{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.26763013515321826,
  "eval_steps": 500,
  "global_step": 250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 166.625,
      "epoch": 0.001070520540612873,
      "grad_norm": 1.2984755039215088,
      "kl": 0.0,
      "learning_rate": 5.319148936170213e-08,
      "loss": -0.0,
      "reward": 0.24379686824977398,
      "reward_std": 0.43905802024528384,
      "rewards/correctness_reward_func": 0.1875,
      "rewards/int_reward_func": 0.0546875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.0016093750018626451,
      "step": 1
    },
    {
      "completion_length": 155.09375,
      "epoch": 0.002141041081225746,
      "grad_norm": 5.433530330657959,
      "kl": 0.0,
      "learning_rate": 1.0638297872340426e-07,
      "loss": -0.0,
      "reward": 0.7448437176644802,
      "reward_std": 0.824664918705821,
      "rewards/correctness_reward_func": 0.59375,
      "rewards/int_reward_func": 0.1640625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.012968750204890966,
      "step": 2
    },
    {
      "completion_length": 156.65625,
      "epoch": 0.003211561621838619,
      "grad_norm": 1.5975662469863892,
      "kl": 0.0003120364726783009,
      "learning_rate": 1.5957446808510638e-07,
      "loss": 0.0,
      "reward": 0.5617187460884452,
      "reward_std": 0.6682680626399815,
      "rewards/correctness_reward_func": 0.40625,
      "rewards/int_reward_func": 0.1328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.02265625004656613,
      "step": 3
    },
    {
      "completion_length": 151.328125,
      "epoch": 0.004282082162451492,
      "grad_norm": 1.3767573833465576,
      "kl": 0.00039493154326919466,
      "learning_rate": 2.1276595744680852e-07,
      "loss": 0.0,
      "reward": 0.14101563091389835,
      "reward_std": 0.40191352693364024,
      "rewards/correctness_reward_func": 0.125,
      "rewards/int_reward_func": 0.0546875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.038671874441206455,
      "step": 4
    },
    {
      "completion_length": 143.8125,
      "epoch": 0.005352602703064365,
      "grad_norm": 1.7029815912246704,
      "kl": 0.0003161150925734546,
      "learning_rate": 2.6595744680851066e-07,
      "loss": 0.0,
      "reward": 0.6667499775066972,
      "reward_std": 0.943404046818614,
      "rewards/correctness_reward_func": 0.5,
      "rewards/int_reward_func": 0.1328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.03393749863607809,
      "step": 5
    },
    {
      "completion_length": 146.5625,
      "epoch": 0.006423123243677238,
      "grad_norm": 4.5992960929870605,
      "kl": 0.0014150730130495504,
      "learning_rate": 3.1914893617021275e-07,
      "loss": 0.0001,
      "reward": 0.6142812594771385,
      "reward_std": 0.8342159832827747,
      "rewards/correctness_reward_func": 0.46875,
      "rewards/int_reward_func": 0.1328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.01271874993108213,
      "step": 6
    },
    {
      "completion_length": 163.9375,
      "epoch": 0.007493643784290111,
      "grad_norm": 7.240438461303711,
      "kl": 0.0018134960264433175,
      "learning_rate": 3.723404255319149e-07,
      "loss": 0.0001,
      "reward": 0.24757812730967999,
      "reward_std": 0.5592395211569965,
      "rewards/correctness_reward_func": 0.1875,
      "rewards/int_reward_func": 0.0703125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.010234375484287739,
      "step": 7
    },
    {
      "completion_length": 151.84375,
      "epoch": 0.008564164324902984,
      "grad_norm": 4.826539993286133,
      "kl": 0.0011592731952987378,
      "learning_rate": 4.2553191489361704e-07,
      "loss": 0.0,
      "reward": 0.24482813104987144,
      "reward_std": 0.4538201582618058,
      "rewards/correctness_reward_func": 0.15625,
      "rewards/int_reward_func": 0.046875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.04170312359929085,
      "step": 8
    },
    {
      "completion_length": 148.828125,
      "epoch": 0.009634684865515858,
      "grad_norm": 4.097943305969238,
      "kl": 0.0009884996707114624,
      "learning_rate": 4.787234042553192e-07,
      "loss": 0.0,
      "reward": 0.6986562423408031,
      "reward_std": 1.089426226913929,
      "rewards/correctness_reward_func": 0.5625,
      "rewards/int_reward_func": 0.1171875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.018968748801853508,
      "step": 9
    },
    {
      "completion_length": 160.3125,
      "epoch": 0.01070520540612873,
      "grad_norm": 3.2978594303131104,
      "kl": 0.0008196280577976722,
      "learning_rate": 5.319148936170213e-07,
      "loss": 0.0,
      "reward": 0.48834376223385334,
      "reward_std": 0.790836479049176,
      "rewards/correctness_reward_func": 0.375,
      "rewards/int_reward_func": 0.1171875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.003843750571832061,
      "step": 10
    },
    {
      "completion_length": 146.96875,
      "epoch": 0.011775725946741603,
      "grad_norm": 3.537848472595215,
      "kl": 0.0007083387099555694,
      "learning_rate": 5.851063829787235e-07,
      "loss": 0.0,
      "reward": 0.5805312437005341,
      "reward_std": 0.7569947894662619,
      "rewards/correctness_reward_func": 0.46875,
      "rewards/int_reward_func": 0.1328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.021031250711530447,
      "step": 11
    },
    {
      "completion_length": 145.890625,
      "epoch": 0.012846246487354477,
      "grad_norm": 3.8045260906219482,
      "kl": 0.0013589818336186,
      "learning_rate": 6.382978723404255e-07,
      "loss": 0.0001,
      "reward": 0.46695311937946826,
      "reward_std": 0.6352905407547951,
      "rewards/correctness_reward_func": 0.3125,
      "rewards/int_reward_func": 0.109375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.04507812508381903,
      "step": 12
    },
    {
      "completion_length": 146.5,
      "epoch": 0.013916767027967349,
      "grad_norm": 1.7494301795959473,
      "kl": 0.000400338125473354,
      "learning_rate": 6.914893617021278e-07,
      "loss": 0.0,
      "reward": 0.5322968787513673,
      "reward_std": 0.7736401874572039,
      "rewards/correctness_reward_func": 0.40625,
      "rewards/int_reward_func": 0.1171875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.00885937490966171,
      "step": 13
    },
    {
      "completion_length": 133.671875,
      "epoch": 0.014987287568580221,
      "grad_norm": 1.8054606914520264,
      "kl": 0.00036212212944519706,
      "learning_rate": 7.446808510638298e-07,
      "loss": 0.0,
      "reward": 0.6521718641743064,
      "reward_std": 1.0104734068736434,
      "rewards/correctness_reward_func": 0.53125,
      "rewards/int_reward_func": 0.1328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.011890625639352947,
      "step": 14
    },
    {
      "completion_length": 130.0,
      "epoch": 0.016057808109193095,
      "grad_norm": 1.5193248987197876,
      "kl": 0.00036491416904027574,
      "learning_rate": 7.97872340425532e-07,
      "loss": 0.0,
      "reward": 0.5991093763150275,
      "reward_std": 0.8120721196755767,
      "rewards/correctness_reward_func": 0.46875,
      "rewards/int_reward_func": 0.125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.005359375616535544,
      "step": 15
    },
    {
      "completion_length": 162.640625,
      "epoch": 0.017128328649805968,
      "grad_norm": 4.938427448272705,
      "kl": 0.0014603480958612636,
      "learning_rate": 8.510638297872341e-07,
      "loss": 0.0001,
      "reward": 0.2470156280323863,
      "reward_std": 0.5583461234346032,
      "rewards/correctness_reward_func": 0.1875,
      "rewards/int_reward_func": 0.0703125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.010796874179504812,
      "step": 16
    },
    {
      "completion_length": 165.796875,
      "epoch": 0.01819884919041884,
      "grad_norm": 3.923428535461426,
      "kl": 0.001025654159093392,
      "learning_rate": 9.042553191489363e-07,
      "loss": 0.0,
      "reward": 0.5893437387421727,
      "reward_std": 0.8829143429175019,
      "rewards/correctness_reward_func": 0.46875,
      "rewards/int_reward_func": 0.125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.004406251944601536,
      "step": 17
    },
    {
      "completion_length": 152.359375,
      "epoch": 0.019269369731031716,
      "grad_norm": 5.248744487762451,
      "kl": 0.0020197606609144714,
      "learning_rate": 9.574468085106384e-07,
      "loss": 0.0001,
      "reward": 0.46303125098347664,
      "reward_std": 0.9115365371108055,
      "rewards/correctness_reward_func": 0.34375,
      "rewards/int_reward_func": 0.1015625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.017718749470077455,
      "step": 18
    },
    {
      "completion_length": 151.359375,
      "epoch": 0.020339890271644588,
      "grad_norm": 1.7978895902633667,
      "kl": 0.00032554956487729214,
      "learning_rate": 1.0106382978723404e-06,
      "loss": 0.0,
      "reward": 0.48937500442843884,
      "reward_std": 0.7031663246452808,
      "rewards/correctness_reward_func": 0.375,
      "rewards/int_reward_func": 0.109375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.005000000121071935,
      "step": 19
    },
    {
      "completion_length": 175.90625,
      "epoch": 0.02141041081225746,
      "grad_norm": 2.1679224967956543,
      "kl": 0.0004576210667437408,
      "learning_rate": 1.0638297872340427e-06,
      "loss": 0.0,
      "reward": 0.44935936853289604,
      "reward_std": 0.6872247559949756,
      "rewards/correctness_reward_func": 0.34375,
      "rewards/int_reward_func": 0.09375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.011859375052154064,
      "step": 20
    },
    {
      "completion_length": 167.265625,
      "epoch": 0.022480931352870333,
      "grad_norm": 1.475581407546997,
      "kl": 0.0007880991906858981,
      "learning_rate": 1.1170212765957447e-06,
      "loss": 0.0,
      "reward": 0.28507812274619937,
      "reward_std": 0.5523091573268175,
      "rewards/correctness_reward_func": 0.21875,
      "rewards/int_reward_func": 0.0625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.003828124259598553,
      "step": 21
    },
    {
      "completion_length": 153.296875,
      "epoch": 0.023551451893483205,
      "grad_norm": 1.6419923305511475,
      "kl": 0.00039223546627908945,
      "learning_rate": 1.170212765957447e-06,
      "loss": 0.0,
      "reward": 0.6624062322080135,
      "reward_std": 0.9240671265870333,
      "rewards/correctness_reward_func": 0.5,
      "rewards/int_reward_func": 0.1328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.0295937517657876,
      "step": 22
    },
    {
      "completion_length": 143.515625,
      "epoch": 0.02462197243409608,
      "grad_norm": 1.3294110298156738,
      "kl": 0.00030555322518921457,
      "learning_rate": 1.223404255319149e-06,
      "loss": 0.0,
      "reward": 0.26720312132965773,
      "reward_std": 0.4894332850817591,
      "rewards/correctness_reward_func": 0.1875,
      "rewards/int_reward_func": 0.0546875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.025015624007210135,
      "step": 23
    },
    {
      "completion_length": 148.3125,
      "epoch": 0.025692492974708953,
      "grad_norm": 6.915380954742432,
      "kl": 0.0062694076787011,
      "learning_rate": 1.276595744680851e-06,
      "loss": 0.0003,
      "reward": 0.28068749560043216,
      "reward_std": 0.5439753192476928,
      "rewards/correctness_reward_func": 0.1875,
      "rewards/int_reward_func": 0.0625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.030687499791383743,
      "step": 24
    },
    {
      "completion_length": 145.421875,
      "epoch": 0.026763013515321826,
      "grad_norm": 1.5796961784362793,
      "kl": 0.0008160970010067103,
      "learning_rate": 1.3297872340425533e-06,
      "loss": 0.0,
      "reward": 0.604515643324703,
      "reward_std": 0.6846362175419927,
      "rewards/correctness_reward_func": 0.4375,
      "rewards/int_reward_func": 0.1171875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.049828124698251486,
      "step": 25
    },
    {
      "completion_length": 144.21875,
      "epoch": 0.027833534055934698,
      "grad_norm": 3.3064963817596436,
      "kl": 0.0009157936146948487,
      "learning_rate": 1.3829787234042555e-06,
      "loss": 0.0,
      "reward": 0.5731093874201179,
      "reward_std": 0.8484273846261203,
      "rewards/correctness_reward_func": 0.4375,
      "rewards/int_reward_func": 0.1171875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.018421874032355845,
      "step": 26
    },
    {
      "completion_length": 135.921875,
      "epoch": 0.02890405459654757,
      "grad_norm": 1.8949307203292847,
      "kl": 0.0004942502673657145,
      "learning_rate": 1.4361702127659578e-06,
      "loss": 0.0,
      "reward": 0.6309375101700425,
      "reward_std": 0.9806207492947578,
      "rewards/correctness_reward_func": 0.5,
      "rewards/int_reward_func": 0.125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.005937499925494194,
      "step": 27
    },
    {
      "completion_length": 153.59375,
      "epoch": 0.029974575137160443,
      "grad_norm": 1.7615420818328857,
      "kl": 0.0004919220991723705,
      "learning_rate": 1.4893617021276596e-06,
      "loss": 0.0,
      "reward": 0.1835937526775524,
      "reward_std": 0.45627398509532213,
      "rewards/correctness_reward_func": 0.15625,
      "rewards/int_reward_func": 0.0546875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.027343749883584678,
      "step": 28
    },
    {
      "completion_length": 164.640625,
      "epoch": 0.03104509567777332,
      "grad_norm": 4.3193230628967285,
      "kl": 0.0013858377351425588,
      "learning_rate": 1.5425531914893618e-06,
      "loss": 0.0001,
      "reward": 0.40034375386312604,
      "reward_std": 0.7546388749033213,
      "rewards/correctness_reward_func": 0.28125,
      "rewards/int_reward_func": 0.109375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.009718750137835741,
      "step": 29
    },
    {
      "completion_length": 128.234375,
      "epoch": 0.03211561621838619,
      "grad_norm": 5.113959312438965,
      "kl": 0.0021859680928173475,
      "learning_rate": 1.595744680851064e-06,
      "loss": 0.0001,
      "reward": 0.8881718653719872,
      "reward_std": 0.9410315058194101,
      "rewards/correctness_reward_func": 0.6875,
      "rewards/int_reward_func": 0.1796875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.02098437474342063,
      "step": 30
