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9039822220802307,
      "rewards/forensics_tempsamp_iou_reward": 0.2772260904312134,
      "step": 1396
    },
    {
      "completion_length": 81.890625,
      "epoch": 3.7163781624500665,
      "grad_norm": 5.979106903076172,
      "kl": 0.75390625,
      "learning_rate": 6.933333333333333e-08,
      "loss": 0.0302,
      "reward": 0.15338758379220963,
      "reward_std": 0.09224838577210903,
      "reward_std_all": 0.08300774917006493,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6173586547374725,
      "rewards/forensics_tempsamp_iou_reward": 0.15338758379220963,
      "step": 1397
    },
    {
      "completion_length": 66.859375,
      "epoch": 3.719041278295606,
      "grad_norm": 5.935772895812988,
      "kl": 0.6015625,
      "learning_rate": 6.866666666666666e-08,
      "loss": 0.0241,
      "reward": 0.2811465859413147,
      "reward_std": 0.11835392192006111,
      "reward_std_all": 0.13311586529016495,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8319978415966034,
      "rewards/forensics_tempsamp_iou_reward": 0.2811465859413147,
      "step": 1398
    },
    {
      "completion_length": 123.28125,
      "epoch": 3.7217043941411454,
      "grad_norm": 7.259429454803467,
      "kl": 0.62109375,
      "learning_rate": 6.8e-08,
      "loss": 0.0249,
      "reward": 0.2811242938041687,
      "reward_std": 0.12627768144011497,
      "reward_std_all": 0.12240491062402725,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9689243137836456,
      "rewards/forensics_tempsamp_iou_reward": 0.2811242938041687,
      "step": 1399
    },
    {
      "completion_length": 64.78125,
      "epoch": 3.7243675099866844,
      "grad_norm": 6.175867557525635,
      "kl": 0.708984375,
      "learning_rate": 6.733333333333333e-08,
      "loss": 0.0283,
      "reward": 0.25725598633289337,
      "reward_std": 0.12459329515695572,
      "reward_std_all": 0.11625881493091583,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7548534870147705,
      "rewards/forensics_tempsamp_iou_reward": 0.25725598633289337,
      "step": 1400
    },
    {
      "completion_length": 35.25,
      "epoch": 3.7270306258322234,
      "grad_norm": 7.379660606384277,
      "kl": 0.3955078125,
      "learning_rate": 6.666666666666667e-08,
      "loss": 0.0159,
      "reward": 0.22456099838018417,
      "reward_std": 0.09320906549692154,
      "reward_std_all": 0.08505652844905853,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7562930881977081,
      "rewards/forensics_tempsamp_iou_reward": 0.22456099838018417,
      "step": 1401
    },
    {
      "completion_length": 90.859375,
      "epoch": 3.729693741677763,
      "grad_norm": 6.666146278381348,
      "kl": 0.5126953125,
      "learning_rate": 6.6e-08,
      "loss": 0.0205,
      "reward": 0.3294059485197067,
      "reward_std": 0.13553017005324364,
      "reward_std_all": 0.1436486840248108,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9920212626457214,
      "rewards/forensics_tempsamp_iou_reward": 0.3294059485197067,
      "step": 1402
    },
    {
      "completion_length": 80.734375,
      "epoch": 3.7323568575233024,
      "grad_norm": 6.423831462860107,
      "kl": 0.576171875,
      "learning_rate": 6.533333333333332e-08,
      "loss": 0.023,
      "reward": 0.24147462099790573,
      "reward_std": 0.1349584236741066,
      "reward_std_all": 0.1265207640826702,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6958231776952744,
      "rewards/forensics_tempsamp_iou_reward": 0.24147462099790573,
      "step": 1403
    },
    {
      "completion_length": 64.84375,
      "epoch": 3.7350199733688414,
      "grad_norm": 7.92807674407959,
      "kl": 0.921875,
      "learning_rate": 6.466666666666666e-08,
      "loss": 0.037,
      "reward": 0.33071038126945496,
      "reward_std": 0.07660849764943123,
      "reward_std_all": 0.08829229325056076,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8545586466789246,
      "rewards/forensics_tempsamp_iou_reward": 0.33071038126945496,
      "step": 1404
    },
    {
      "completion_length": 110.375,
      "epoch": 3.737683089214381,
      "grad_norm": 8.06119441986084,
      "kl": 0.67578125,
      "learning_rate": 6.4e-08,
      "loss": 0.027,
      "reward": 0.19822578132152557,
      "reward_std": 0.10866658762097359,
      "reward_std_all": 0.12182365357875824,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6877467930316925,
      "rewards/forensics_tempsamp_iou_reward": 0.19822578132152557,
      "step": 1405
    },
    {
      "completion_length": 58.71875,
      "epoch": 3.74034620505992,
      "grad_norm": 10.706305503845215,
      "kl": 0.7578125,
      "learning_rate": 6.333333333333333e-08,
      "loss": 0.0303,
      "reward": 0.3126893937587738,
      "reward_std": 0.168880432844162,
      "reward_std_all": 0.18710077553987503,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8620719909667969,
      "rewards/forensics_tempsamp_iou_reward": 0.3126893937587738,
      "step": 1406
    },
    {
      "completion_length": 77.609375,
      "epoch": 3.7430093209054593,
      "grad_norm": 6.985531330108643,
      "kl": 0.5947265625,
      "learning_rate": 6.266666666666666e-08,
      "loss": 0.0238,
      "reward": 0.24688231199979782,
      "reward_std": 0.0787038579583168,
      "reward_std_all": 0.07193859666585922,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9215156733989716,
      "rewards/forensics_tempsamp_iou_reward": 0.24688231199979782,
      "step": 1407
    },
    {
      "completion_length": 87.5,
      "epoch": 3.745672436750999,
      "grad_norm": 8.101543426513672,
      "kl": 0.6220703125,
      "learning_rate": 6.2e-08,
      "loss": 0.0249,
      "reward": 0.2404617890715599,
      "reward_std": 0.12700499594211578,
      "reward_std_all": 0.13176733255386353,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8386728167533875,
      "rewards/forensics_tempsamp_iou_reward": 0.2404617890715599,
      "step": 1408
    },
    {
      "completion_length": 114.828125,
      "epoch": 3.748335552596538,
      "grad_norm": 6.640694618225098,
      "kl": 0.841796875,
      "learning_rate": 6.133333333333333e-08,
      "loss": 0.0337,
      "reward": 0.1823241040110588,
      "reward_std": 0.0830938071012497,
      "reward_std_all": 0.09162938594818115,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6668749898672104,