    },
    {
      "completion_length": 140.8125,
      "epoch": 0.03318613675899906,
      "grad_norm": 7.337815284729004,
      "kl": 0.00254826245145523,
      "learning_rate": 1.648936170212766e-06,
      "loss": 0.0001,
      "reward": 0.7087812423706055,
      "reward_std": 0.9508242532610893,
      "rewards/correctness_reward_func": 0.53125,
      "rewards/int_reward_func": 0.15625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.021281250286847353,
      "step": 31
    },
    {
      "completion_length": 141.0625,
      "epoch": 0.034256657299611935,
      "grad_norm": 4.055324554443359,
      "kl": 0.0016690000156813767,
      "learning_rate": 1.7021276595744682e-06,
      "loss": 0.0001,
      "reward": 0.7299062572419643,
      "reward_std": 0.8444140013307333,
      "rewards/correctness_reward_func": 0.5625,
      "rewards/int_reward_func": 0.1328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.03459375072270632,
      "step": 32
    },
    {
      "completion_length": 154.296875,
      "epoch": 0.03532717784022481,
      "grad_norm": 3.565863847732544,
      "kl": 0.0011733200299204327,
      "learning_rate": 1.7553191489361704e-06,
      "loss": 0.0,
      "reward": 0.7152343707857653,
      "reward_std": 0.986182201653719,
      "rewards/correctness_reward_func": 0.5625,
      "rewards/int_reward_func": 0.1640625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.011328124441206455,
      "step": 33
    },
    {
      "completion_length": 156.484375,
      "epoch": 0.03639769838083768,
      "grad_norm": 2.782845973968506,
      "kl": 0.001061345017660642,
      "learning_rate": 1.8085106382978727e-06,
      "loss": 0.0,
      "reward": 0.4085781138855964,
      "reward_std": 0.6946883676573634,
      "rewards/correctness_reward_func": 0.3125,
      "rewards/int_reward_func": 0.09375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.0023281261092051864,
      "step": 34
    },
    {
      "completion_length": 166.203125,
      "epoch": 0.03746821892145055,
      "grad_norm": 1.6768676042556763,
      "kl": 0.0008204600453609601,
      "learning_rate": 1.8617021276595745e-06,
      "loss": 0.0,
      "reward": 0.5197031321004033,
      "reward_std": 0.8639188781380653,
      "rewards/correctness_reward_func": 0.40625,
      "rewards/int_reward_func": 0.1171875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.0037343755830079317,
      "step": 35
    },
    {
      "completion_length": 168.828125,
      "epoch": 0.03853873946206343,
      "grad_norm": 7.892724990844727,
      "kl": 0.0029609855264425278,
      "learning_rate": 1.9148936170212767e-06,
      "loss": 0.0001,
      "reward": 0.23489062942098826,
      "reward_std": 0.356358939781785,
      "rewards/correctness_reward_func": 0.15625,
      "rewards/int_reward_func": 0.046875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.031765625230036676,
      "step": 36
    },
    {
      "completion_length": 123.53125,
      "epoch": 0.039609260002676304,
      "grad_norm": 2.5262649059295654,
      "kl": 0.0010402118496131152,
      "learning_rate": 1.968085106382979e-06,
      "loss": 0.0,
      "reward": 0.9631250270176679,
      "reward_std": 1.0183926988393068,
      "rewards/correctness_reward_func": 0.71875,
      "rewards/int_reward_func": 0.1796875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.06468750000931323,
      "step": 37
    },
    {
      "completion_length": 159.46875,
      "epoch": 0.040679780543289176,
      "grad_norm": 1.3111544847488403,
      "kl": 0.0011480498651508242,
      "learning_rate": 2.021276595744681e-06,
      "loss": 0.0,
      "reward": 0.4472187543287873,
      "reward_std": 0.6874936055392027,
      "rewards/correctness_reward_func": 0.34375,
      "rewards/int_reward_func": 0.09375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.009718748508021235,
      "step": 38
    },
    {
      "completion_length": 142.25,
      "epoch": 0.04175030108390205,
      "grad_norm": 2.2276878356933594,
      "kl": 0.0014690053576487117,
      "learning_rate": 2.074468085106383e-06,
      "loss": 0.0001,
      "reward": 1.0481719109229743,
      "reward_std": 0.7983668614178896,
      "rewards/correctness_reward_func": 0.8125,
      "rewards/int_reward_func": 0.1796875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.05598437529988587,
      "step": 39
    },
    {
      "completion_length": 149.71875,
      "epoch": 0.04282082162451492,
      "grad_norm": 3.2188539505004883,
      "kl": 0.002757413443760015,
      "learning_rate": 2.1276595744680853e-06,
      "loss": 0.0001,
      "reward": 0.8439687644131482,
      "reward_std": 0.9366709599271417,
      "rewards/correctness_reward_func": 0.625,
      "rewards/int_reward_func": 0.171875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.04709374951198697,
      "step": 40
    },
    {
      "completion_length": 153.75,
      "epoch": 0.04389134216512779,
      "grad_norm": 1.609779953956604,
      "kl": 0.002587423972727265,
      "learning_rate": 2.1808510638297876e-06,
      "loss": 0.0001,
      "reward": 0.9616250339895487,
      "reward_std": 0.8783023115247488,
      "rewards/correctness_reward_func": 0.75,
      "rewards/int_reward_func": 0.1953125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.01631249929778278,
      "step": 41
    },
    {
      "completion_length": 156.15625,
      "epoch": 0.044961862705740666,
      "grad_norm": 5.789142608642578,
      "kl": 0.0045223182532936335,
      "learning_rate": 2.2340425531914894e-06,
      "loss": 0.0002,
      "reward": 0.8380624754354358,
      "reward_std": 0.881235895678401,
      "rewards/correctness_reward_func": 0.625,
      "rewards/int_reward_func": 0.1796875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.03337499825283885,
      "step": 42
    },
    {
      "completion_length": 161.390625,
      "epoch": 0.04603238324635354,
      "grad_norm": 5.589737892150879,
      "kl": 0.005504744782228954,
      "learning_rate": 2.2872340425531916e-06,
      "loss": 0.0002,
      "reward": 0.8749218666926026,
      "reward_std": 0.9216371476650238,
      "rewards/correctness_reward_func": 0.6875,
      "rewards/int_reward_func": 0.1796875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.007734375540167093,
      "step": 43
    },
    {
      "completion_length": 161.796875,
      "epoch": 0.04710290378696641,
      "grad_norm": 5.604761600494385,
      "kl": 0.0046821657015243545,
      "learning_rate": 2.340425531914894e-06,
      "loss": 0.0002,
      "reward": 0.2252656314522028,
      "reward_std": 0.47946364153176546,
      "rewards/correctness_reward_func": 0.15625,
      "rewards/int_reward_func": 0.0546875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.014328125165775418,
      "step": 44
    },
    {
      "completion_length": 154.453125,
      "epoch": 0.04817342432757928,
      "grad_norm": 9.506386756896973,
      "kl": 0.007164878101320937,
      "learning_rate": 2.393617021276596e-06,
      "loss": 0.0003,
      "reward": 0.6195624829269946,
      "reward_std": 1.0045473407953978,
      "rewards/correctness_reward_func": 0.46875,
      "rewards/int_reward_func": 0.1328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.01799999945797026,
      "step": 45
    },
    {
      "completion_length": 140.390625,
      "epoch": 0.04924394486819216,
      "grad_norm": 3.766388416290283,
      "kl": 0.00619677483337,
      "learning_rate": 2.446808510638298e-06,
      "loss": 0.0002,
      "reward": 0.9423750173300505,
      "reward_std": 0.5746848955750465,
      "rewards/correctness_reward_func": 0.65625,
      "rewards/int_reward_func": 0.203125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.08299999847076833,
      "step": 46
    },
    {
      "completion_length": 143.09375,
      "epoch": 0.050314465408805034,
      "grad_norm": 4.4466705322265625,
      "kl": 0.009431202546693385,
      "learning_rate": 2.5e-06,
      "loss": 0.0004,
      "reward": 0.9380937227979302,
      "reward_std": 0.658873830921948,
      "rewards/correctness_reward_func": 0.6875,
      "rewards/int_reward_func": 0.1796875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07090624794363976,
      "step": 47
    },
    {
      "completion_length": 165.0625,
      "epoch": 0.051384985949417906,
      "grad_norm": 1.3459303379058838,
      "kl": 0.00600922666490078,
      "learning_rate": 2.553191489361702e-06,
      "loss": 0.0002,
      "reward": 0.6878281269455329,
      "reward_std": 0.9325386872515082,
      "rewards/correctness_reward_func": 0.5,
      "rewards/int_reward_func": 0.15625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.03157812531571835,
      "step": 48
    },
    {
      "completion_length": 134.421875,
      "epoch": 0.05245550649003078,
      "grad_norm": 1.628507375717163,
      "kl": 0.008930853742640465,
      "learning_rate": 2.6063829787234047e-06,
      "loss": 0.0004,
      "reward": 0.8599374926416203,
      "reward_std": 0.9755922555923462,
      "rewards/correctness_reward_func": 0.625,
      "rewards/int_reward_func": 0.1796875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.05525000113993883,
      "step": 49
    },
    {
      "completion_length": 149.5625,
      "epoch": 0.05352602703064365,
      "grad_norm": 1.1767226457595825,
      "kl": 0.012399342958815396,
      "learning_rate": 2.6595744680851065e-06,
      "loss": 0.0005,
      "reward": 1.0091875102370977,
      "reward_std": 0.8712345249950886,
      "rewards/correctness_reward_func": 0.71875,
      "rewards/int_reward_func": 0.203125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.08731249999254942,
      "step": 50
    },
    {
      "completion_length": 169.078125,
      "epoch": 0.05459654757125652,
      "grad_norm": 2.5901050567626953,
      "kl": 0.007969769008923322,
      "learning_rate": 2.7127659574468084e-06,
      "loss": 0.0003,
      "reward": 0.9503125064074993,
      "reward_std": 0.9690856691449881,
      "rewards/correctness_reward_func": 0.6875,
      "rewards/int_reward_func": 0.234375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.028437498025596142,
      "step": 51
    },
    {
      "completion_length": 166.390625,
      "epoch": 0.055667068111869396,
      "grad_norm": 3.764841079711914,
      "kl": 0.011693944863509387,
      "learning_rate": 2.765957446808511e-06,
      "loss": 0.0005,
      "reward": 0.7889687474817038,
      "reward_std": 0.8161248974502087,
      "rewards/correctness_reward_func": 0.5625,
      "rewards/int_reward_func": 0.1875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.03896875097416341,
      "step": 52
    },
    {
      "completion_length": 158.3125,
      "epoch": 0.05673758865248227,
      "grad_norm": 4.34719705581665,
      "kl": 0.00852247714647092,
      "learning_rate": 2.819148936170213e-06,
      "loss": 0.0003,
      "reward": 0.5854531275108457,
      "reward_std": 0.579142062459141,
      "rewards/correctness_reward_func": 0.4375,
      "rewards/int_reward_func": 0.1484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.00048437435179948807,
      "step": 53
    },
    {
      "completion_length": 149.15625,
      "epoch": 0.05780810919309514,
      "grad_norm": 1.3242722749710083,
      "kl": 0.006340013263979927,
      "learning_rate": 2.8723404255319155e-06,
      "loss": 0.0003,
      "reward": 0.7202812694013119,
      "reward_std": 0.6025513117201626,
      "rewards/correctness_reward_func": 0.5,
      "rewards/int_reward_func": 0.140625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07965625170618296,
      "step": 54
    },
    {
      "completion_length": 131.0,
      "epoch": 0.05887862973370801,
      "grad_norm": 1.2887805700302124,
      "kl": 0.005301086028339341,
      "learning_rate": 2.9255319148936174e-06,
      "loss": 0.0002,
      "reward": 1.1997031308710575,
      "reward_std": 0.8454109290614724,
      "rewards/correctness_reward_func": 0.875,
      "rewards/int_reward_func": 0.21875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.10595312505029142,
      "step": 55
    },
    {
      "completion_length": 132.828125,
      "epoch": 0.059949150274320885,
      "grad_norm": 3.085810899734497,
      "kl": 0.03685568018408958,
      "learning_rate": 2.978723404255319e-06,
      "loss": 0.0015,
      "reward": 1.32792192324996,
      "reward_std": 0.8808077229186893,
      "rewards/correctness_reward_func": 1.0,
      "rewards/int_reward_func": 0.265625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.062296870397403836,
      "step": 56
    },
    {
      "completion_length": 169.5,
      "epoch": 0.061019670814933764,
      "grad_norm": 3.218595266342163,
      "kl": 0.019670582871185616,
      "learning_rate": 3.031914893617022e-06,
      "loss": 0.0008,
      "reward": 1.0494843795895576,
      "reward_std": 0.8383767995983362,
      "rewards/correctness_reward_func": 0.78125,
      "rewards/int_reward_func": 0.1953125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07292187376879156,
      "step": 57
    },
    {
      "completion_length": 164.71875,
      "epoch": 0.06209019135554664,
      "grad_norm": 1.189056634902954,
      "kl": 0.010589714744128287,
      "learning_rate": 3.0851063829787237e-06,
      "loss": 0.0004,
      "reward": 0.9995937808416784,
      "reward_std": 0.8763523772358894,
      "rewards/correctness_reward_func": 0.75,
      "rewards/int_reward_func": 0.234375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.015218749409541488,
      "step": 58
    },
    {
      "completion_length": 173.109375,
      "epoch": 0.0631607118961595,
      "grad_norm": 2.9426801204681396,
      "kl": 0.010116680678038392,
      "learning_rate": 3.1382978723404255e-06,
      "loss": 0.0004,
      "reward": 0.5431718821637332,
      "reward_std": 0.5429769204929471,
      "rewards/correctness_reward_func": 0.40625,
      "rewards/int_reward_func": 0.1015625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.03535937680862844,
      "step": 59
    },
    {
      "completion_length": 148.328125,
      "epoch": 0.06423123243677238,
      "grad_norm": 1.2039525508880615,
      "kl": 0.012681124440860003,
      "learning_rate": 3.191489361702128e-06,
      "loss": 0.0005,
      "reward": 0.7227343516424298,
      "reward_std": 0.7870303755626082,
      "rewards/correctness_reward_func": 0.53125,
      "rewards/int_reward_func": 0.125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.06648437259718776,
      "step": 60
    },
    {
      "completion_length": 125.5625,
      "epoch": 0.06530175297738525,
      "grad_norm": 1.2423548698425293,