      "rewards/forensics_tempsamp_iou_reward": 0.1823241040110588,
      "step": 1409
    },
    {
      "completion_length": 62.359375,
      "epoch": 3.7509986684420773,
      "grad_norm": 7.735201358795166,
      "kl": 0.5302734375,
      "learning_rate": 6.066666666666667e-08,
      "loss": 0.0212,
      "reward": 0.3202361911535263,
      "reward_std": 0.10017607361078262,
      "reward_std_all": 0.10063987970352173,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7505738437175751,
      "rewards/forensics_tempsamp_iou_reward": 0.3202361911535263,
      "step": 1410
    },
    {
      "completion_length": 68.28125,
      "epoch": 3.7536617842876163,
      "grad_norm": 7.605510711669922,
      "kl": 0.556640625,
      "learning_rate": 6e-08,
      "loss": 0.0223,
      "reward": 0.3808708190917969,
      "reward_std": 0.13111061602830887,
      "reward_std_all": 0.1262989453971386,
      "rewards-max/forensics_tempsamp_iou_reward": 0.991104394197464,
      "rewards/forensics_tempsamp_iou_reward": 0.3808708190917969,
      "step": 1411
    },
    {
      "completion_length": 81.1875,
      "epoch": 3.756324900133156,
      "grad_norm": 9.201498985290527,
      "kl": 0.66015625,
      "learning_rate": 5.9333333333333335e-08,
      "loss": 0.0264,
      "reward": 0.2503710836172104,
      "reward_std": 0.0794895701110363,
      "reward_std_all": 0.08124836534261703,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7475031912326813,
      "rewards/forensics_tempsamp_iou_reward": 0.2503710836172104,
      "step": 1412
    },
    {
      "completion_length": 43.296875,
      "epoch": 3.7589880159786953,
      "grad_norm": 8.481043815612793,
      "kl": 0.59765625,
      "learning_rate": 5.866666666666666e-08,
      "loss": 0.0239,
      "reward": 0.32899754494428635,
      "reward_std": 0.09494398534297943,
      "reward_std_all": 0.08932143822312355,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8909804224967957,
      "rewards/forensics_tempsamp_iou_reward": 0.32899754494428635,
      "step": 1413
    },
    {
      "completion_length": 74.875,
      "epoch": 3.7616511318242343,
      "grad_norm": 9.873522758483887,
      "kl": 0.755859375,
      "learning_rate": 5.8e-08,
      "loss": 0.0302,
      "reward": 0.3174866735935211,
      "reward_std": 0.12960810959339142,
      "reward_std_all": 0.12218261510133743,
      "rewards-max/forensics_tempsamp_iou_reward": 0.963345855474472,
      "rewards/forensics_tempsamp_iou_reward": 0.3174866735935211,
      "step": 1414
    },
    {
      "completion_length": 85.609375,
      "epoch": 3.7643142476697737,
      "grad_norm": 7.056171417236328,
      "kl": 0.689453125,
      "learning_rate": 5.733333333333333e-08,
      "loss": 0.0276,
      "reward": 0.19741545245051384,
      "reward_std": 0.0553523488342762,
      "reward_std_all": 0.07658253237605095,
      "rewards-max/forensics_tempsamp_iou_reward": 0.589157797396183,
      "rewards/forensics_tempsamp_iou_reward": 0.19741545245051384,
      "step": 1415
    },
    {
      "completion_length": 54.96875,
      "epoch": 3.7669773635153128,
      "grad_norm": 8.390982627868652,
      "kl": 0.546875,
      "learning_rate": 5.666666666666666e-08,
      "loss": 0.0219,
      "reward": 0.2866446077823639,
      "reward_std": 0.0962323471903801,
      "reward_std_all": 0.10019703954458237,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8030822277069092,
      "rewards/forensics_tempsamp_iou_reward": 0.2866446077823639,
      "step": 1416
    },
    {
      "completion_length": 47.109375,
      "epoch": 3.7696404793608522,
      "grad_norm": 24.29985237121582,
      "kl": 1.5703125,
      "learning_rate": 5.6e-08,
      "loss": 0.0629,
      "reward": 0.28522633016109467,
      "reward_std": 0.11429783329367638,
      "reward_std_all": 0.12064677104353905,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6930082440376282,
      "rewards/forensics_tempsamp_iou_reward": 0.28522633016109467,
      "step": 1417
    },
    {
      "completion_length": 59.125,
      "epoch": 3.7723035952063917,
      "grad_norm": 6.4276862144470215,
      "kl": 0.4267578125,
      "learning_rate": 5.5333333333333326e-08,
      "loss": 0.0171,
      "reward": 0.28373339027166367,
      "reward_std": 0.08501791208982468,
      "reward_std_all": 0.09292265400290489,
      "rewards-max/forensics_tempsamp_iou_reward": 0.89402174949646,
      "rewards/forensics_tempsamp_iou_reward": 0.28373339027166367,
      "step": 1418
    },
    {
      "completion_length": 48.71875,
      "epoch": 3.7749667110519307,
      "grad_norm": 9.365870475769043,
      "kl": 0.794921875,
      "learning_rate": 5.4666666666666666e-08,
      "loss": 0.0317,
      "reward": 0.22497066110372543,
      "reward_std": 0.0970868431031704,
      "reward_std_all": 0.09929521754384041,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5997474789619446,
      "rewards/forensics_tempsamp_iou_reward": 0.22497066110372543,
      "step": 1419
    },
    {
      "completion_length": 91.953125,
      "epoch": 3.77762982689747,
      "grad_norm": 6.227244853973389,
      "kl": 1.0625,
      "learning_rate": 5.3999999999999994e-08,
      "loss": 0.0424,
      "reward": 0.18047649413347244,
      "reward_std": 0.07116115465760231,
      "reward_std_all": 0.06577415019273758,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6912878900766373,
      "rewards/forensics_tempsamp_iou_reward": 0.18047649413347244,
      "step": 1420
    },
    {
      "completion_length": 70.640625,
      "epoch": 3.780292942743009,
      "grad_norm": 6.867727279663086,
      "kl": 0.73828125,
      "learning_rate": 5.3333333333333334e-08,
      "loss": 0.0295,
      "reward": 0.2141377478837967,
      "reward_std": 0.08859726414084435,
      "reward_std_all": 0.09236960113048553,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6848442554473877,
      "rewards/forensics_tempsamp_iou_reward": 0.2141377478837967,
      "step": 1421
    },
    {
      "completion_length": 85.8125,
      "epoch": 3.7829560585885487,
      "grad_norm": 8.307931900024414,
      "kl": 0.470703125,
      "learning_rate": 5.266666666666666e-08,
      "loss": 0.0188,
      "reward": 0.2231016606092453,
      "reward_std": 0.11207074671983719,
      "reward_std_all": 0.11235000193119049,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8294642865657806,