      "kl": 0.007948026963276789,
      "learning_rate": 3.24468085106383e-06,
      "loss": 0.0003,
      "reward": 1.3104218831285834,
      "reward_std": 0.6878003547899425,
      "rewards/correctness_reward_func": 0.9375,
      "rewards/int_reward_func": 0.25,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.12292187649291009,
      "step": 61
    },
    {
      "completion_length": 152.296875,
      "epoch": 0.06637227351799813,
      "grad_norm": 5.324446678161621,
      "kl": 0.043227474874584004,
      "learning_rate": 3.297872340425532e-06,
      "loss": 0.0017,
      "reward": 1.0643124831840396,
      "reward_std": 0.9667724259197712,
      "rewards/correctness_reward_func": 0.78125,
      "rewards/int_reward_func": 0.2109375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07212500204332173,
      "step": 62
    },
    {
      "completion_length": 157.328125,
      "epoch": 0.067442794058611,
      "grad_norm": 1.2027546167373657,
      "kl": 0.004988896660506725,
      "learning_rate": 3.3510638297872345e-06,
      "loss": 0.0002,
      "reward": 1.1987031551543623,
      "reward_std": 0.8610715009272099,
      "rewards/correctness_reward_func": 0.90625,
      "rewards/int_reward_func": 0.2578125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.034640624886378646,
      "step": 63
    },
    {
      "completion_length": 163.9375,
      "epoch": 0.06851331459922387,
      "grad_norm": 1.126978874206543,
      "kl": 0.004329273069743067,
      "learning_rate": 3.4042553191489363e-06,
      "loss": 0.0002,
      "reward": 0.8524375010747463,
      "reward_std": 0.742443086579442,
      "rewards/correctness_reward_func": 0.65625,
      "rewards/int_reward_func": 0.171875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.024312500143423676,
      "step": 64
    },
    {
      "completion_length": 135.765625,
      "epoch": 0.06958383513983675,
      "grad_norm": 4.760307788848877,
      "kl": 0.0423761896090582,
      "learning_rate": 3.457446808510639e-06,
      "loss": 0.0017,
      "reward": 1.4641093388199806,
      "reward_std": 1.0102775804698467,
      "rewards/correctness_reward_func": 1.0625,
      "rewards/int_reward_func": 0.296875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.1047343765385449,
      "step": 65
    },
    {
      "completion_length": 155.796875,
      "epoch": 0.07065435568044962,
      "grad_norm": 3.1110665798187256,
      "kl": 0.061538238427601755,
      "learning_rate": 3.510638297872341e-06,
      "loss": 0.0025,
      "reward": 1.033312514424324,
      "reward_std": 0.8865859052166343,
      "rewards/correctness_reward_func": 0.78125,
      "rewards/int_reward_func": 0.203125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.048937500920146704,
      "step": 66
    },
    {
      "completion_length": 164.578125,
      "epoch": 0.0717248762210625,
      "grad_norm": 1.2862604856491089,
      "kl": 0.0052650388242909685,
      "learning_rate": 3.5638297872340426e-06,
      "loss": 0.0002,
      "reward": 0.7733906293287873,
      "reward_std": 0.8027388863265514,
      "rewards/correctness_reward_func": 0.59375,
      "rewards/int_reward_func": 0.1640625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.015578125370666385,
      "step": 67
    },
    {
      "completion_length": 163.75,
      "epoch": 0.07279539676167536,
      "grad_norm": 6.033888816833496,
      "kl": 0.07815390304313041,
      "learning_rate": 3.6170212765957453e-06,
      "loss": 0.0031,
      "reward": 1.1917500102426857,
      "reward_std": 0.7897173650562763,
      "rewards/correctness_reward_func": 0.90625,
      "rewards/int_reward_func": 0.265625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.019875000230968,
      "step": 68
    },
    {
      "completion_length": 149.40625,
      "epoch": 0.07386591730228824,
      "grad_norm": 1.29887056350708,
      "kl": 0.01987961767008528,
      "learning_rate": 3.670212765957447e-06,
      "loss": 0.0008,
      "reward": 1.1079531812574714,
      "reward_std": 0.7181853111833334,
      "rewards/correctness_reward_func": 0.78125,
      "rewards/int_reward_func": 0.265625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.06107812421396375,
      "step": 69
    },
    {
      "completion_length": 141.859375,
      "epoch": 0.0749364378429011,
      "grad_norm": 5.329657077789307,
      "kl": 0.0800003606127575,
      "learning_rate": 3.723404255319149e-06,
      "loss": 0.0032,
      "reward": 0.9977656248956919,
      "reward_std": 0.5876570995897055,
      "rewards/correctness_reward_func": 0.6875,
      "rewards/int_reward_func": 0.21875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.09151562419719994,
      "step": 70
    },
    {
      "completion_length": 138.5625,
      "epoch": 0.07600695838351398,
      "grad_norm": 2.7627108097076416,
      "kl": 0.03701730686589144,
      "learning_rate": 3.7765957446808516e-06,
      "loss": 0.0015,
      "reward": 1.2351874904707074,
      "reward_std": 0.8267962019890547,
      "rewards/correctness_reward_func": 0.90625,
      "rewards/int_reward_func": 0.25,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07893750071525574,
      "step": 71
    },
    {
      "completion_length": 133.984375,
      "epoch": 0.07707747892412686,
      "grad_norm": 1.414104700088501,
      "kl": 0.016362678608857095,
      "learning_rate": 3.8297872340425535e-06,
      "loss": 0.0007,
      "reward": 1.827203094959259,
      "reward_std": 1.0086707267910242,
      "rewards/correctness_reward_func": 1.375,
      "rewards/int_reward_func": 0.359375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.0928281235974282,
      "step": 72
    },
    {
      "completion_length": 148.0625,
      "epoch": 0.07814799946473973,
      "grad_norm": 1.3797897100448608,
      "kl": 0.010796019807457924,
      "learning_rate": 3.882978723404256e-06,
      "loss": 0.0004,
      "reward": 1.1499999817460775,
      "reward_std": 0.9028994599357247,
      "rewards/correctness_reward_func": 0.84375,
      "rewards/int_reward_func": 0.2109375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.09531250316649675,
      "step": 73
    },
    {
      "completion_length": 155.828125,
      "epoch": 0.07921852000535261,
      "grad_norm": 1.2884279489517212,
      "kl": 0.017199350346345454,
      "learning_rate": 3.936170212765958e-06,
      "loss": 0.0007,
      "reward": 1.0700937574729323,
      "reward_std": 1.0012164115905762,
      "rewards/correctness_reward_func": 0.8125,
      "rewards/int_reward_func": 0.234375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.023218751302920282,
      "step": 74
    },
    {
      "completion_length": 146.96875,
      "epoch": 0.08028904054596547,
      "grad_norm": 3.5342090129852295,
      "kl": 0.05702902490156703,
      "learning_rate": 3.98936170212766e-06,
      "loss": 0.0023,
      "reward": 1.2074843887239695,
      "reward_std": 0.8127174219116569,
      "rewards/correctness_reward_func": 0.875,
      "rewards/int_reward_func": 0.2578125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07467187196016312,
      "step": 75
    },
    {
      "completion_length": 135.140625,
      "epoch": 0.08135956108657835,
      "grad_norm": 1.7845228910446167,
      "kl": 0.01959521723620128,
      "learning_rate": 4.042553191489362e-06,
      "loss": 0.0008,
      "reward": 1.4087968692183495,
      "reward_std": 0.9582029562443495,
      "rewards/correctness_reward_func": 1.03125,
      "rewards/int_reward_func": 0.2734375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.10410937643609941,
      "step": 76
    },
    {
      "completion_length": 126.125,
      "epoch": 0.08243008162719122,
      "grad_norm": 1.3039053678512573,
      "kl": 0.008768495463300496,
      "learning_rate": 4.095744680851064e-06,
      "loss": 0.0004,
      "reward": 1.3187343887984753,
      "reward_std": 0.5690027270466089,
      "rewards/correctness_reward_func": 0.9375,
      "rewards/int_reward_func": 0.2578125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.12342187319882214,
      "step": 77
    },
    {
      "completion_length": 148.6875,
      "epoch": 0.0835006021678041,
      "grad_norm": 3.3175690174102783,
      "kl": 0.014638990571256727,
      "learning_rate": 4.148936170212766e-06,
      "loss": 0.0006,
      "reward": 1.4877656111493707,
      "reward_std": 0.8957763649523258,
      "rewards/correctness_reward_func": 1.09375,
      "rewards/int_reward_func": 0.2890625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.10495312558487058,
      "step": 78
    },
    {
      "completion_length": 171.0625,
      "epoch": 0.08457112270841696,
      "grad_norm": 5.574616432189941,
      "kl": 0.01483242801623419,
      "learning_rate": 4.202127659574468e-06,
      "loss": 0.0006,
      "reward": 1.0053124725818634,
      "reward_std": 1.1163944154977798,
      "rewards/correctness_reward_func": 0.75,
      "rewards/int_reward_func": 0.2265625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.02874999982304871,
      "step": 79
    },
    {
      "completion_length": 150.5625,
      "epoch": 0.08564164324902984,
      "grad_norm": 3.4894111156463623,
      "kl": 0.020942480681696907,
      "learning_rate": 4.255319148936171e-06,
      "loss": 0.0008,
      "reward": 1.081656239926815,
      "reward_std": 0.8474766900762916,
      "rewards/correctness_reward_func": 0.75,
      "rewards/int_reward_func": 0.234375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.09728124877437949,
      "step": 80
    },
    {
      "completion_length": 145.8125,
      "epoch": 0.08671216378964271,
      "grad_norm": 4.479030609130859,
      "kl": 0.02067101007560268,
      "learning_rate": 4.308510638297873e-06,
      "loss": 0.0008,
      "reward": 0.9469531225040555,
      "reward_std": 0.5945024443790317,
      "rewards/correctness_reward_func": 0.625,
      "rewards/int_reward_func": 0.2421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07976562529802322,
      "step": 81
    },
    {
      "completion_length": 142.609375,
      "epoch": 0.08778268433025559,
      "grad_norm": 9.429178237915039,
      "kl": 0.032599265803582966,
      "learning_rate": 4.361702127659575e-06,
      "loss": 0.0013,
      "reward": 0.7288906406611204,
      "reward_std": 0.8307479582726955,
      "rewards/correctness_reward_func": 0.46875,
      "rewards/int_reward_func": 0.171875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.08826562575995922,
      "step": 82
    },
    {
      "completion_length": 143.84375,
      "epoch": 0.08885320487086847,
      "grad_norm": 1.5900259017944336,
      "kl": 0.013988179998705164,
      "learning_rate": 4.414893617021277e-06,
      "loss": 0.0006,
      "reward": 1.591531228274107,
      "reward_std": 0.8249969203025103,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.3203125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.021218748996034265,
      "step": 83
    },
    {
      "completion_length": 135.8125,
      "epoch": 0.08992372541148133,
      "grad_norm": 4.648512363433838,
      "kl": 0.06424052006332204,
      "learning_rate": 4.468085106382979e-06,
      "loss": 0.0026,
      "reward": 1.4651249905582517,
      "reward_std": 0.8705566665157676,
      "rewards/correctness_reward_func": 1.0625,
      "rewards/int_reward_func": 0.2890625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.11356249963864684,
      "step": 84
    },
    {
      "completion_length": 123.90625,
      "epoch": 0.09099424595209421,
      "grad_norm": 6.136663913726807,
      "kl": 0.02576264040544629,
      "learning_rate": 4.521276595744681e-06,
      "loss": 0.001,
      "reward": 1.7882343754172325,
      "reward_std": 0.7715246099978685,
      "rewards/correctness_reward_func": 1.3125,
      "rewards/int_reward_func": 0.3671875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.10854687169194221,
      "step": 85
    },
    {
      "completion_length": 132.65625,
      "epoch": 0.09206476649270708,
      "grad_norm": 2.7052152156829834,
      "kl": 0.021323165216017514,
      "learning_rate": 4.574468085106383e-06,
      "loss": 0.0009,
      "reward": 1.2060781214386225,
      "reward_std": 0.74014887586236,
      "rewards/correctness_reward_func": 0.84375,
      "rewards/int_reward_func": 0.25,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.11232812539674342,
      "step": 86
    },
    {
      "completion_length": 142.3125,
      "epoch": 0.09313528703331996,
      "grad_norm": 6.834721088409424,
      "kl": 0.15714708802988753,
      "learning_rate": 4.6276595744680855e-06,
      "loss": 0.0063,
      "reward": 1.4907656013965607,
      "reward_std": 1.1851906776428223,
      "rewards/correctness_reward_func": 1.09375,
      "rewards/int_reward_func": 0.3203125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07670312328264117,
      "step": 87
    },
    {
      "completion_length": 141.625,
      "epoch": 0.09420580757393282,
      "grad_norm": 1.5215015411376953,
      "kl": 0.012561204843223095,
      "learning_rate": 4.680851063829788e-06,
      "loss": 0.0005,
      "reward": 0.973562479019165,
      "reward_std": 0.8407629178836942,
      "rewards/correctness_reward_func": 0.6875,
      "rewards/int_reward_func": 0.2109375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07512499787844718,
      "step": 88
    },
    {
      "completion_length": 131.109375,
      "epoch": 0.0952763281145457,
      "grad_norm": 4.186502456665039,
      "kl": 0.029811605345457792,
      "learning_rate": 4.73404255319149e-06,
      "loss": 0.0012,
      "reward": 1.571968775242567,
      "reward_std": 0.8088337788358331,
      "rewards/correctness_reward_func": 1.125,
      "rewards/int_reward_func": 0.3125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.13446875009685755,
      "step": 89
    },
    {
      "completion_length": 137.5625,
      "epoch": 0.09634684865515857,
      "grad_norm": 3.539090633392334,
      "kl": 0.025141435849945992,
      "learning_rate": 4.787234042553192e-06,
      "loss": 0.001,
      "reward": 1.1439843773841858,
      "reward_std": 1.1783022359013557,
      "rewards/correctness_reward_func": 0.8125,
      "rewards/int_reward_func": 0.2421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.08929687412455678,
      "step": 90
    },
    {
      "completion_length": 142.0,
      "epoch": 0.09741736919577144,
      "grad_norm": 3.9599854946136475,
      "kl": 0.03133453679038212,
      "learning_rate": 4.840425531914894e-06,
      "loss": 0.0013,
      "reward": 0.9123750082217157,