      "rewards/forensics_tempsamp_iou_reward": 0.2231016606092453,
      "step": 1422
    },
    {
      "completion_length": 66.0625,
      "epoch": 3.785619174434088,
      "grad_norm": 7.443784236907959,
      "kl": 0.66015625,
      "learning_rate": 5.1999999999999996e-08,
      "loss": 0.0264,
      "reward": 0.24480071663856506,
      "reward_std": 0.12952952086925507,
      "reward_std_all": 0.12533502280712128,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7616941928863525,
      "rewards/forensics_tempsamp_iou_reward": 0.24480071663856506,
      "step": 1423
    },
    {
      "completion_length": 89.859375,
      "epoch": 3.788282290279627,
      "grad_norm": 8.561107635498047,
      "kl": 0.57421875,
      "learning_rate": 5.133333333333333e-08,
      "loss": 0.023,
      "reward": 0.33709439635276794,
      "reward_std": 0.17423061281442642,
      "reward_std_all": 0.15484514087438583,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8593803346157074,
      "rewards/forensics_tempsamp_iou_reward": 0.33709439635276794,
      "step": 1424
    },
    {
      "completion_length": 107.5625,
      "epoch": 3.790945406125166,
      "grad_norm": 6.769015312194824,
      "kl": 0.6845703125,
      "learning_rate": 5.0666666666666664e-08,
      "loss": 0.0274,
      "reward": 0.1935296207666397,
      "reward_std": 0.10034444183111191,
      "reward_std_all": 0.09724152833223343,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7586266398429871,
      "rewards/forensics_tempsamp_iou_reward": 0.1935296207666397,
      "step": 1425
    },
    {
      "completion_length": 57.03125,
      "epoch": 3.7936085219707056,
      "grad_norm": 5.666106700897217,
      "kl": 0.4423828125,
      "learning_rate": 5e-08,
      "loss": 0.0177,
      "reward": 0.4102146476507187,
      "reward_std": 0.04608460143208504,
      "reward_std_all": 0.06488636136054993,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9648790657520294,
      "rewards/forensics_tempsamp_iou_reward": 0.4102146476507187,
      "step": 1426
    },
    {
      "completion_length": 58.1875,
      "epoch": 3.796271637816245,
      "grad_norm": 6.267786026000977,
      "kl": 0.4853515625,
      "learning_rate": 4.933333333333333e-08,
      "loss": 0.0194,
      "reward": 0.19097967445850372,
      "reward_std": 0.07388913258910179,
      "reward_std_all": 0.07903417199850082,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6519680917263031,
      "rewards/forensics_tempsamp_iou_reward": 0.19097967445850372,
      "step": 1427
    },
    {
      "completion_length": 59.046875,
      "epoch": 3.798934753661784,
      "grad_norm": 7.305250644683838,
      "kl": 0.5068359375,
      "learning_rate": 4.866666666666666e-08,
      "loss": 0.0203,
      "reward": 0.2417309284210205,
      "reward_std": 0.0688055083155632,
      "reward_std_all": 0.08535336330533028,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7066666632890701,
      "rewards/forensics_tempsamp_iou_reward": 0.2417309284210205,
      "step": 1428
    },
    {
      "completion_length": 72.8125,
      "epoch": 3.8015978695073236,
      "grad_norm": 8.080949783325195,
      "kl": 0.66796875,
      "learning_rate": 4.8e-08,
      "loss": 0.0267,
      "reward": 0.23832973837852478,
      "reward_std": 0.1015089601278305,
      "reward_std_all": 0.1029728539288044,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8640545010566711,
      "rewards/forensics_tempsamp_iou_reward": 0.23832973837852478,
      "step": 1429
    },
    {
      "completion_length": 50.609375,
      "epoch": 3.8042609853528626,
      "grad_norm": 7.251799583435059,
      "kl": 0.4677734375,
      "learning_rate": 4.733333333333333e-08,
      "loss": 0.0187,
      "reward": 0.16030287742614746,
      "reward_std": 0.07971447892487049,
      "reward_std_all": 0.07934064790606499,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6125802397727966,
      "rewards/forensics_tempsamp_iou_reward": 0.16030287742614746,
      "step": 1430
    },
    {
      "completion_length": 67.96875,
      "epoch": 3.806924101198402,
      "grad_norm": 6.929407119750977,
      "kl": 0.6484375,
      "learning_rate": 4.666666666666667e-08,
      "loss": 0.0259,
      "reward": 0.2644018679857254,
      "reward_std": 0.07007848471403122,
      "reward_std_all": 0.08424598723649979,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8579118251800537,
      "rewards/forensics_tempsamp_iou_reward": 0.2644018679857254,
      "step": 1431
    },
    {
      "completion_length": 51.484375,
      "epoch": 3.8095872170439415,
      "grad_norm": 7.590999126434326,
      "kl": 0.5107421875,
      "learning_rate": 4.5999999999999995e-08,
      "loss": 0.0204,
      "reward": 0.2861444801092148,
      "reward_std": 0.1395893469452858,
      "reward_std_all": 0.13640540838241577,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9633021950721741,
      "rewards/forensics_tempsamp_iou_reward": 0.2861444801092148,
      "step": 1432
    },
    {
      "completion_length": 51.734375,
      "epoch": 3.8122503328894806,
      "grad_norm": 6.952057838439941,
      "kl": 0.4755859375,
      "learning_rate": 4.5333333333333336e-08,
      "loss": 0.019,
      "reward": 0.31784213334321976,
      "reward_std": 0.11369860172271729,
      "reward_std_all": 0.11146035045385361,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9885112941265106,
      "rewards/forensics_tempsamp_iou_reward": 0.31784213334321976,
      "step": 1433
    },
    {
      "completion_length": 50.046875,
      "epoch": 3.81491344873502,
      "grad_norm": 8.502201080322266,
      "kl": 0.537109375,
      "learning_rate": 4.466666666666666e-08,
      "loss": 0.0216,
      "reward": 0.30128222703933716,
      "reward_std": 0.1285492703318596,
      "reward_std_all": 0.11622438579797745,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8016268610954285,
      "rewards/forensics_tempsamp_iou_reward": 0.30128222703933716,
      "step": 1434
    },
    {
      "completion_length": 73.796875,
      "epoch": 3.817576564580559,
      "grad_norm": 6.068613529205322,
      "kl": 0.52734375,
      "learning_rate": 4.4e-08,
      "loss": 0.0211,
      "reward": 0.3559842109680176,
      "reward_std": 0.1441579982638359,
      "reward_std_all": 0.13801200687885284,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9574430286884308,
      "rewards/forensics_tempsamp_iou_reward": 0.3559842109680176,
      "step": 1435
    },
    {
      "completion_length": 36.3125,
      "epoch": 3.8202396804260985,
      "grad_norm": 7.399556636810303,
      "kl": 0.703125,
      "learning_rate": 4.333333333333333e-08,
      "loss": 0.0281,
      "reward": 0.47084125876426697,
      "reward_std": 0.1420947052538395,
      "reward_std_all": 0.13537975400686264,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9810074865818024,
      "rewards/forensics_tempsamp_iou_reward": 0.47084125876426697,
      "step": 1436
    },
    {
      "completion_length": 93.40625,
      "epoch": 3.822902796271638,
      "grad_norm": 6.836562633514404,
      "kl": 0.7236328125,
      "learning_rate": 4.2666666666666665e-08,
      "loss": 0.029,
      "reward": 0.2787610590457916,
      "reward_std": 0.163833387196064,
      "reward_std_all": 0.16775567829608917,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9559286534786224,
      "rewards/forensics_tempsamp_iou_reward": 0.2787610590457916,
      "step": 1437
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.825565912117177,
      "grad_norm": 6.759432315826416,
      "kl": 0.67578125,
      "learning_rate": 4.2e-08,
      "loss": 0.027,
      "reward": 0.4058362543582916,
      "reward_std": 0.13162709027528763,
      "reward_std_all": 0.14130032807588577,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9904488623142242,
      "rewards/forensics_tempsamp_iou_reward": 0.4058362543582916,
      "step": 1438
    },
    {
      "completion_length": 60.75,
      "epoch": 3.8282290279627165,
      "grad_norm": 8.023249626159668,
      "kl": 0.6953125,
      "learning_rate": 4.133333333333333e-08,
      "loss": 0.0278,
      "reward": 0.3095630407333374,
      "reward_std": 0.13239258527755737,
      "reward_std_all": 0.12898240610957146,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8444444239139557,
      "rewards/forensics_tempsamp_iou_reward": 0.3095630407333374,
      "step": 1439
    },
    {
      "completion_length": 31.359375,
      "epoch": 3.8308921438082555,
      "grad_norm": 7.552297592163086,
      "kl": 0.4599609375,
      "learning_rate": 4.066666666666666e-08,
      "loss": 0.0184,
      "reward": 0.3563598692417145,
      "reward_std": 0.0635056123137474,
      "reward_std_all": 0.09672678261995316,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9936606287956238,
      "rewards/forensics_tempsamp_iou_reward": 0.3563598692417145,
      "step": 1440
    },
    {
      "completion_length": 105.875,
      "epoch": 3.833555259653795,
      "grad_norm": 7.082237720489502,
      "kl": 0.591796875,
      "learning_rate": 4e-08,
      "loss": 0.0237,
      "reward": 0.2440130114555359,
      "reward_std": 0.1095648929476738,
      "reward_std_all": 0.12298275530338287,
      "rewards-max/forensics_tempsamp_iou_reward": 0.800309956073761,
      "rewards/forensics_tempsamp_iou_reward": 0.2440130114555359,
      "step": 1441
    },
    {
      "completion_length": 48.375,
      "epoch": 3.8362183754993344,
      "grad_norm": 7.519127368927002,
      "kl": 0.521484375,
      "learning_rate": 3.933333333333333e-08,
      "loss": 0.0209,
      "reward": 0.1748124063014984,
      "reward_std": 0.05040998198091984,
      "reward_std_all": 0.05578366667032242,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7536468505859375,
      "rewards/forensics_tempsamp_iou_reward": 0.1748124063014984,
      "step": 1442
    },
    {
      "completion_length": 82.21875,
      "epoch": 3.8388814913448734,
      "grad_norm": 9.794856071472168,
      "kl": 0.814453125,
      "learning_rate": 3.866666666666667e-08,
      "loss": 0.0326,
      "reward": 0.3242434710264206,
      "reward_std": 0.08881944045424461,
      "reward_std_all": 0.08190956711769104,
      "rewards-max/forensics_tempsamp_iou_reward": 0.990340918302536,
      "rewards/forensics_tempsamp_iou_reward": 0.3242434710264206,
      "step": 1443
    },
    {
      "completion_length": 56.71875,
      "epoch": 3.841544607190413,
      "grad_norm": 6.894487380981445,
      "kl": 0.3564453125,
      "learning_rate": 3.7999999999999996e-08,
      "loss": 0.0143,
      "reward": 0.36652426421642303,
      "reward_std": 0.08980544283986092,
      "reward_std_all": 0.08559947088360786,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9932494461536407,
      "rewards/forensics_tempsamp_iou_reward": 0.36652426421642303,
      "step": 1444
    },
    {
      "completion_length": 55.59375,
      "epoch": 3.844207723035952,
      "grad_norm": 7.921837329864502,
      "kl": 0.498046875,
      "learning_rate": 3.733333333333334e-08,
      "loss": 0.0199,
      "reward": 0.3282070457935333,
      "reward_std": 0.09983186051249504,
      "reward_std_all": 0.09226395934820175,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9344806671142578,
      "rewards/forensics_tempsamp_iou_reward": 0.3282070457935333,
      "step": 1445
    },
    {
      "completion_length": 72.40625,
      "epoch": 3.8468708388814914,
      "grad_norm": 6.057131767272949,
      "kl": 0.771484375,
      "learning_rate": 3.6666666666666664e-08,
      "loss": 0.0308,
      "reward": 0.30152513086795807,
      "reward_std": 0.07555434107780457,
      "reward_std_all": 0.07195603847503662,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9790315330028534,
      "rewards/forensics_tempsamp_iou_reward": 0.30152513086795807,
      "step": 1446
    },
    {
      "completion_length": 50.34375,
      "epoch": 3.849533954727031,
      "grad_norm": 10.11002254486084,
      "kl": 0.53515625,
      "learning_rate": 3.6e-08,
      "loss": 0.0214,
      "reward": 0.38302164524793625,
      "reward_std": 0.07966551184654236,
      "reward_std_all": 0.09419894777238369,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9500892460346222,
      "rewards/forensics_tempsamp_iou_reward": 0.38302164524793625,
      "step": 1447
    },
    {
      "completion_length": 111.46875,
      "epoch": 3.85219707057257,
      "grad_norm": 10.284345626831055,
      "kl": 1.01171875,
      "learning_rate": 3.533333333333333e-08,
      "loss": 0.0404,
      "reward": 0.2334664762020111,
      "reward_std": 0.14676057547330856,
      "reward_std_all": 0.14016326516866684,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7600090205669403,
      "rewards/forensics_tempsamp_iou_reward": 0.2334664762020111,
      "step": 1448
    },