      "reward_std": 0.8763793092221022,
      "rewards/correctness_reward_func": 0.65625,
      "rewards/int_reward_func": 0.21875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.037375001003965735,
      "step": 91
    },
    {
      "completion_length": 127.0625,
      "epoch": 0.09848788973638432,
      "grad_norm": 5.762697219848633,
      "kl": 0.02441024547442794,
      "learning_rate": 4.893617021276596e-06,
      "loss": 0.001,
      "reward": 1.4343437626957893,
      "reward_std": 1.0151535924524069,
      "rewards/correctness_reward_func": 1.0,
      "rewards/int_reward_func": 0.328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.10621875151991844,
      "step": 92
    },
    {
      "completion_length": 128.421875,
      "epoch": 0.09955841027699719,
      "grad_norm": 1.7006865739822388,
      "kl": 0.014295783417765051,
      "learning_rate": 4.946808510638298e-06,
      "loss": 0.0006,
      "reward": 1.4152031522244215,
      "reward_std": 1.0377220567315817,
      "rewards/correctness_reward_func": 1.03125,
      "rewards/int_reward_func": 0.265625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.11832812381908298,
      "step": 93
    },
    {
      "completion_length": 126.703125,
      "epoch": 0.10062893081761007,
      "grad_norm": 2.6587624549865723,
      "kl": 0.022590334410779178,
      "learning_rate": 5e-06,
      "loss": 0.0009,
      "reward": 1.0790781378746033,
      "reward_std": 0.8811032259836793,
      "rewards/correctness_reward_func": 0.75,
      "rewards/int_reward_func": 0.203125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.12595312716439366,
      "step": 94
    },
    {
      "completion_length": 134.25,
      "epoch": 0.10169945135822293,
      "grad_norm": 5.803914546966553,
      "kl": 0.0693736044340767,
      "learning_rate": 4.999982515602153e-06,
      "loss": 0.0028,
      "reward": 1.6232343390583992,
      "reward_std": 1.171187661588192,
      "rewards/correctness_reward_func": 1.1875,
      "rewards/int_reward_func": 0.34375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.09198437177110463,
      "step": 95
    },
    {
      "completion_length": 119.09375,
      "epoch": 0.10276997189883581,
      "grad_norm": 2.6233108043670654,
      "kl": 0.025776030379347503,
      "learning_rate": 4.999930062653175e-06,
      "loss": 0.001,
      "reward": 1.4469374530017376,
      "reward_std": 0.717207751236856,
      "rewards/correctness_reward_func": 1.03125,
      "rewards/int_reward_func": 0.2890625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.12662499537691474,
      "step": 96
    },
    {
      "completion_length": 135.625,
      "epoch": 0.10384049243944868,
      "grad_norm": 1.4691489934921265,
      "kl": 0.018308754893951118,
      "learning_rate": 4.999842641886752e-06,
      "loss": 0.0007,
      "reward": 1.7408281043171883,
      "reward_std": 0.6604799125343561,
      "rewards/correctness_reward_func": 1.3125,
      "rewards/int_reward_func": 0.328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.10020312923006713,
      "step": 97
    },
    {
      "completion_length": 142.53125,
      "epoch": 0.10491101298006156,
      "grad_norm": 3.7322754859924316,
      "kl": 0.031659536180086434,
      "learning_rate": 4.999720254525684e-06,
      "loss": 0.0013,
      "reward": 1.3402031436562538,
      "reward_std": 1.0035525700077415,
      "rewards/correctness_reward_func": 0.96875,
      "rewards/int_reward_func": 0.3203125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.051140623865649104,
      "step": 98
    },
    {
      "completion_length": 136.859375,
      "epoch": 0.10598153352067442,
      "grad_norm": 1.445330023765564,
      "kl": 0.016985120251774788,
      "learning_rate": 4.999562902281866e-06,
      "loss": 0.0007,
      "reward": 1.4283437356352806,
      "reward_std": 0.9263063753023744,
      "rewards/correctness_reward_func": 1.0,
      "rewards/int_reward_func": 0.296875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.13146874867379665,
      "step": 99
    },
    {
      "completion_length": 145.234375,
      "epoch": 0.1070520540612873,
      "grad_norm": 4.298182010650635,
      "kl": 0.03539641568204388,
      "learning_rate": 4.999370587356267e-06,
      "loss": 0.0014,
      "reward": 1.5638906005769968,
      "reward_std": 0.9018815001472831,
      "rewards/correctness_reward_func": 1.15625,
      "rewards/int_reward_func": 0.328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.07951562479138374,
      "step": 100
    },
    {
      "completion_length": 126.46875,
      "epoch": 0.10812257460190017,
      "grad_norm": 1.7051573991775513,
      "kl": 0.0210904503474012,
      "learning_rate": 4.999143312438893e-06,
      "loss": 0.0008,
      "reward": 2.032781273126602,
      "reward_std": 0.882479477673769,
      "rewards/correctness_reward_func": 1.5,
      "rewards/int_reward_func": 0.4296875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.10309374984353781,
      "step": 101
    },
    {
      "completion_length": 138.4375,
      "epoch": 0.10919309514251305,
      "grad_norm": 5.384690284729004,
      "kl": 0.05343873624224216,
      "learning_rate": 4.998881080708759e-06,
      "loss": 0.0021,
      "reward": 1.7520312666893005,
      "reward_std": 1.0768068991601467,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.3203125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.08796874759718776,
      "step": 102
    },
    {
      "completion_length": 121.265625,
      "epoch": 0.11026361568312593,
      "grad_norm": 3.4453506469726562,
      "kl": 0.054925739066675305,
      "learning_rate": 4.998583895833834e-06,
      "loss": 0.0022,
      "reward": 1.6850781589746475,
      "reward_std": 0.6663676341995597,
      "rewards/correctness_reward_func": 1.125,
      "rewards/int_reward_func": 0.4296875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.1303906268440187,
      "step": 103
    },
    {
      "completion_length": 114.71875,
      "epoch": 0.11133413622373879,
      "grad_norm": 1.5185428857803345,
      "kl": 0.022636244888417423,
      "learning_rate": 4.998251761970997e-06,
      "loss": 0.0009,
      "reward": 1.7328437007963657,
      "reward_std": 0.7276732774917036,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.34375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.13909374829381704,
      "step": 104
    },
    {
      "completion_length": 114.359375,
      "epoch": 0.11240465676435167,
      "grad_norm": 4.903259754180908,
      "kl": 0.0455300398170948,
      "learning_rate": 4.997884683765977e-06,
      "loss": 0.0018,
      "reward": 2.070781234651804,
      "reward_std": 0.5790006909519434,
      "rewards/correctness_reward_func": 1.5,
      "rewards/int_reward_func": 0.4296875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.14109374955296516,
      "step": 105
    },
    {
      "completion_length": 117.578125,
      "epoch": 0.11347517730496454,
      "grad_norm": 7.719193458557129,
      "kl": 0.07494375784881413,
      "learning_rate": 4.997482666353287e-06,
      "loss": 0.003,
      "reward": 1.8749061971902847,
      "reward_std": 0.762595918495208,
      "rewards/correctness_reward_func": 1.3125,
      "rewards/int_reward_func": 0.390625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.1717812498100102,
      "step": 106
    },
    {
      "completion_length": 132.71875,
      "epoch": 0.11454569784557742,
      "grad_norm": 1.9178704023361206,
      "kl": 0.026816099416464567,
      "learning_rate": 4.997045715356153e-06,
      "loss": 0.0011,
      "reward": 1.4049999862909317,
      "reward_std": 1.0492134541273117,
      "rewards/correctness_reward_func": 0.9375,
      "rewards/int_reward_func": 0.328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.13937499932944775,
      "step": 107
    },
    {
      "completion_length": 121.21875,
      "epoch": 0.11561621838619028,
      "grad_norm": 3.1541106700897217,
      "kl": 0.04801671905443072,
      "learning_rate": 4.9965738368864345e-06,
      "loss": 0.0019,
      "reward": 1.7343593537807465,
      "reward_std": 0.9352071397006512,
      "rewards/correctness_reward_func": 1.21875,
      "rewards/int_reward_func": 0.359375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.1562343705445528,
      "step": 108
    },
    {
      "completion_length": 134.40625,
      "epoch": 0.11668673892680316,
      "grad_norm": 3.4600815773010254,
      "kl": 0.07537143386434764,
      "learning_rate": 4.996067037544542e-06,
      "loss": 0.003,
      "reward": 1.6357812024652958,
      "reward_std": 0.7928643207997084,
      "rewards/correctness_reward_func": 1.15625,
      "rewards/int_reward_func": 0.3359375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.14359374903142452,
      "step": 109
    },
    {
      "completion_length": 121.25,
      "epoch": 0.11775725946741603,
      "grad_norm": 1.8081263303756714,
      "kl": 0.03252584161236882,
      "learning_rate": 4.995525324419338e-06,
      "loss": 0.0013,
      "reward": 1.6038124561309814,
      "reward_std": 0.9315616749227047,
      "rewards/correctness_reward_func": 1.09375,
      "rewards/int_reward_func": 0.3671875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.14287500269711018,
      "step": 110
    },
    {
      "completion_length": 113.734375,
      "epoch": 0.1188277800080289,
      "grad_norm": 3.427846670150757,
      "kl": 0.029214507434517145,
      "learning_rate": 4.994948705088047e-06,
      "loss": 0.0012,
      "reward": 1.7096093818545341,
      "reward_std": 0.9403769830241799,
      "rewards/correctness_reward_func": 1.1875,
      "rewards/int_reward_func": 0.375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.14710937440395355,
      "step": 111
    },
    {
      "completion_length": 138.140625,
      "epoch": 0.11989830054864177,
      "grad_norm": 5.425636291503906,
      "kl": 0.11221261869650334,
      "learning_rate": 4.99433718761614e-06,
      "loss": 0.0045,
      "reward": 1.4035156145691872,
      "reward_std": 1.1017278581857681,
      "rewards/correctness_reward_func": 0.90625,
      "rewards/int_reward_func": 0.3515625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.14570312481373549,
      "step": 112
    },
    {
      "completion_length": 128.828125,
      "epoch": 0.12096882108925465,
      "grad_norm": 9.655882835388184,
      "kl": 0.14679548889398575,
      "learning_rate": 4.993690780557232e-06,
      "loss": 0.0059,
      "reward": 1.5967968963086605,
      "reward_std": 0.9262449182569981,
      "rewards/correctness_reward_func": 1.09375,
      "rewards/int_reward_func": 0.3359375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.16710936930030584,
      "step": 113
    },
    {
      "completion_length": 114.3125,
      "epoch": 0.12203934162986753,
      "grad_norm": 2.9527218341827393,
      "kl": 0.03957346314564347,
      "learning_rate": 4.993009492952951e-06,
      "loss": 0.0016,
      "reward": 1.7140468880534172,
      "reward_std": 0.9286471158266068,
      "rewards/correctness_reward_func": 1.125,
      "rewards/int_reward_func": 0.421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.16717187454923987,
      "step": 114
    },
    {
      "completion_length": 122.515625,
      "epoch": 0.1231098621704804,
      "grad_norm": 4.749152660369873,
      "kl": 0.05118492292240262,
      "learning_rate": 4.992293334332821e-06,
      "loss": 0.002,
      "reward": 1.9886562526226044,
      "reward_std": 0.9419979229569435,
      "rewards/correctness_reward_func": 1.4375,
      "rewards/int_reward_func": 0.40625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.14490625634789467,
      "step": 115
    },
    {
      "completion_length": 124.171875,
      "epoch": 0.12418038271109327,
      "grad_norm": 4.01917028427124,
      "kl": 0.10346966434735805,
      "learning_rate": 4.991542314714122e-06,
      "loss": 0.0041,
      "reward": 1.837890625,
      "reward_std": 0.912613769993186,
      "rewards/correctness_reward_func": 1.28125,
      "rewards/int_reward_func": 0.40625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.15039062406867743,
      "step": 116
    },
    {
      "completion_length": 124.375,
      "epoch": 0.12525090325170615,
      "grad_norm": 3.7853267192840576,
      "kl": 0.07023024489171803,
      "learning_rate": 4.990756444601757e-06,
      "loss": 0.0028,
      "reward": 1.6455781627446413,
      "reward_std": 0.7009308515116572,
      "rewards/correctness_reward_func": 1.125,
      "rewards/int_reward_func": 0.3671875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.15339062316343188,
      "step": 117
    },
    {
      "completion_length": 136.359375,
      "epoch": 0.126321423792319,
      "grad_norm": 2.0698678493499756,
      "kl": 0.03698924113996327,
      "learning_rate": 4.989935734988098e-06,
      "loss": 0.0015,
      "reward": 1.8074062652885914,
      "reward_std": 0.8670060317963362,
      "rewards/correctness_reward_func": 1.28125,
      "rewards/int_reward_func": 0.3671875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.1589687503874302,
      "step": 118
    },
    {
      "completion_length": 148.15625,
      "epoch": 0.12739194433293188,
      "grad_norm": 7.199705123901367,
      "kl": 0.180443427991122,
      "learning_rate": 4.989080197352834e-06,
      "loss": 0.0072,
      "reward": 1.179109364748001,
      "reward_std": 0.7882043793797493,
      "rewards/correctness_reward_func": 0.71875,
      "rewards/int_reward_func": 0.328125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.13223437825217843,
      "step": 119
    },
    {
      "completion_length": 122.46875,
      "epoch": 0.12846246487354476,
      "grad_norm": 1.6841520071029663,
      "kl": 0.0505296983756125,
      "learning_rate": 4.9881898436628165e-06,
      "loss": 0.002,
      "reward": 1.9381406530737877,
      "reward_std": 0.8331695850938559,
      "rewards/correctness_reward_func": 1.375,
      "rewards/int_reward_func": 0.390625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.17251562420278788,
      "step": 120
    },
    {
      "completion_length": 108.1875,
      "epoch": 0.12953298541415764,
      "grad_norm": 1.8260003328323364,
      "kl": 0.059317339677363634,
      "learning_rate": 4.987264686371881e-06,
      "loss": 0.0024,
      "reward": 2.02584370970726,
      "reward_std": 1.0399171710014343,
      "rewards/correctness_reward_func": 1.40625,
      "rewards/int_reward_func": 0.421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.19771874882280827,
      "step": 121
    },
    {
      "completion_length": 95.46875,
      "epoch": 0.1306035059547705,
      "grad_norm": 2.1317758560180664,
      "kl": 0.053682942409068346,
      "learning_rate": 4.986304738420684e-06,
      "loss": 0.0021,
      "reward": 2.1616249792277813,
      "reward_std": 0.6366997184231877,
      "rewards/correctness_reward_func": 1.5,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.22412499878555536,
      "step": 122
    },
    {
      "completion_length": 118.453125,
      "epoch": 0.13167402649538337,
      "grad_norm": 1.487423062324524,
      "kl": 0.04084368539042771,
      "learning_rate": 4.985310013236514e-06,
      "loss": 0.0016,
      "reward": 2.4017499536275864,
      "reward_std": 0.6733962241560221,
      "rewards/correctness_reward_func": 1.71875,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2064374964684248,
      "step": 123
    },
    {
      "completion_length": 136.75,
      "epoch": 0.13274454703599625,
      "grad_norm": 6.275110244750977,
      "kl": 0.10213979217223823,
      "learning_rate": 4.984280524733107e-06,
      "loss": 0.0041,
      "reward": 1.5195625126361847,
      "reward_std": 0.8954427968710661,
      "rewards/correctness_reward_func": 1.0625,
      "rewards/int_reward_func": 0.3125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.1445625051856041,
      "step": 124
    },
    {
      "completion_length": 107.84375,
      "epoch": 0.13381506757660913,
      "grad_norm": 4.696693420410156,
      "kl": 0.1205500855576247,
      "learning_rate": 4.983216287310453e-06,
      "loss": 0.0048,
      "reward": 1.828781247138977,
      "reward_std": 0.92250463552773,
      "rewards/correctness_reward_func": 1.1875,
      "rewards/int_reward_func": 0.4296875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.21159374713897705,
      "step": 125
    },
    {
      "completion_length": 90.984375,
      "epoch": 0.134885588117222,
      "grad_norm": 5.726349830627441,
      "kl": 0.14590927632525563,
      "learning_rate": 4.982117315854594e-06,
      "loss": 0.0058,
      "reward": 2.218953087925911,
      "reward_std": 0.855751893715933,
      "rewards/correctness_reward_func": 1.53125,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.24239062331616879,
      "step": 126
    },
    {
      "completion_length": 102.28125,
      "epoch": 0.13595610865783486,
      "grad_norm": 2.1168274879455566,
      "kl": 0.07060433947481215,
      "learning_rate": 4.980983625737411e-06,
      "loss": 0.0028,
      "reward": 2.1402343213558197,
      "reward_std": 0.6787437000311911,
      "rewards/correctness_reward_func": 1.46875,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.20273437071591616,
      "step": 127
    },
    {
      "completion_length": 122.0625,
      "epoch": 0.13702662919844774,
      "grad_norm": 9.184843063354492,
      "kl": 0.15047278022393584,
      "learning_rate": 4.9798152328164165e-06,
      "loss": 0.006,
      "reward": 1.5844999551773071,
      "reward_std": 1.025202952325344,
      "rewards/correctness_reward_func": 1.03125,
      "rewards/int_reward_func": 0.359375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.19387499801814556,
      "step": 128
    },
    {
      "completion_length": 107.21875,
      "epoch": 0.13809714973906062,
      "grad_norm": 1.8967880010604858,
      "kl": 0.04243561811745167,
      "learning_rate": 4.978612153434527e-06,
      "loss": 0.0017,
      "reward": 2.1196718886494637,
      "reward_std": 0.5547986216843128,
      "rewards/correctness_reward_func": 1.46875,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2134218756109476,
      "step": 129
    },
    {
      "completion_length": 108.03125,
      "epoch": 0.1391676702796735,
      "grad_norm": 2.011303186416626,
      "kl": 0.04303696344140917,
      "learning_rate": 4.977374404419838e-06,
      "loss": 0.0017,
      "reward": 2.1730000376701355,
      "reward_std": 0.9301177933812141,
      "rewards/correctness_reward_func": 1.53125,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2042500004172325,
      "step": 130
    },
    {
      "completion_length": 110.453125,
      "epoch": 0.14023819082028635,
      "grad_norm": 7.903467178344727,
      "kl": 0.22786898026242852,
      "learning_rate": 4.9761020030853854e-06,
      "loss": 0.0091,
      "reward": 2.0016875714063644,
      "reward_std": 0.8600476859137416,
      "rewards/correctness_reward_func": 1.40625,
      "rewards/int_reward_func": 0.3984375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.19699999503791332,
      "step": 131
    },
    {
      "completion_length": 132.625,
      "epoch": 0.14130871136089923,
      "grad_norm": 5.2613444328308105,
      "kl": 0.23378165811300278,
      "learning_rate": 4.9747949672289075e-06,
      "loss": 0.0094,
      "reward": 1.4954218715429306,
      "reward_std": 0.9026085883378983,
      "rewards/correctness_reward_func": 1.03125,
      "rewards/int_reward_func": 0.3203125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.14385937619954348,
      "step": 132
    },
    {
      "completion_length": 101.125,
      "epoch": 0.1423792319015121,
      "grad_norm": 4.111429214477539,
      "kl": 0.18476470839232206,
      "learning_rate": 4.973453315132592e-06,
      "loss": 0.0074,
      "reward": 2.419406235218048,
      "reward_std": 0.6811097683385015,
      "rewards/correctness_reward_func": 1.75,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.19284374825656414,
      "step": 133
    },
    {
      "completion_length": 110.0625,
      "epoch": 0.143449752442125,
      "grad_norm": 4.763584613800049,
      "kl": 0.1902949649374932,
      "learning_rate": 4.9720770655628216e-06,
      "loss": 0.0076,
      "reward": 2.049671910703182,
      "reward_std": 0.8413272872567177,
      "rewards/correctness_reward_func": 1.40625,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.19029687531292439,
      "step": 134
    },
    {
      "completion_length": 95.578125,
      "epoch": 0.14452027298273787,
      "grad_norm": 4.345425128936768,
      "kl": 0.18916460033506155,
      "learning_rate": 4.970666237769913e-06,
      "loss": 0.0076,
      "reward": 2.191656231880188,
      "reward_std": 0.7033369969576597,
      "rewards/correctness_reward_func": 1.53125,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.20728124678134918,
      "step": 135
    },
    {
      "completion_length": 84.078125,
      "epoch": 0.14559079352335072,
      "grad_norm": 2.940378427505493,
      "kl": 0.06369929504580796,
      "learning_rate": 4.9692208514878445e-06,
      "loss": 0.0025,
      "reward": 2.3128594160079956,
      "reward_std": 0.6257542409002781,
      "rewards/correctness_reward_func": 1.625,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.21910937316715717,
      "step": 136
    },
    {
      "completion_length": 120.328125,
      "epoch": 0.1466613140639636,
      "grad_norm": 7.51654052734375,
      "kl": 0.3002074658870697,
      "learning_rate": 4.967740926933985e-06,
      "loss": 0.012,
      "reward": 1.7032031267881393,
      "reward_std": 1.0220621526241302,
      "rewards/correctness_reward_func": 1.15625,
      "rewards/int_reward_func": 0.3828125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.16414062399417162,
      "step": 137
    },
    {
      "completion_length": 114.046875,
      "epoch": 0.14773183460457648,
      "grad_norm": 2.956787347793579,
      "kl": 0.10229182336479425,
      "learning_rate": 4.966226484808804e-06,
      "loss": 0.0041,
      "reward": 1.6846249997615814,
      "reward_std": 0.8771754652261734,
      "rewards/correctness_reward_func": 1.09375,
      "rewards/int_reward_func": 0.3984375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.19243750255554914,
      "step": 138
    },
    {
      "completion_length": 102.34375,
      "epoch": 0.14880235514518936,
      "grad_norm": 4.8385114669799805,
      "kl": 0.1648537963628769,
      "learning_rate": 4.96467754629559e-06,
      "loss": 0.0066,
      "reward": 1.9405781105160713,
      "reward_std": 0.6661778870038688,
      "rewards/correctness_reward_func": 1.3125,
      "rewards/int_reward_func": 0.40625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.22182812727987766,
      "step": 139
    },
    {
      "completion_length": 99.921875,
      "epoch": 0.1498728756858022,
      "grad_norm": 2.6203041076660156,
      "kl": 0.06277278368361294,
      "learning_rate": 4.963094133060148e-06,
      "loss": 0.0025,
      "reward": 2.1328750401735306,
      "reward_std": 0.6620223973877728,
      "rewards/correctness_reward_func": 1.4375,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.23443749826401472,
      "step": 140
    },
    {
      "completion_length": 92.15625,
      "epoch": 0.1509433962264151,
      "grad_norm": 2.787095069885254,
      "kl": 0.07504080841317773,
      "learning_rate": 4.961476267250501e-06,
      "loss": 0.003,
      "reward": 2.2870156168937683,
      "reward_std": 0.6934415455907583,
      "rewards/correctness_reward_func": 1.59375,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.23232812341302633,
      "step": 141
    },
    {
      "completion_length": 111.0,
      "epoch": 0.15201391676702797,
      "grad_norm": 2.570060968399048,
      "kl": 0.06025985535234213,
      "learning_rate": 4.959823971496575e-06,
      "loss": 0.0024,
      "reward": 1.9751719227060676,
      "reward_std": 0.7296689655631781,
      "rewards/correctness_reward_func": 1.375,
      "rewards/int_reward_func": 0.40625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.1939218717161566,
      "step": 142
    },
    {
      "completion_length": 127.40625,
      "epoch": 0.15308443730764085,
      "grad_norm": 4.903911590576172,
      "kl": 0.20708202896639705,
      "learning_rate": 4.958137268909887e-06,
      "loss": 0.0083,
      "reward": 2.063531205058098,
      "reward_std": 0.8510163221508265,
      "rewards/correctness_reward_func": 1.5,
      "rewards/int_reward_func": 0.421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.14165624976158142,
      "step": 143
    },
    {
      "completion_length": 98.5625,
      "epoch": 0.15415495784825373,
      "grad_norm": 4.191147804260254,
      "kl": 0.2446250948123634,
      "learning_rate": 4.9564161830832214e-06,
      "loss": 0.0098,
      "reward": 2.337281256914139,
      "reward_std": 0.6763627836480737,
      "rewards/correctness_reward_func": 1.65625,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2200937452726066,
      "step": 144
    },
    {
      "completion_length": 88.984375,
      "epoch": 0.15522547838886658,
      "grad_norm": 168.51283264160156,
      "kl": 0.34985177870839834,
      "learning_rate": 4.954660738090297e-06,
      "loss": 0.014,
      "reward": 1.6290156617760658,
      "reward_std": 0.7621745709329844,
      "rewards/correctness_reward_func": 1.03125,
      "rewards/int_reward_func": 0.375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.222765626385808,
      "step": 145
    },
    {
      "completion_length": 116.078125,
      "epoch": 0.15629599892947946,
      "grad_norm": 4.605787754058838,
      "kl": 0.20393797848373652,
      "learning_rate": 4.9528709584854316e-06,
      "loss": 0.0082,
      "reward": 1.5738593488931656,
      "reward_std": 1.115996390581131,
      "rewards/correctness_reward_func": 1.0,
      "rewards/int_reward_func": 0.3984375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.17542187124490738,
      "step": 146
    },
    {
      "completion_length": 89.328125,
      "epoch": 0.15736651947009234,
      "grad_norm": 4.029605865478516,
      "kl": 0.21852776128798723,
      "learning_rate": 4.951046869303202e-06,
      "loss": 0.0087,
      "reward": 1.9777500182390213,
      "reward_std": 0.9872381817549467,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.40625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.22774999774992466,
      "step": 147
    },
    {
      "completion_length": 106.75,
      "epoch": 0.15843704001070522,
      "grad_norm": 3.108353853225708,
      "kl": 0.07371893431991339,
      "learning_rate": 4.949188496058089e-06,
      "loss": 0.0029,
      "reward": 1.7372031211853027,
      "reward_std": 0.9112571626901627,
      "rewards/correctness_reward_func": 1.09375,
      "rewards/int_reward_func": 0.421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2215781332924962,
      "step": 148
    },
    {
      "completion_length": 78.40625,
      "epoch": 0.15950756055131807,
      "grad_norm": 5.034030914306641,
      "kl": 0.11514789052307606,
      "learning_rate": 4.947295864744121e-06,
      "loss": 0.0046,
      "reward": 2.055265612900257,
      "reward_std": 0.6963230553083122,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.258390624076128,
      "step": 149
    },
    {
      "completion_length": 90.484375,
      "epoch": 0.16057808109193095,
      "grad_norm": 2.6227409839630127,
      "kl": 0.09901809925213456,
      "learning_rate": 4.9453690018345144e-06,
      "loss": 0.004,
      "reward": 2.031171888113022,
      "reward_std": 0.9270219663158059,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2499218750745058,
      "step": 150
    },
    {
      "completion_length": 111.640625,
      "epoch": 0.16164860163254383,
      "grad_norm": 5.403820037841797,
      "kl": 0.24172887252643704,
      "learning_rate": 4.943407934281298e-06,
      "loss": 0.0097,
      "reward": 1.693390630185604,
      "reward_std": 1.0351360142230988,
      "rewards/correctness_reward_func": 1.0625,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.1699531227350235,
      "step": 151
    },
    {
      "completion_length": 92.890625,
      "epoch": 0.1627191221731567,
      "grad_norm": 2.174766778945923,
      "kl": 0.07736781658604741,
      "learning_rate": 4.941412689514941e-06,
      "loss": 0.0031,
      "reward": 2.1249531507492065,
      "reward_std": 0.82035240996629,
      "rewards/correctness_reward_func": 1.40625,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2499531265348196,
      "step": 152
    },
    {
      "completion_length": 96.546875,
      "epoch": 0.16378964271376958,
      "grad_norm": 2.039672374725342,
      "kl": 0.07323360512964427,
      "learning_rate": 4.939383295443966e-06,
      "loss": 0.0029,
      "reward": 2.026875004172325,
      "reward_std": 0.7493576873093843,
      "rewards/correctness_reward_func": 1.375,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.20656250044703484,
      "step": 153
    },
    {
      "completion_length": 106.59375,
      "epoch": 0.16486016325438244,
      "grad_norm": 2.2819831371307373,