    {
      "completion_length": 49.125,
      "epoch": 3.8548601864181093,
      "grad_norm": 8.248061180114746,
      "kl": 0.4970703125,
      "learning_rate": 3.4666666666666666e-08,
      "loss": 0.0198,
      "reward": 0.38225962221622467,
      "reward_std": 0.08998743817210197,
      "reward_std_all": 0.10272883996367455,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8297527730464935,
      "rewards/forensics_tempsamp_iou_reward": 0.38225962221622467,
      "step": 1449
    },
    {
      "completion_length": 70.609375,
      "epoch": 3.8575233022636484,
      "grad_norm": 6.348260402679443,
      "kl": 0.77734375,
      "learning_rate": 3.4e-08,
      "loss": 0.0311,
      "reward": 0.2902192771434784,
      "reward_std": 0.12217871472239494,
      "reward_std_all": 0.12407732009887695,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8148109018802643,
      "rewards/forensics_tempsamp_iou_reward": 0.2902192771434784,
      "step": 1450
    },
    {
      "completion_length": 91.28125,
      "epoch": 3.860186418109188,
      "grad_norm": 7.04579496383667,
      "kl": 0.7890625,
      "learning_rate": 3.3333333333333334e-08,
      "loss": 0.0316,
      "reward": 0.27296774089336395,
      "reward_std": 0.08851085603237152,
      "reward_std_all": 0.08823995478451252,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9609279334545135,
      "rewards/forensics_tempsamp_iou_reward": 0.27296774089336395,
      "step": 1451
    },
    {
      "completion_length": 92.390625,
      "epoch": 3.8628495339547273,
      "grad_norm": 7.480047225952148,
      "kl": 0.828125,
      "learning_rate": 3.266666666666666e-08,
      "loss": 0.0331,
      "reward": 0.24386244267225266,
      "reward_std": 0.14507479220628738,
      "reward_std_all": 0.12487347424030304,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7839181125164032,
      "rewards/forensics_tempsamp_iou_reward": 0.24386244267225266,
      "step": 1452
    },
    {
      "completion_length": 99.734375,
      "epoch": 3.8655126498002663,
      "grad_norm": 5.99587345123291,
      "kl": 0.75,
      "learning_rate": 3.2e-08,
      "loss": 0.03,
      "reward": 0.28860943764448166,
      "reward_std": 0.11905363947153091,
      "reward_std_all": 0.11701997369527817,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9420523941516876,
      "rewards/forensics_tempsamp_iou_reward": 0.28860943764448166,
      "step": 1453
    },
    {
      "completion_length": 40.671875,
      "epoch": 3.8681757656458053,
      "grad_norm": 8.644628524780273,
      "kl": 0.498046875,
      "learning_rate": 3.133333333333333e-08,
      "loss": 0.0199,
      "reward": 0.24970079958438873,
      "reward_std": 0.10568784922361374,
      "reward_std_all": 0.10359913483262062,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7778356671333313,
      "rewards/forensics_tempsamp_iou_reward": 0.24970079958438873,
      "step": 1454
    },
    {
      "completion_length": 48.96875,
      "epoch": 3.870838881491345,
      "grad_norm": 12.077093124389648,
      "kl": 0.69921875,
      "learning_rate": 3.0666666666666663e-08,
      "loss": 0.028,
      "reward": 0.30714885890483856,
      "reward_std": 0.07953616604208946,
      "reward_std_all": 0.09321337193250656,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7224595844745636,
      "rewards/forensics_tempsamp_iou_reward": 0.30714885890483856,
      "step": 1455
    },
    {
      "completion_length": 99.015625,
      "epoch": 3.8735019973368843,
      "grad_norm": 10.250205039978027,
      "kl": 0.802734375,
      "learning_rate": 3e-08,
      "loss": 0.0321,
      "reward": 0.3148933947086334,
      "reward_std": 0.11393548548221588,
      "reward_std_all": 0.10402914509177208,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9800512492656708,
      "rewards/forensics_tempsamp_iou_reward": 0.3148933947086334,
      "step": 1456
    },
    {
      "completion_length": 69.828125,
      "epoch": 3.8761651131824233,
      "grad_norm": 8.915035247802734,
      "kl": 0.513671875,
      "learning_rate": 2.933333333333333e-08,
      "loss": 0.0205,
      "reward": 0.256245993077755,
      "reward_std": 0.11475708708167076,
      "reward_std_all": 0.11051029711961746,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8676335215568542,
      "rewards/forensics_tempsamp_iou_reward": 0.256245993077755,
      "step": 1457
    },
    {
      "completion_length": 54.828125,
      "epoch": 3.8788282290279628,
      "grad_norm": 8.385981559753418,
      "kl": 0.556640625,
      "learning_rate": 2.8666666666666665e-08,
      "loss": 0.0222,
      "reward": 0.35876624286174774,
      "reward_std": 0.07651467993855476,
      "reward_std_all": 0.08203140273690224,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9847008287906647,
      "rewards/forensics_tempsamp_iou_reward": 0.35876624286174774,
      "step": 1458
    },
    {
      "completion_length": 42.609375,
      "epoch": 3.881491344873502,
      "grad_norm": 5.882723808288574,
      "kl": 0.576171875,
      "learning_rate": 2.8e-08,
      "loss": 0.0231,
      "reward": 0.2534218803048134,
      "reward_std": 0.07474559545516968,
      "reward_std_all": 0.07344690710306168,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8586516082286835,
      "rewards/forensics_tempsamp_iou_reward": 0.2534218803048134,
      "step": 1459
    },
    {
      "completion_length": 47.90625,
      "epoch": 3.8841544607190412,
      "grad_norm": 8.701482772827148,
      "kl": 0.794921875,
      "learning_rate": 2.7333333333333333e-08,
      "loss": 0.0318,
      "reward": 0.2664850205183029,
      "reward_std": 0.11576187610626221,
      "reward_std_all": 0.11739012226462364,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6784188151359558,
      "rewards/forensics_tempsamp_iou_reward": 0.2664850205183029,
      "step": 1460
    },
    {
      "completion_length": 39.6875,
      "epoch": 3.8868175765645807,
      "grad_norm": 6.707154273986816,
      "kl": 0.6103515625,
      "learning_rate": 2.6666666666666667e-08,
      "loss": 0.0244,
      "reward": 0.3307982534170151,
      "reward_std": 0.0832194946706295,
      "reward_std_all": 0.07977622374892235,
      "rewards-max/forensics_tempsamp_iou_reward": 0.974765956401825,
      "rewards/forensics_tempsamp_iou_reward": 0.3307982534170151,
      "step": 1461
    },
    {
      "completion_length": 52.96875,