      "kl": 0.07374695758335292,
      "learning_rate": 4.937319780454559e-06,
      "loss": 0.0029,
      "reward": 1.7915781140327454,
      "reward_std": 0.8505431758239865,
      "rewards/correctness_reward_func": 1.15625,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.1900156200863421,
      "step": 154
    },
    {
      "completion_length": 91.90625,
      "epoch": 0.16593068379499532,
      "grad_norm": 5.251499176025391,
      "kl": 0.40140265179798007,
      "learning_rate": 4.9352221734101745e-06,
      "loss": 0.0161,
      "reward": 2.1450937539339066,
      "reward_std": 0.7235295535065234,
      "rewards/correctness_reward_func": 1.46875,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2388437483459711,
      "step": 155
    },
    {
      "completion_length": 91.375,
      "epoch": 0.1670012043356082,
      "grad_norm": 2.6729063987731934,
      "kl": 0.06791439699009061,
      "learning_rate": 4.933090503651129e-06,
      "loss": 0.0027,
      "reward": 2.2278437092900276,
      "reward_std": 0.7469283854588866,
      "rewards/correctness_reward_func": 1.5625,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.22003125306218863,
      "step": 156
    },
    {
      "completion_length": 83.53125,
      "epoch": 0.16807172487622107,
      "grad_norm": 2.890995979309082,
      "kl": 0.08941763360053301,
      "learning_rate": 4.930924800994192e-06,
      "loss": 0.0036,
      "reward": 2.040843792259693,
      "reward_std": 0.6482276869937778,
      "rewards/correctness_reward_func": 1.3125,
      "rewards/int_reward_func": 0.4921875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.23615625128149986,
      "step": 157
    },
    {
      "completion_length": 103.4375,
      "epoch": 0.16914224541683393,
      "grad_norm": 4.847292900085449,
      "kl": 0.25062092347070575,
      "learning_rate": 4.9287250957321685e-06,
      "loss": 0.01,
      "reward": 1.8447500094771385,
      "reward_std": 0.8658850640058517,
      "rewards/correctness_reward_func": 1.1875,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.21193750761449337,
      "step": 158
    },
    {
      "completion_length": 87.78125,
      "epoch": 0.1702127659574468,
      "grad_norm": 2.486091136932373,
      "kl": 0.10755344619974494,
      "learning_rate": 4.9264914186334775e-06,
      "loss": 0.0043,
      "reward": 2.14860936999321,
      "reward_std": 0.7705556647852063,
      "rewards/correctness_reward_func": 1.4375,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.25017187278717756,
      "step": 159
    },
    {
      "completion_length": 74.8125,
      "epoch": 0.17128328649805968,
      "grad_norm": 2.6566174030303955,
      "kl": 0.09692100062966347,
      "learning_rate": 4.924223800941718e-06,
      "loss": 0.0039,
      "reward": 2.179875001311302,
      "reward_std": 0.647944641765207,
      "rewards/correctness_reward_func": 1.40625,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.28924999572336674,
      "step": 160
    },
    {
      "completion_length": 112.390625,
      "epoch": 0.17235380703867256,
      "grad_norm": 7.983546257019043,
      "kl": 0.34650124446488917,
      "learning_rate": 4.921922274375232e-06,
      "loss": 0.0139,
      "reward": 1.7768593654036522,
      "reward_std": 0.6629852540791035,
      "rewards/correctness_reward_func": 1.1875,
      "rewards/int_reward_func": 0.4140625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.17529687471687794,
      "step": 161
    },
    {
      "completion_length": 81.609375,
      "epoch": 0.17342432757928541,
      "grad_norm": 3.0785446166992188,
      "kl": 0.09424351761117578,
      "learning_rate": 4.919586871126667e-06,
      "loss": 0.0038,
      "reward": 2.1120937913656235,
      "reward_std": 0.947939082980156,
      "rewards/correctness_reward_func": 1.375,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0078125,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.26834374107420444,
      "step": 162
    },
    {
      "completion_length": 90.421875,
      "epoch": 0.1744948481198983,
      "grad_norm": 4.486505508422852,
      "kl": 0.2375591630116105,
      "learning_rate": 4.917217623862516e-06,
      "loss": 0.0095,
      "reward": 1.9345000088214874,
      "reward_std": 0.7378783877938986,
      "rewards/correctness_reward_func": 1.21875,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.23918749950826168,
      "step": 163
    },
    {
      "completion_length": 89.34375,
      "epoch": 0.17556536866051117,
      "grad_norm": 5.476224422454834,
      "kl": 0.3916892586275935,
      "learning_rate": 4.914814565722671e-06,
      "loss": 0.0157,
      "reward": 1.6488437578082085,
      "reward_std": 0.861721821129322,
      "rewards/correctness_reward_func": 0.9375,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2582187484949827,
      "step": 164
    },
    {
      "completion_length": 76.15625,
      "epoch": 0.17663588920112405,
      "grad_norm": 2.504824161529541,
      "kl": 0.09898415254428983,
      "learning_rate": 4.912377730319951e-06,
      "loss": 0.004,
      "reward": 2.149609424173832,
      "reward_std": 0.8080255158711225,
      "rewards/correctness_reward_func": 1.4375,
      "rewards/int_reward_func": 0.421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.29023437574505806,
      "step": 165
    },
    {
      "completion_length": 86.703125,
      "epoch": 0.17770640974173693,
      "grad_norm": 2.5262093544006348,
      "kl": 0.10452345060184598,
      "learning_rate": 4.909907151739634e-06,
      "loss": 0.0042,
      "reward": 2.0487187057733536,
      "reward_std": 1.013342872262001,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.25965624768286943,
      "step": 166
    },
    {
      "completion_length": 83.109375,
      "epoch": 0.17877693028234978,
      "grad_norm": 3.6836423873901367,
      "kl": 0.12263510143384337,
      "learning_rate": 4.907402864538984e-06,
      "loss": 0.0049,
      "reward": 2.0983437597751617,
      "reward_std": 0.8467487432062626,
      "rewards/correctness_reward_func": 1.375,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2545937467366457,
      "step": 167
    },
    {
      "completion_length": 105.828125,
      "epoch": 0.17984745082296266,
      "grad_norm": 3.9994001388549805,
      "kl": 0.2718197964131832,
      "learning_rate": 4.904864903746765e-06,
      "loss": 0.0109,
      "reward": 1.590890608727932,
      "reward_std": 1.0225011110305786,
      "rewards/correctness_reward_func": 0.96875,
      "rewards/int_reward_func": 0.3984375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.21589063154533505,
      "step": 168
    },
    {
      "completion_length": 80.734375,
      "epoch": 0.18091797136357554,
      "grad_norm": 3.370049476623535,
      "kl": 0.0908288094215095,
      "learning_rate": 4.9022933048627496e-06,
      "loss": 0.0036,
      "reward": 1.9604843854904175,
      "reward_std": 0.6807838249951601,
      "rewards/correctness_reward_func": 1.1875,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2886093705892563,
      "step": 169
    },
    {
      "completion_length": 89.03125,
      "epoch": 0.18198849190418842,
      "grad_norm": 2.7841224670410156,
      "kl": 0.09452959662303329,
      "learning_rate": 4.899688103857223e-06,
      "loss": 0.0038,
      "reward": 2.052734389901161,
      "reward_std": 0.789879210293293,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.25585937313735485,
      "step": 170
    },
    {
      "completion_length": 99.46875,
      "epoch": 0.18305901244480127,
      "grad_norm": 2.4192087650299072,
      "kl": 0.10028906259685755,
      "learning_rate": 4.897049337170483e-06,
      "loss": 0.004,
      "reward": 1.9263124987483025,
      "reward_std": 0.4447530438192189,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.23881249967962503,
      "step": 171
    },
    {
      "completion_length": 85.359375,
      "epoch": 0.18412953298541415,
      "grad_norm": 2.552004814147949,
      "kl": 0.10651395656168461,
      "learning_rate": 4.894377041712327e-06,
      "loss": 0.0043,
      "reward": 2.072437509894371,
      "reward_std": 0.8923071715980768,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.244312503375113,
      "step": 172
    },
    {
      "completion_length": 86.828125,
      "epoch": 0.18520005352602703,
      "grad_norm": 2.8208720684051514,
      "kl": 0.08628266118466854,
      "learning_rate": 4.891671254861535e-06,
      "loss": 0.0035,
      "reward": 2.123812586069107,
      "reward_std": 0.7962243193760514,
      "rewards/correctness_reward_func": 1.40625,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2722499957308173,
      "step": 173
    },
    {
      "completion_length": 68.59375,
      "epoch": 0.1862705740666399,
      "grad_norm": 3.199782133102417,
      "kl": 0.09673942252993584,
      "learning_rate": 4.8889320144653525e-06,
      "loss": 0.0039,
      "reward": 2.4416562616825104,
      "reward_std": 0.6845003152266145,
      "rewards/correctness_reward_func": 1.65625,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3010312579572201,
      "step": 174
    },
    {
      "completion_length": 80.265625,
      "epoch": 0.1873410946072528,
      "grad_norm": 2.8050692081451416,
      "kl": 0.09070709394291043,
      "learning_rate": 4.886159358838952e-06,
      "loss": 0.0036,
      "reward": 2.2726562321186066,
      "reward_std": 0.713380170520395,
      "rewards/correctness_reward_func": 1.53125,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2804687600582838,
      "step": 175
    },
    {
      "completion_length": 95.078125,
      "epoch": 0.18841161514786564,
      "grad_norm": 3.0878233909606934,
      "kl": 0.22780032362788916,
      "learning_rate": 4.883353326764907e-06,
      "loss": 0.0091,
      "reward": 2.0888593643903732,
      "reward_std": 0.632993305567652,
      "rewards/correctness_reward_func": 1.40625,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.22948437556624413,
      "step": 176
    },
    {
      "completion_length": 92.03125,
      "epoch": 0.18948213568847852,
      "grad_norm": 6.2119035720825195,
      "kl": 0.3357097846455872,
      "learning_rate": 4.880513957492641e-06,
      "loss": 0.0134,
      "reward": 1.9138593599200249,
      "reward_std": 0.8630610294640064,
      "rewards/correctness_reward_func": 1.21875,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.24979687482118607,
      "step": 177
    },
    {
      "completion_length": 84.390625,
      "epoch": 0.1905526562290914,
      "grad_norm": 4.748230457305908,
      "kl": 0.26475911401212215,
      "learning_rate": 4.8776412907378845e-06,
      "loss": 0.0106,
      "reward": 1.8656718656420708,
      "reward_std": 0.7014469979330897,
      "rewards/correctness_reward_func": 1.125,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2641093786805868,
      "step": 178
    },
    {
      "completion_length": 95.0625,
      "epoch": 0.19162317676970428,
      "grad_norm": 2.9860572814941406,
      "kl": 0.1027436142321676,
      "learning_rate": 4.8747353666821155e-06,
      "loss": 0.0041,
      "reward": 1.8935781568288803,
      "reward_std": 0.8618385540321469,
      "rewards/correctness_reward_func": 1.15625,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.25295313261449337,
      "step": 179
    },
    {
      "completion_length": 78.359375,
      "epoch": 0.19269369731031713,
      "grad_norm": 4.26141881942749,
      "kl": 0.19185639871284366,
      "learning_rate": 4.871796225972e-06,
      "loss": 0.0077,
      "reward": 1.882093757390976,
      "reward_std": 0.762749788351357,
      "rewards/correctness_reward_func": 1.1875,
      "rewards/int_reward_func": 0.4140625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2805312527343631,
      "step": 180
    },
    {
      "completion_length": 86.953125,
      "epoch": 0.19376421785093,
      "grad_norm": 2.9578616619110107,
      "kl": 0.09208998270332813,
      "learning_rate": 4.868823909718823e-06,
      "loss": 0.0037,
      "reward": 2.173124998807907,
      "reward_std": 0.8102906746789813,
      "rewards/correctness_reward_func": 1.46875,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.25125000439584255,
      "step": 181
    },
    {
      "completion_length": 71.671875,
      "epoch": 0.1948347383915429,
      "grad_norm": 2.6002249717712402,
      "kl": 0.10551499295979738,
      "learning_rate": 4.865818459497911e-06,
      "loss": 0.0042,
      "reward": 2.2820468470454216,
      "reward_std": 0.555876774713397,
      "rewards/correctness_reward_func": 1.46875,
      "rewards/int_reward_func": 0.5,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3132968805730343,
      "step": 182
    },
    {
      "completion_length": 92.03125,
      "epoch": 0.19590525893215577,
      "grad_norm": 6.459475517272949,
      "kl": 0.2931561325676739,
      "learning_rate": 4.862779917348055e-06,
      "loss": 0.0117,
      "reward": 1.95626562833786,
      "reward_std": 0.8608931167982519,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2453281208872795,
      "step": 183
    },
    {
      "completion_length": 102.515625,
      "epoch": 0.19697577947276865,
      "grad_norm": 1.968064308166504,
      "kl": 0.09014055877923965,
      "learning_rate": 4.859708325770919e-06,
      "loss": 0.0036,
      "reward": 1.539296880364418,
      "reward_std": 0.8695460446178913,
      "rewards/correctness_reward_func": 0.875,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.22679687477648258,
      "step": 184
    },
    {
      "completion_length": 82.21875,
      "epoch": 0.1980463000133815,
      "grad_norm": 5.502157688140869,
      "kl": 0.24213434057310224,
      "learning_rate": 4.856603727730446e-06,
      "loss": 0.0097,
      "reward": 2.103000044822693,
      "reward_std": 0.7971827173605561,
      "rewards/correctness_reward_func": 1.375,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.27487499825656414,
      "step": 185
    },
    {
      "completion_length": 96.796875,
      "epoch": 0.19911682055399438,
      "grad_norm": 4.163439750671387,
      "kl": 0.2488319119438529,
      "learning_rate": 4.853466166652259e-06,
      "loss": 0.01,
      "reward": 1.8271406143903732,
      "reward_std": 0.8965076357126236,
      "rewards/correctness_reward_func": 1.125,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.22557812742888927,
      "step": 186
    },
    {
      "completion_length": 109.671875,
      "epoch": 0.20018734109460726,
      "grad_norm": 2.380798816680908,
      "kl": 0.07831509876996279,