      "epoch": 3.8894806924101197,
      "grad_norm": 9.311327934265137,
      "kl": 1.1171875,
      "learning_rate": 2.5999999999999998e-08,
      "loss": 0.0447,
      "reward": 0.25577350705862045,
      "reward_std": 0.055205587297677994,
      "reward_std_all": 0.06608258001506329,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6642969995737076,
      "rewards/forensics_tempsamp_iou_reward": 0.25577350705862045,
      "step": 1462
    },
    {
      "completion_length": 59.671875,
      "epoch": 3.892143808255659,
      "grad_norm": 5.731664180755615,
      "kl": 0.734375,
      "learning_rate": 2.5333333333333332e-08,
      "loss": 0.0294,
      "reward": 0.3130272477865219,
      "reward_std": 0.12002746760845184,
      "reward_std_all": 0.1106007993221283,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9504595696926117,
      "rewards/forensics_tempsamp_iou_reward": 0.3130272477865219,
      "step": 1463
    },
    {
      "completion_length": 49.375,
      "epoch": 3.894806924101198,
      "grad_norm": 7.152234077453613,
      "kl": 0.787109375,
      "learning_rate": 2.4666666666666666e-08,
      "loss": 0.0314,
      "reward": 0.2532842308282852,
      "reward_std": 0.08651967346668243,
      "reward_std_all": 0.0835922546684742,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8472914695739746,
      "rewards/forensics_tempsamp_iou_reward": 0.2532842308282852,
      "step": 1464
    },
    {
      "completion_length": 69.546875,
      "epoch": 3.8974700399467377,
      "grad_norm": 8.481136322021484,
      "kl": 0.689453125,
      "learning_rate": 2.4e-08,
      "loss": 0.0276,
      "reward": 0.23788084089756012,
      "reward_std": 0.07825170271098614,
      "reward_std_all": 0.10142708569765091,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6287878751754761,
      "rewards/forensics_tempsamp_iou_reward": 0.23788084089756012,
      "step": 1465
    },
    {
      "completion_length": 83.984375,
      "epoch": 3.900133155792277,
      "grad_norm": 7.296108245849609,
      "kl": 0.642578125,
      "learning_rate": 2.3333333333333334e-08,
      "loss": 0.0257,
      "reward": 0.24199078232049942,
      "reward_std": 0.10136577486991882,
      "reward_std_all": 0.08892293646931648,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9364981949329376,
      "rewards/forensics_tempsamp_iou_reward": 0.24199078232049942,
      "step": 1466
    },
    {
      "completion_length": 68.421875,
      "epoch": 3.902796271637816,
      "grad_norm": 8.13603401184082,
      "kl": 0.65234375,
      "learning_rate": 2.2666666666666668e-08,
      "loss": 0.0261,
      "reward": 0.3154732584953308,
      "reward_std": 0.07942404970526695,
      "reward_std_all": 0.10383393242955208,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9476930797100067,
      "rewards/forensics_tempsamp_iou_reward": 0.3154732584953308,
      "step": 1467
    },
    {
      "completion_length": 62.4375,
      "epoch": 3.9054593874833556,
      "grad_norm": 11.06646728515625,
      "kl": 0.7578125,
      "learning_rate": 2.2e-08,
      "loss": 0.0304,
      "reward": 0.22418934106826782,
      "reward_std": 0.06701544485986233,
      "reward_std_all": 0.07447571866214275,
      "rewards-max/forensics_tempsamp_iou_reward": 0.799346387386322,
      "rewards/forensics_tempsamp_iou_reward": 0.22418934106826782,
      "step": 1468
    },
    {
      "completion_length": 70.0,
      "epoch": 3.9081225033288947,
      "grad_norm": 10.363840103149414,
      "kl": 0.6767578125,
      "learning_rate": 2.1333333333333332e-08,
      "loss": 0.0271,
      "reward": 0.28829824924468994,
      "reward_std": 0.0830892026424408,
      "reward_std_all": 0.101648960262537,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9180911481380463,
      "rewards/forensics_tempsamp_iou_reward": 0.28829824924468994,
      "step": 1469
    },
    {
      "completion_length": 97.59375,
      "epoch": 3.910785619174434,
      "grad_norm": 7.027428150177002,
      "kl": 0.5234375,
      "learning_rate": 2.0666666666666666e-08,
      "loss": 0.0209,
      "reward": 0.27187423408031464,
      "reward_std": 0.09286906011402607,
      "reward_std_all": 0.11975802853703499,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9652162790298462,
      "rewards/forensics_tempsamp_iou_reward": 0.27187423408031464,
      "step": 1470
    },
    {
      "completion_length": 70.046875,
      "epoch": 3.9134487350199736,
      "grad_norm": 7.3919806480407715,
      "kl": 0.73828125,
      "learning_rate": 2e-08,
      "loss": 0.0295,
      "reward": 0.1888095885515213,
      "reward_std": 0.10902810096740723,
      "reward_std_all": 0.11271839216351509,
      "rewards-max/forensics_tempsamp_iou_reward": 0.6724418699741364,
      "rewards/forensics_tempsamp_iou_reward": 0.1888095885515213,
      "step": 1471
    },
    {
      "completion_length": 56.328125,
      "epoch": 3.9161118508655126,
      "grad_norm": 13.789094924926758,
      "kl": 0.48828125,
      "learning_rate": 1.9333333333333334e-08,
      "loss": 0.0196,
      "reward": 0.33725666999816895,
      "reward_std": 0.08510963246226311,
      "reward_std_all": 0.08771350234746933,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9803397059440613,
      "rewards/forensics_tempsamp_iou_reward": 0.33725666999816895,
      "step": 1472
    },
    {
      "completion_length": 85.46875,
      "epoch": 3.918774966711052,
      "grad_norm": 7.806558609008789,
      "kl": 0.7421875,
      "learning_rate": 1.866666666666667e-08,
      "loss": 0.0296,
      "reward": 0.22170411050319672,
      "reward_std": 0.07857520133256912,
      "reward_std_all": 0.1088581457734108,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5781825184822083,
      "rewards/forensics_tempsamp_iou_reward": 0.22170411050319672,
      "step": 1473
    },
    {
      "completion_length": 94.75,
      "epoch": 3.921438082556591,
      "grad_norm": 4.902889728546143,
      "kl": 0.650390625,
      "learning_rate": 1.8e-08,
      "loss": 0.0261,
      "reward": 0.323398619890213,
      "reward_std": 0.11235952377319336,
      "reward_std_all": 0.1460832804441452,
      "rewards-max/forensics_tempsamp_iou_reward": 0.97662353515625,
      "rewards/forensics_tempsamp_iou_reward": 0.323398619890213,
      "step": 1474
    },
    {
      "completion_length": 72.96875,
      "epoch": 3.9241011984021306,
      "grad_norm": 7.394687175750732,
      "kl": 0.806640625,
      "learning_rate": 1.7333333333333333e-08,
      "loss": 0.0323,
      "reward": 0.2750549763441086,
      "reward_std": 0.07120634615421295,
      "reward_std_all": 0.07541938871145248,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9861778020858765,
      "rewards/forensics_tempsamp_iou_reward": 0.2750549763441086,
      "step": 1475
    },
    {
      "completion_length": 60.90625,
      "epoch": 3.92676431424767,
      "grad_norm": 6.379154205322266,
      "kl": 0.638671875,
      "learning_rate": 1.6666666666666667e-08,
      "loss": 0.0256,
      "reward": 0.5465699136257172,
      "reward_std": 0.12688514590263367,
      "reward_std_all": 0.20276273787021637,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9963188469409943,
      "rewards/forensics_tempsamp_iou_reward": 0.5465699136257172,
      "step": 1476
    },
    {
      "completion_length": 54.171875,
      "epoch": 3.929427430093209,
      "grad_norm": 8.537092208862305,
      "kl": 0.6484375,
      "learning_rate": 1.6e-08,
      "loss": 0.026,
      "reward": 0.2663114219903946,
      "reward_std": 0.05330933444201946,
      "reward_std_all": 0.050438059493899345,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9714808762073517,
      "rewards/forensics_tempsamp_iou_reward": 0.2663114219903946,
      "step": 1477
    },
    {
      "completion_length": 118.359375,
      "epoch": 3.932090545938748,
      "grad_norm": 6.766648292541504,
      "kl": 0.75390625,
      "learning_rate": 1.5333333333333332e-08,
      "loss": 0.0302,
      "reward": 0.1444733515381813,
      "reward_std": 0.095237597823143,
      "reward_std_all": 0.09116290882229805,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5085806399583817,
      "rewards/forensics_tempsamp_iou_reward": 0.1444733515381813,
      "step": 1478
    },
    {
      "completion_length": 58.859375,
      "epoch": 3.9347536617842875,
      "grad_norm": 16.617746353149414,
      "kl": 0.98046875,
      "learning_rate": 1.4666666666666666e-08,
      "loss": 0.0393,
      "reward": 0.15224765241146088,
      "reward_std": 0.0938822515308857,
      "reward_std_all": 0.0916118323802948,
      "rewards-max/forensics_tempsamp_iou_reward": 0.48048049211502075,
      "rewards/forensics_tempsamp_iou_reward": 0.15224765241146088,
      "step": 1479
    },
    {
      "completion_length": 104.96875,
      "epoch": 3.937416777629827,
      "grad_norm": 9.765542030334473,
      "kl": 0.923828125,
      "learning_rate": 1.4e-08,
      "loss": 0.0369,
      "reward": 0.2169748693704605,
      "reward_std": 0.08068733662366867,
      "reward_std_all": 0.09287139400839806,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7381205558776855,
      "rewards/forensics_tempsamp_iou_reward": 0.2169748693704605,
      "step": 1480
    },
    {
      "completion_length": 69.359375,
      "epoch": 3.940079893475366,
      "grad_norm": 7.018608093261719,
      "kl": 0.755859375,
      "learning_rate": 1.3333333333333334e-08,
      "loss": 0.0302,
      "reward": 0.14672774821519852,
      "reward_std": 0.0871586874127388,
      "reward_std_all": 0.08751373738050461,
      "rewards-max/forensics_tempsamp_iou_reward": 0.42530570924282074,
      "rewards/forensics_tempsamp_iou_reward": 0.14672774821519852,
      "step": 1481
    },
    {
      "completion_length": 80.59375,
      "epoch": 3.9427430093209055,
      "grad_norm": 10.795493125915527,
      "kl": 0.720703125,
      "learning_rate": 1.2666666666666666e-08,
      "loss": 0.0288,
      "reward": 0.26216650009155273,
      "reward_std": 0.11896692216396332,
      "reward_std_all": 0.12341323867440224,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8057285845279694,
      "rewards/forensics_tempsamp_iou_reward": 0.26216650009155273,
      "step": 1482
    },
    {
      "completion_length": 49.59375,
      "epoch": 3.9454061251664445,
      "grad_norm": 9.892342567443848,
      "kl": 0.4931640625,
      "learning_rate": 1.2e-08,
      "loss": 0.0197,
      "reward": 0.25079404562711716,
      "reward_std": 0.05270187184214592,
      "reward_std_all": 0.07314950786530972,
      "rewards-max/forensics_tempsamp_iou_reward": 0.7535484135150909,
      "rewards/forensics_tempsamp_iou_reward": 0.25079404562711716,
      "step": 1483
    },
    {
      "completion_length": 63.25,
      "epoch": 3.948069241011984,
      "grad_norm": 6.919610977172852,
      "kl": 0.5439453125,
      "learning_rate": 1.1333333333333334e-08,
      "loss": 0.0217,
      "reward": 0.35409966111183167,
      "reward_std": 0.1623065248131752,
      "reward_std_all": 0.1675962619483471,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9721753895282745,
      "rewards/forensics_tempsamp_iou_reward": 0.35409966111183167,
      "step": 1484
    },
    {
      "completion_length": 51.265625,
      "epoch": 3.9507323568575234,
      "grad_norm": 8.02136516571045,
      "kl": 0.5478515625,
      "learning_rate": 1.0666666666666666e-08,
      "loss": 0.0219,
      "reward": 0.33415988087654114,
      "reward_std": 0.05220809020102024,
      "reward_std_all": 0.05780207738280296,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9767783582210541,
      "rewards/forensics_tempsamp_iou_reward": 0.33415988087654114,
      "step": 1485
    },
    {
      "completion_length": 79.578125,
      "epoch": 3.9533954727030625,
      "grad_norm": 6.6024580001831055,
      "kl": 0.5498046875,
      "learning_rate": 1e-08,
      "loss": 0.022,
      "reward": 0.2964852899312973,
      "reward_std": 0.08962912112474442,
      "reward_std_all": 0.09233392402529716,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9508658945560455,
      "rewards/forensics_tempsamp_iou_reward": 0.2964852899312973,
      "step": 1486
    },
    {
      "completion_length": 74.203125,
      "epoch": 3.956058588548602,
      "grad_norm": 7.2511796951293945,
      "kl": 0.859375,
      "learning_rate": 9.333333333333334e-09,
      "loss": 0.0345,
      "reward": 0.3245937079191208,
      "reward_std": 0.11521265655755997,
      "reward_std_all": 0.11502501368522644,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8251011967658997,
      "rewards/forensics_tempsamp_iou_reward": 0.3245937079191208,
      "step": 1487
    },
    {
      "completion_length": 63.40625,
      "epoch": 3.958721704394141,
      "grad_norm": 7.357708930969238,