      "learning_rate": 4.850295686423048e-06,
      "loss": 0.0031,
      "reward": 1.7582030892372131,
      "reward_std": 0.8107579126954079,
      "rewards/correctness_reward_func": 1.125,
      "rewards/int_reward_func": 0.4140625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.21132812649011612,
      "step": 187
    },
    {
      "completion_length": 96.46875,
      "epoch": 0.20125786163522014,
      "grad_norm": 5.34075927734375,
      "kl": 0.3953818525187671,
      "learning_rate": 4.8470923313899655e-06,
      "loss": 0.0158,
      "reward": 1.9607812352478504,
      "reward_std": 0.5701902243308723,
      "rewards/correctness_reward_func": 1.3125,
      "rewards/int_reward_func": 0.421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.22640624921768904,
      "step": 188
    },
    {
      "completion_length": 79.265625,
      "epoch": 0.202328382175833,
      "grad_norm": 2.1705665588378906,
      "kl": 0.10117745213210583,
      "learning_rate": 4.843856146359999e-06,
      "loss": 0.004,
      "reward": 2.0710155963897705,
      "reward_std": 0.6723045469261706,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2741406299173832,
      "step": 189
    },
    {
      "completion_length": 72.46875,
      "epoch": 0.20339890271644587,
      "grad_norm": 2.7847604751586914,
      "kl": 0.12489751679822803,
      "learning_rate": 4.8405871765993435e-06,
      "loss": 0.005,
      "reward": 1.8936093151569366,
      "reward_std": 0.8906522025354207,
      "rewards/correctness_reward_func": 1.15625,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.2920468747615814,
      "step": 190
    },
    {
      "completion_length": 74.3125,
      "epoch": 0.20446942325705875,
      "grad_norm": 7.144665241241455,
      "kl": 0.5474197333678603,
      "learning_rate": 4.837285467832775e-06,
      "loss": 0.0219,
      "reward": 1.9980624914169312,
      "reward_std": 1.1599431410431862,
      "rewards/correctness_reward_func": 1.28125,
      "rewards/int_reward_func": 0.4140625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3027499932795763,
      "step": 191
    },
    {
      "completion_length": 77.015625,
      "epoch": 0.20553994379767163,
      "grad_norm": 2.282410144805908,
      "kl": 0.12488419935107231,
      "learning_rate": 4.833951066243004e-06,
      "loss": 0.005,
      "reward": 2.0859062671661377,
      "reward_std": 0.7159075043164194,
      "rewards/correctness_reward_func": 1.3125,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3124687448143959,
      "step": 192
    },
    {
      "completion_length": 74.6875,
      "epoch": 0.20661046433828448,
      "grad_norm": 5.521911144256592,
      "kl": 0.38207234255969524,
      "learning_rate": 4.830584018470036e-06,
      "loss": 0.0153,
      "reward": 2.290625035762787,
      "reward_std": 0.6622507013380527,
      "rewards/correctness_reward_func": 1.5,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.2984374985098839,
      "step": 193
    },
    {
      "completion_length": 81.640625,
      "epoch": 0.20768098487889736,
      "grad_norm": 5.832190036773682,
      "kl": 0.24756696447730064,
      "learning_rate": 4.827184371610511e-06,
      "loss": 0.0099,
      "reward": 2.2973125129938126,
      "reward_std": 0.6708025210537016,
      "rewards/correctness_reward_func": 1.53125,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.28168749529868364,
      "step": 194
    },
    {
      "completion_length": 74.671875,
      "epoch": 0.20875150541951024,
      "grad_norm": 7.309933662414551,
      "kl": 0.12391441874206066,
      "learning_rate": 4.8237521732170525e-06,
      "loss": 0.005,
      "reward": 2.016703099012375,
      "reward_std": 1.132209412753582,
      "rewards/correctness_reward_func": 1.28125,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.29014062508940697,
      "step": 195
    },
    {
      "completion_length": 73.34375,
      "epoch": 0.20982202596012312,
      "grad_norm": 3.4460299015045166,
      "kl": 0.12927269656211138,
      "learning_rate": 4.820287471297598e-06,
      "loss": 0.0052,
      "reward": 1.9848750308156013,
      "reward_std": 0.7779360907152295,
      "rewards/correctness_reward_func": 1.21875,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2895624926313758,
      "step": 196
    },
    {
      "completion_length": 88.28125,
      "epoch": 0.210892546500736,
      "grad_norm": 4.137183666229248,
      "kl": 0.3276713816449046,
      "learning_rate": 4.816790314314729e-06,
      "loss": 0.0131,
      "reward": 1.964468702673912,
      "reward_std": 0.7385209053754807,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2535312492400408,
      "step": 197
    },
    {
      "completion_length": 73.453125,
      "epoch": 0.21196306704134885,
      "grad_norm": 6.063220500946045,
      "kl": 0.36109886690974236,
      "learning_rate": 4.813260751184992e-06,
      "loss": 0.0144,
      "reward": 2.3978749811649323,
      "reward_std": 0.7160034999251366,
      "rewards/correctness_reward_func": 1.59375,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.31975000351667404,
      "step": 198
    },
    {
      "completion_length": 90.359375,
      "epoch": 0.21303358758196173,
      "grad_norm": 9.171374320983887,
      "kl": 0.7352069662883878,
      "learning_rate": 4.809698831278217e-06,
      "loss": 0.0294,
      "reward": 1.9928593933582306,
      "reward_std": 0.8052435261197388,
      "rewards/correctness_reward_func": 1.28125,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2662968710064888,
      "step": 199
    },
    {
      "completion_length": 69.953125,
      "epoch": 0.2141041081225746,
      "grad_norm": 3.0394396781921387,
      "kl": 0.14388780342414975,
      "learning_rate": 4.806104604416824e-06,
      "loss": 0.0058,
      "reward": 2.525015652179718,
      "reward_std": 0.4300219719298184,
      "rewards/correctness_reward_func": 1.75,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2906406167894602,
      "step": 200
    },
    {
      "completion_length": 92.03125,
      "epoch": 0.21517462866318748,
      "grad_norm": 4.5592265129089355,
      "kl": 0.3974157813936472,
      "learning_rate": 4.802478120875125e-06,
      "loss": 0.0159,
      "reward": 1.534609392285347,
      "reward_std": 1.030179588124156,
      "rewards/correctness_reward_func": 0.875,
      "rewards/int_reward_func": 0.3828125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2767968699336052,
      "step": 201
    },
    {
      "completion_length": 76.625,
      "epoch": 0.21624514920380034,
      "grad_norm": 2.740817070007324,
      "kl": 0.14991699904203415,
      "learning_rate": 4.7988194313786275e-06,
      "loss": 0.006,
      "reward": 2.134265646338463,
      "reward_std": 0.8027890680823475,
      "rewards/correctness_reward_func": 1.375,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.28270312771201134,
      "step": 202
    },
    {
      "completion_length": 86.53125,
      "epoch": 0.21731566974441321,
      "grad_norm": 5.491028785705566,
      "kl": 0.32305468805134296,
      "learning_rate": 4.795128587103315e-06,
      "loss": 0.0129,
      "reward": 2.177546873688698,
      "reward_std": 0.862015737220645,
      "rewards/correctness_reward_func": 1.46875,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2556718774139881,
      "step": 203
    },
    {
      "completion_length": 80.640625,
      "epoch": 0.2183861902850261,
      "grad_norm": 3.542194366455078,
      "kl": 0.13027278054505587,
      "learning_rate": 4.791405639674941e-06,
      "loss": 0.0052,
      "reward": 1.6867187470197678,
      "reward_std": 0.8688598442822695,
      "rewards/correctness_reward_func": 0.9375,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.28046874329447746,
      "step": 204
    },
    {
      "completion_length": 75.96875,
      "epoch": 0.21945671082563897,
      "grad_norm": 3.078350305557251,
      "kl": 0.1293167658150196,
      "learning_rate": 4.7876506411683e-06,
      "loss": 0.0052,
      "reward": 2.0141249895095825,
      "reward_std": 0.7147551532834768,
      "rewards/correctness_reward_func": 1.21875,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.31881249509751797,
      "step": 205
    },
    {
      "completion_length": 62.734375,
      "epoch": 0.22052723136625185,
      "grad_norm": 7.453506946563721,
      "kl": 0.8046054858714342,
      "learning_rate": 4.783863644106502e-06,
      "loss": 0.0322,
      "reward": 1.6788437813520432,
      "reward_std": 0.8941609086468816,
      "rewards/correctness_reward_func": 0.96875,
      "rewards/int_reward_func": 0.40625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3038437496870756,
      "step": 206
    },
    {
      "completion_length": 68.671875,
      "epoch": 0.2215977519068647,
      "grad_norm": 4.4995436668396,
      "kl": 0.14437556639313698,
      "learning_rate": 4.780044701460239e-06,
      "loss": 0.0058,
      "reward": 2.126671925187111,
      "reward_std": 0.7776627587154508,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.29854688607156277,
      "step": 207
    },
    {
      "completion_length": 54.75,
      "epoch": 0.22266827244747758,
      "grad_norm": 3.271150588989258,
      "kl": 0.2097001215443015,
      "learning_rate": 4.7761938666470405e-06,
      "loss": 0.0084,
      "reward": 1.956812545657158,
      "reward_std": 0.9061004631221294,
      "rewards/correctness_reward_func": 1.125,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.3552500009536743,
      "step": 208
    },
    {
      "completion_length": 72.578125,
      "epoch": 0.22373879298809046,
      "grad_norm": 3.0446856021881104,
      "kl": 0.15493952203541994,
      "learning_rate": 4.7723111935305275e-06,
      "loss": 0.0062,
      "reward": 2.270968735218048,
      "reward_std": 0.8504615277051926,
      "rewards/correctness_reward_func": 1.46875,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.32565624453127384,
      "step": 209
    },
    {
      "completion_length": 69.046875,
      "epoch": 0.22480931352870334,
      "grad_norm": 5.6605095863342285,
      "kl": 0.44642951618880033,
      "learning_rate": 4.7683967364196624e-06,
      "loss": 0.0179,
      "reward": 2.0304374992847443,
      "reward_std": 0.7277556583285332,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3116874936968088,
      "step": 210
    },
    {
      "completion_length": 64.9375,
      "epoch": 0.2258798340693162,
      "grad_norm": 4.651093482971191,
      "kl": 0.17161214351654053,
      "learning_rate": 4.764450550067986e-06,
      "loss": 0.0069,
      "reward": 1.935359388589859,
      "reward_std": 0.850550489500165,
      "rewards/correctness_reward_func": 1.15625,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3259843699634075,
      "step": 211
    },
    {
      "completion_length": 99.921875,
      "epoch": 0.22695035460992907,
      "grad_norm": 7.470244407653809,
      "kl": 0.5526934135705233,
      "learning_rate": 4.760472689672851e-06,
      "loss": 0.0221,
      "reward": 1.62957813590765,
      "reward_std": 0.9631365463137627,
      "rewards/correctness_reward_func": 1.0,
      "rewards/int_reward_func": 0.4140625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2155156247317791,
      "step": 212
    },
    {
      "completion_length": 64.296875,
      "epoch": 0.22802087515054195,
      "grad_norm": 4.030452251434326,
      "kl": 0.1725650643929839,
      "learning_rate": 4.7564632108746524e-06,
      "loss": 0.0069,
      "reward": 2.4495781660079956,
      "reward_std": 0.7331900605931878,
      "rewards/correctness_reward_func": 1.625,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.3402031324803829,
      "step": 213
    },
    {
      "completion_length": 57.0625,
      "epoch": 0.22909139569115483,
      "grad_norm": 2.44975209236145,
      "kl": 0.1745634926483035,
      "learning_rate": 4.752422169756048e-06,
      "loss": 0.007,
      "reward": 2.49567186832428,
      "reward_std": 0.585249027935788,
      "rewards/correctness_reward_func": 1.65625,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3628593757748604,
      "step": 214
    },
    {
      "completion_length": 80.609375,
      "epoch": 0.2301619162317677,
      "grad_norm": 3.696011781692505,
      "kl": 0.1569390268996358,
      "learning_rate": 4.7483496228411754e-06,
      "loss": 0.0063,
      "reward": 2.0820469111204147,
      "reward_std": 0.7507896656170487,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.26173438131809235,
      "step": 215
    },
    {
      "completion_length": 67.90625,
      "epoch": 0.23123243677238056,
      "grad_norm": 3.6006627082824707,
      "kl": 0.1786866094917059,
      "learning_rate": 4.744245627094859e-06,
      "loss": 0.0071,
      "reward": 2.0825624614953995,
      "reward_std": 0.8309154035523534,
      "rewards/correctness_reward_func": 1.28125,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.015625,
      "rewards/xmlcount_reward_func": 0.33256249874830246,
      "step": 216
    },
    {
      "completion_length": 82.53125,
      "epoch": 0.23230295731299344,
      "grad_norm": 4.304072856903076,
      "kl": 0.31106262002140284,
      "learning_rate": 4.740110239921813e-06,
      "loss": 0.0124,
      "reward": 1.406374990940094,
      "reward_std": 0.8875350207090378,
      "rewards/correctness_reward_func": 0.6875,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.2657500021159649,
      "step": 217
    },
    {
      "completion_length": 57.75,
      "epoch": 0.23337347785360632,
      "grad_norm": 3.51265025138855,
      "kl": 0.18941342923790216,
      "learning_rate": 4.735943519165843e-06,
      "loss": 0.0076,
      "reward": 2.366296797990799,
      "reward_std": 0.6840489963069558,
      "rewards/correctness_reward_func": 1.5625,
      "rewards/int_reward_func": 0.4609375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3428593724966049,
      "step": 218
    },
    {
      "completion_length": 69.6875,
      "epoch": 0.2344439983942192,
      "grad_norm": 3.1696012020111084,
      "kl": 0.144703084602952,
      "learning_rate": 4.731745523109029e-06,
      "loss": 0.0058,
      "reward": 1.7885156497359276,
      "reward_std": 0.8288924656808376,
      "rewards/correctness_reward_func": 1.0,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.3432031273841858,
      "step": 219
    },
    {
      "completion_length": 79.453125,
      "epoch": 0.23551451893483205,
      "grad_norm": 3.0204484462738037,
      "kl": 0.16810880228877068,
      "learning_rate": 4.72751631047092e-06,
      "loss": 0.0067,
      "reward": 1.817734345793724,