      "kl": 0.666015625,
      "learning_rate": 8.666666666666667e-09,
      "loss": 0.0267,
      "reward": 0.3273387551307678,
      "reward_std": 0.1215183399617672,
      "reward_std_all": 0.11916732043027878,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9284515976905823,
      "rewards/forensics_tempsamp_iou_reward": 0.3273387551307678,
      "step": 1488
    },
    {
      "completion_length": 63.53125,
      "epoch": 3.9613848202396804,
      "grad_norm": 6.838582515716553,
      "kl": 0.5927734375,
      "learning_rate": 8e-09,
      "loss": 0.0237,
      "reward": 0.3394510746002197,
      "reward_std": 0.10407137870788574,
      "reward_std_all": 0.1085910052061081,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9163791537284851,
      "rewards/forensics_tempsamp_iou_reward": 0.3394510746002197,
      "step": 1489
    },
    {
      "completion_length": 76.03125,
      "epoch": 3.96404793608522,
      "grad_norm": 9.80378246307373,
      "kl": 0.599609375,
      "learning_rate": 7.333333333333333e-09,
      "loss": 0.024,
      "reward": 0.2723238170146942,
      "reward_std": 0.12250593304634094,
      "reward_std_all": 0.13769879937171936,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8562971949577332,
      "rewards/forensics_tempsamp_iou_reward": 0.2723238170146942,
      "step": 1490
    },
    {
      "completion_length": 55.75,
      "epoch": 3.966711051930759,
      "grad_norm": 5.754693031311035,
      "kl": 0.537109375,
      "learning_rate": 6.666666666666667e-09,
      "loss": 0.0214,
      "reward": 0.3160756528377533,
      "reward_std": 0.09190076217055321,
      "reward_std_all": 0.09657830744981766,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9932350218296051,
      "rewards/forensics_tempsamp_iou_reward": 0.3160756528377533,
      "step": 1491
    },
    {
      "completion_length": 58.9375,
      "epoch": 3.9693741677762984,
      "grad_norm": 17.15450668334961,
      "kl": 0.76953125,
      "learning_rate": 6e-09,
      "loss": 0.0308,
      "reward": 0.3248796910047531,
      "reward_std": 0.10322285071015358,
      "reward_std_all": 0.10052924230694771,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9947368502616882,
      "rewards/forensics_tempsamp_iou_reward": 0.3248796910047531,
      "step": 1492
    },
    {
      "completion_length": 57.734375,
      "epoch": 3.9720372836218374,
      "grad_norm": 7.6521315574646,
      "kl": 0.6630859375,
      "learning_rate": 5.333333333333333e-09,
      "loss": 0.0265,
      "reward": 0.21196426451206207,
      "reward_std": 0.0825972631573677,
      "reward_std_all": 0.07570833340287209,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9093061983585358,
      "rewards/forensics_tempsamp_iou_reward": 0.21196426451206207,
      "step": 1493
    },
    {
      "completion_length": 39.046875,
      "epoch": 3.974700399467377,
      "grad_norm": 7.162972927093506,
      "kl": 0.4794921875,
      "learning_rate": 4.666666666666667e-09,
      "loss": 0.0192,
      "reward": 0.31236162781715393,
      "reward_std": 0.05349629558622837,
      "reward_std_all": 0.05383843369781971,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9890438914299011,
      "rewards/forensics_tempsamp_iou_reward": 0.31236162781715393,
      "step": 1494
    },
    {
      "completion_length": 52.5625,
      "epoch": 3.9773635153129163,
      "grad_norm": 11.620828628540039,
      "kl": 0.61328125,
      "learning_rate": 4e-09,
      "loss": 0.0245,
      "reward": 0.27486637234687805,
      "reward_std": 0.11646375805139542,
      "reward_std_all": 0.11453945189714432,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8526972234249115,
      "rewards/forensics_tempsamp_iou_reward": 0.27486637234687805,
      "step": 1495
    },
    {
      "completion_length": 67.9375,
      "epoch": 3.9800266311584553,
      "grad_norm": 6.375306129455566,
      "kl": 0.697265625,
      "learning_rate": 3.3333333333333334e-09,
      "loss": 0.0279,
      "reward": 0.3383793979883194,
      "reward_std": 0.1362774781882763,
      "reward_std_all": 0.12691298872232437,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9969668984413147,
      "rewards/forensics_tempsamp_iou_reward": 0.3383793979883194,
      "step": 1496
    },
    {
      "completion_length": 61.375,
      "epoch": 3.982689747003995,
      "grad_norm": 5.313143253326416,
      "kl": 0.490234375,
      "learning_rate": 2.6666666666666666e-09,
      "loss": 0.0197,
      "reward": 0.32571984827518463,
      "reward_std": 0.09282652661204338,
      "reward_std_all": 0.08491916954517365,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9952455759048462,
      "rewards/forensics_tempsamp_iou_reward": 0.32571984827518463,
      "step": 1497
    },
    {
      "completion_length": 60.515625,
      "epoch": 3.985352862849534,
      "grad_norm": 8.113720893859863,
      "kl": 0.591796875,
      "learning_rate": 2e-09,
      "loss": 0.0237,
      "reward": 0.16112249717116356,
      "reward_std": 0.06843116134405136,
      "reward_std_all": 0.07584844157099724,
      "rewards-max/forensics_tempsamp_iou_reward": 0.5852852761745453,
      "rewards/forensics_tempsamp_iou_reward": 0.16112249717116356,
      "step": 1498
    },
    {
      "completion_length": 62.0,
      "epoch": 3.9880159786950733,
      "grad_norm": 8.745377540588379,
      "kl": 0.478515625,
      "learning_rate": 1.3333333333333333e-09,
      "loss": 0.0191,
      "reward": 0.33214840292930603,
      "reward_std": 0.07808126881718636,
      "reward_std_all": 0.07538432627916336,
      "rewards-max/forensics_tempsamp_iou_reward": 0.9889093041419983,
      "rewards/forensics_tempsamp_iou_reward": 0.33214840292930603,
      "step": 1499
    },
    {
      "completion_length": 84.8125,
      "epoch": 3.9906790945406128,
      "grad_norm": 8.200748443603516,
      "kl": 0.591796875,
      "learning_rate": 6.666666666666666e-10,
      "loss": 0.0237,
      "reward": 0.19289538264274597,
      "reward_std": 0.11974532902240753,
      "reward_std_all": 0.1249452456831932,
      "rewards-max/forensics_tempsamp_iou_reward": 0.8292483687400818,
      "rewards/forensics_tempsamp_iou_reward": 0.19289538264274597,
      "step": 1500
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 1500,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 30,
  "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
}