      "reward_std": 1.0543543472886086,
      "rewards/correctness_reward_func": 1.0625,
      "rewards/int_reward_func": 0.4296875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.302109370008111,
      "step": 220
    },
    {
      "completion_length": 89.984375,
      "epoch": 0.23658503947544493,
      "grad_norm": 9.887064933776855,
      "kl": 0.571788308210671,
      "learning_rate": 4.723255940407704e-06,
      "loss": 0.0229,
      "reward": 2.150187447667122,
      "reward_std": 0.7866512620821595,
      "rewards/correctness_reward_func": 1.375,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.27518749609589577,
      "step": 221
    },
    {
      "completion_length": 82.15625,
      "epoch": 0.2376555600160578,
      "grad_norm": 3.509312868118286,
      "kl": 0.16501779574900866,
      "learning_rate": 4.718964472511386e-06,
      "loss": 0.0066,
      "reward": 2.0418750420212746,
      "reward_std": 0.9721324890851974,
      "rewards/correctness_reward_func": 1.28125,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.29187500290572643,
      "step": 222
    },
    {
      "completion_length": 69.703125,
      "epoch": 0.2387260805566707,
      "grad_norm": 3.7087841033935547,
      "kl": 0.18095918465405703,
      "learning_rate": 4.71464196680895e-06,
      "loss": 0.0072,
      "reward": 2.18121874332428,
      "reward_std": 0.870441822335124,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.015625,
      "rewards/xmlcount_reward_func": 0.3452812507748604,
      "step": 223
    },
    {
      "completion_length": 93.453125,
      "epoch": 0.23979660109728354,
      "grad_norm": 9.079336166381836,
      "kl": 0.7006166982464492,
      "learning_rate": 4.710288483761524e-06,
      "loss": 0.028,
      "reward": 2.110390603542328,
      "reward_std": 0.7916512079536915,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.03125,
      "rewards/xmlcount_reward_func": 0.2666406221687794,
      "step": 224
    },
    {
      "completion_length": 89.796875,
      "epoch": 0.24086712163789642,
      "grad_norm": 5.301533222198486,
      "kl": 0.33772587310522795,
      "learning_rate": 4.705904084263534e-06,
      "loss": 0.0135,
      "reward": 1.9977499693632126,
      "reward_std": 0.9460650207474828,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.31025000289082527,
      "step": 225
    },
    {
      "completion_length": 93.375,
      "epoch": 0.2419376421785093,
      "grad_norm": 2.236737012863159,
      "kl": 0.11640047281980515,
      "learning_rate": 4.701488829641845e-06,
      "loss": 0.0047,
      "reward": 2.145703136920929,
      "reward_std": 0.8284804495051503,
      "rewards/correctness_reward_func": 1.40625,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.2863281313329935,
      "step": 226
    },
    {
      "completion_length": 81.8125,
      "epoch": 0.24300816271912218,
      "grad_norm": 3.1928486824035645,
      "kl": 0.135368085000664,
      "learning_rate": 4.697042781654913e-06,
      "loss": 0.0054,
      "reward": 2.135328069329262,
      "reward_std": 0.7982124611735344,
      "rewards/correctness_reward_func": 1.3125,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.33845312520861626,
      "step": 227
    },
    {
      "completion_length": 85.453125,
      "epoch": 0.24407868325973506,
      "grad_norm": 12.730961799621582,
      "kl": 0.4075321350246668,
      "learning_rate": 4.692566002491917e-06,
      "loss": 0.0163,
      "reward": 2.0005937218666077,
      "reward_std": 0.8029458876699209,
      "rewards/correctness_reward_func": 1.21875,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.34434375166893005,
      "step": 228
    },
    {
      "completion_length": 72.875,
      "epoch": 0.2451492038003479,
      "grad_norm": 6.923903942108154,
      "kl": 0.6862649563699961,
      "learning_rate": 4.6880585547718845e-06,
      "loss": 0.0275,
      "reward": 1.7937656044960022,
      "reward_std": 0.915380734950304,
      "rewards/correctness_reward_func": 0.9375,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.015625,
      "rewards/xmlcount_reward_func": 0.371890626847744,
      "step": 229
    },
    {
      "completion_length": 93.75,
      "epoch": 0.2462197243409608,
      "grad_norm": 3.041198968887329,
      "kl": 0.109968694858253,
      "learning_rate": 4.683520501542825e-06,
      "loss": 0.0044,
      "reward": 2.041828043758869,
      "reward_std": 0.828435555100441,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.4296875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.03125,
      "rewards/xmlcount_reward_func": 0.3308906201273203,
      "step": 230
    },
    {
      "completion_length": 78.359375,
      "epoch": 0.24729024488157367,
      "grad_norm": 3.0014753341674805,
      "kl": 0.13458310719579458,
      "learning_rate": 4.67895190628084e-06,
      "loss": 0.0054,
      "reward": 2.4933906197547913,
      "reward_std": 0.8564620353281498,
      "rewards/correctness_reward_func": 1.625,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.37620312348008156,
      "step": 231
    },
    {
      "completion_length": 93.8125,
      "epoch": 0.24836076542218655,
      "grad_norm": 4.2342848777771,
      "kl": 0.1281078103929758,
      "learning_rate": 4.674352832889239e-06,
      "loss": 0.0051,
      "reward": 2.4442031383514404,
      "reward_std": 0.7425322765484452,
      "rewards/correctness_reward_func": 1.59375,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.3426406290382147,
      "step": 232
    },
    {
      "completion_length": 91.703125,
      "epoch": 0.2494312859627994,
      "grad_norm": 4.6626811027526855,
      "kl": 0.30947081558406353,
      "learning_rate": 4.669723345697646e-06,
      "loss": 0.0124,
      "reward": 1.7927343994379044,
      "reward_std": 1.0577923730015755,
      "rewards/correctness_reward_func": 1.0,
      "rewards/int_reward_func": 0.4375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.3317968789488077,
      "step": 233
    },
    {
      "completion_length": 95.1875,
      "epoch": 0.2505018065034123,
      "grad_norm": 2.6523685455322266,
      "kl": 0.12286860542371869,
      "learning_rate": 4.665063509461098e-06,
      "loss": 0.0049,
      "reward": 2.122484341263771,
      "reward_std": 0.6617152327671647,
      "rewards/correctness_reward_func": 1.28125,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.015625,
      "rewards/xmlcount_reward_func": 0.38029688596725464,
      "step": 234
    },
    {
      "completion_length": 100.3125,
      "epoch": 0.25157232704402516,
      "grad_norm": 3.743971586227417,
      "kl": 0.2634156849235296,
      "learning_rate": 4.660373389359137e-06,
      "loss": 0.0105,
      "reward": 1.9493124820291996,
      "reward_std": 0.6555369319394231,
      "rewards/correctness_reward_func": 1.1875,
      "rewards/int_reward_func": 0.4296875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": 0.3321250043809414,
      "step": 235
    },
    {
      "completion_length": 91.890625,
      "epoch": 0.252642847584638,
      "grad_norm": 5.421882629394531,
      "kl": 0.36969919549301267,
      "learning_rate": 4.655653050994907e-06,
      "loss": 0.0148,
      "reward": 2.489093706011772,
      "reward_std": 0.5151624148711562,
      "rewards/correctness_reward_func": 1.59375,
      "rewards/int_reward_func": 0.4765625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.03125,
      "rewards/xmlcount_reward_func": 0.38753124326467514,
      "step": 236
    },
    {
      "completion_length": 107.421875,
      "epoch": 0.2537133681252509,
      "grad_norm": 3.658877372741699,
      "kl": 0.3179207113571465,
      "learning_rate": 4.650902560394225e-06,
      "loss": 0.0127,
      "reward": 2.1520937085151672,
      "reward_std": 0.7971650678664446,
      "rewards/correctness_reward_func": 1.34375,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.3474062494933605,
      "step": 237
    },
    {
      "completion_length": 96.75,
      "epoch": 0.25478388866586377,
      "grad_norm": 2.333758592605591,
      "kl": 0.14108293130993843,
      "learning_rate": 4.646121984004666e-06,
      "loss": 0.0056,
      "reward": 2.4891093373298645,
      "reward_std": 0.7755477353930473,
      "rewards/correctness_reward_func": 1.59375,
      "rewards/int_reward_func": 0.484375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.38754688017070293,
      "step": 238
    },
    {
      "completion_length": 97.796875,
      "epoch": 0.2558544092064767,
      "grad_norm": 7.903679370880127,
      "kl": 0.6723966179415584,
      "learning_rate": 4.641311388694629e-06,
      "loss": 0.0269,
      "reward": 2.23892180621624,
      "reward_std": 0.8879001673776656,
      "rewards/correctness_reward_func": 1.40625,
      "rewards/int_reward_func": 0.4140625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.3951718807220459,
      "step": 239
    },
    {
      "completion_length": 128.1875,
      "epoch": 0.2569249297470895,
      "grad_norm": 2.0204544067382812,
      "kl": 0.09938508365303278,
      "learning_rate": 4.636470841752405e-06,
      "loss": 0.004,
      "reward": 1.7635936960577965,
      "reward_std": 0.65199055057019,
      "rewards/correctness_reward_func": 0.96875,
      "rewards/int_reward_func": 0.390625,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0390625,
      "rewards/xmlcount_reward_func": 0.36515624821186066,
      "step": 240
    },
    {
      "completion_length": 95.03125,
      "epoch": 0.2579954502877024,
      "grad_norm": 2.7629244327545166,
      "kl": 0.16699408926069736,
      "learning_rate": 4.631600410885231e-06,
      "loss": 0.0067,
      "reward": 2.553484320640564,
      "reward_std": 0.6153040612116456,
      "rewards/correctness_reward_func": 1.65625,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.42848438024520874,
      "step": 241
    },
    {
      "completion_length": 91.640625,
      "epoch": 0.2590659708283153,
      "grad_norm": 2.4757933616638184,
      "kl": 0.12886409275233746,
      "learning_rate": 4.626700164218349e-06,
      "loss": 0.0052,
      "reward": 2.60553115606308,
      "reward_std": 0.6941613564267755,
      "rewards/correctness_reward_func": 1.6875,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.015625,
      "rewards/xmlcount_reward_func": 0.4336562491953373,
      "step": 242
    },
    {
      "completion_length": 88.0,
      "epoch": 0.26013649136892814,
      "grad_norm": 5.133749961853027,
      "kl": 0.3455530842766166,
      "learning_rate": 4.621770170294049e-06,
      "loss": 0.0138,
      "reward": 1.829562470316887,
      "reward_std": 0.8116088081151247,
      "rewards/correctness_reward_func": 1.0,
      "rewards/int_reward_func": 0.421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.015625,
      "rewards/xmlcount_reward_func": 0.3920625038444996,
      "step": 243
    },
    {
      "completion_length": 106.015625,
      "epoch": 0.261207011909541,
      "grad_norm": 2.4294216632843018,
      "kl": 0.14171195961534977,
      "learning_rate": 4.6168104980707105e-06,
      "loss": 0.0057,
      "reward": 2.2230467945337296,
      "reward_std": 0.9806447625160217,
      "rewards/correctness_reward_func": 1.3125,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0390625,
      "rewards/xmlcount_reward_func": 0.4261718839406967,
      "step": 244
    },
    {
      "completion_length": 97.625,
      "epoch": 0.2622775324501539,
      "grad_norm": 6.696681976318359,
      "kl": 0.5324421431869268,
      "learning_rate": 4.61182121692184e-06,
      "loss": 0.0213,
      "reward": 2.5623437613248825,
      "reward_std": 0.7769194557331502,
      "rewards/correctness_reward_func": 1.65625,
      "rewards/int_reward_func": 0.453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0234375,
      "rewards/xmlcount_reward_func": 0.4295312501490116,
      "step": 245
    },
    {
      "completion_length": 105.328125,
      "epoch": 0.26334805299076675,
      "grad_norm": 3.7250571250915527,
      "kl": 0.3119491417892277,
      "learning_rate": 4.606802396635098e-06,
      "loss": 0.0125,
      "reward": 2.0679530799388885,
      "reward_std": 1.0327460495755076,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0078125,
      "rewards/xmlcount_reward_func": 0.388265622779727,
      "step": 246
    },
    {
      "completion_length": 101.0625,
      "epoch": 0.26441857353137965,
      "grad_norm": 2.0395541191101074,
      "kl": 0.1265430450439453,
      "learning_rate": 4.601754107411326e-06,
      "loss": 0.0051,
      "reward": 2.3826874494552612,
      "reward_std": 0.881409777328372,
      "rewards/correctness_reward_func": 1.46875,
      "rewards/int_reward_func": 0.46875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.046875,
      "rewards/xmlcount_reward_func": 0.3983125016093254,
      "step": 247
    },
    {
      "completion_length": 119.25,
      "epoch": 0.2654890940719925,
      "grad_norm": 50.28816223144531,
      "kl": 1.0364861502312124,
      "learning_rate": 4.596676419863561e-06,
      "loss": 0.0415,
      "reward": 1.8263437151908875,
      "reward_std": 1.0226206295192242,
      "rewards/correctness_reward_func": 1.03125,
      "rewards/int_reward_func": 0.375,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.046875,
      "rewards/xmlcount_reward_func": 0.37321874499320984,
      "step": 248
    },
    {
      "completion_length": 113.515625,
      "epoch": 0.26655961461260536,
      "grad_norm": 18.58391571044922,
      "kl": 1.4038016851991415,
      "learning_rate": 4.59156940501605e-06,
      "loss": 0.0562,
      "reward": 2.122187450528145,
      "reward_std": 0.7838817811571062,
      "rewards/correctness_reward_func": 1.25,
      "rewards/int_reward_func": 0.421875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.046875,
      "rewards/xmlcount_reward_func": 0.4034374989569187,
      "step": 249
    },
    {
      "completion_length": 77.8125,
      "epoch": 0.26763013515321826,
      "grad_norm": 2.225803852081299,
      "kl": 0.1744153881445527,
      "learning_rate": 4.586433134303257e-06,
      "loss": 0.007,
      "reward": 2.600734308362007,
      "reward_std": 0.6244143173098564,
      "rewards/correctness_reward_func": 1.59375,
      "rewards/int_reward_func": 0.4453125,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.1015625,
      "rewards/xmlcount_reward_func": 0.4601093679666519,
      "step": 250
    }
  ],
  "logging_steps": 1,
  "max_steps": 934,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 250,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}