{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "eval_steps": 27,
  "global_step": 267,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 515.3928833007812,
      "epoch": 0.003745318352059925,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": -0.0,
      "reward": -0.1899932947490015,
      "reward_std": 1.3191318809986115,
      "rewards/correctness_reward_func": 0.2812500149011612,
      "rewards/int_reward_func": 0.318080373108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.789323702454567,
      "step": 1
    },
    {
      "completion_length": 533.0357513427734,
      "epoch": 0.00749063670411985,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.00650669834431028,
      "reward_std": 1.158670336008072,
      "rewards/correctness_reward_func": 0.3125000149011612,
      "rewards/int_reward_func": 0.2756696566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5946763753890991,
      "step": 2
    },
    {
      "completion_length": 519.4286041259766,
      "epoch": 0.011235955056179775,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.07290402640137472,
      "reward_std": 1.1724370419979095,
      "rewards/correctness_reward_func": 0.2500000149011612,
      "rewards/int_reward_func": 0.266741082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5896451026201248,
      "step": 3
    },
    {
      "completion_length": 526.2857513427734,
      "epoch": 0.0149812734082397,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.16426785849034786,
      "reward_std": 1.2251884937286377,
      "rewards/correctness_reward_func": 0.1875000074505806,
      "rewards/int_reward_func": 0.2700892984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.6229732632637024,
      "step": 4
    },
    {
      "completion_length": 529.8928680419922,
      "epoch": 0.018726591760299626,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": -0.0,
      "reward": -0.23104465380311012,
      "reward_std": 1.5072951018810272,
      "rewards/correctness_reward_func": 0.2767857275903225,
      "rewards/int_reward_func": 0.2957589402794838,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8047053962945938,
      "step": 5
    },
    {
      "completion_length": 544.4643249511719,
      "epoch": 0.02247191011235955,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.09965624660253525,
      "reward_std": 1.2428130805492401,
      "rewards/correctness_reward_func": 0.334821455180645,
      "rewards/int_reward_func": 0.266741082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7012187838554382,
      "step": 6
    },
    {
      "completion_length": 522.5000152587891,
      "epoch": 0.026217228464419477,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": -0.0,
      "reward": -0.13676341250538826,
      "reward_std": 1.2565626800060272,
      "rewards/correctness_reward_func": 0.325892873108387,
      "rewards/int_reward_func": 0.2756696566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.738325908780098,
      "step": 7
    },
    {
      "completion_length": 521.4643096923828,
      "epoch": 0.0299625468164794,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.02000223658978939,
      "reward_std": 1.237034946680069,
      "rewards/correctness_reward_func": 0.3705357387661934,
      "rewards/int_reward_func": 0.2678571492433548,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6595112085342407,
      "step": 8
    },
    {
      "completion_length": 516.4643096923828,
      "epoch": 0.033707865168539325,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.19426116452086717,
      "reward_std": 1.3457031548023224,
      "rewards/correctness_reward_func": 0.2544643022119999,
      "rewards/int_reward_func": 0.2801339402794838,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7299755066633224,
      "step": 9
    },
    {
      "completion_length": 521.5357360839844,
      "epoch": 0.03745318352059925,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.08841964812017977,
      "reward_std": 1.1999126970767975,
      "rewards/correctness_reward_func": 0.2812500111758709,
      "rewards/int_reward_func": 0.3035714477300644,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6732410788536072,
      "step": 10
    },
    {
      "completion_length": 536.6428833007812,
      "epoch": 0.04119850187265917,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.2129687536507845,
      "reward_std": 1.3859791457653046,
      "rewards/correctness_reward_func": 0.2678571566939354,
      "rewards/int_reward_func": 0.271205373108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7520312815904617,
      "step": 11
    },
    {
      "completion_length": 534.5000305175781,
      "epoch": 0.0449438202247191,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.14605582039803267,
      "reward_std": 1.3036423921585083,
      "rewards/correctness_reward_func": 0.2589285895228386,
      "rewards/int_reward_func": 0.2957589440047741,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7007433176040649,
      "step": 12
    },
    {
      "completion_length": 517.1786041259766,
      "epoch": 0.04868913857677903,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": -0.0,
      "reward": -0.21629687771201134,
      "reward_std": 1.3215062022209167,
      "rewards/correctness_reward_func": 0.2678571529686451,
      "rewards/int_reward_func": 0.2700892947614193,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7542433589696884,
      "step": 13
    },
    {
      "completion_length": 523.5714569091797,
      "epoch": 0.052434456928838954,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": -0.0,
      "reward": -0.11244195885956287,
      "reward_std": 1.3325236439704895,
      "rewards/correctness_reward_func": 0.3125000111758709,
      "rewards/int_reward_func": 0.2812500149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.7073080092668533,
      "step": 14
    },
    {
      "completion_length": 520.7500305175781,
      "epoch": 0.056179775280898875,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.16676786169409752,
      "reward_std": 1.2786953151226044,
      "rewards/correctness_reward_func": 0.258928582072258,
      "rewards/int_reward_func": 0.2868303656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7125268131494522,
      "step": 15
    },
    {
      "completion_length": 529.5000305175781,
      "epoch": 0.0599250936329588,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": -0.15193304512649775,
      "reward_std": 1.2763742506504059,
      "rewards/correctness_reward_func": 0.2767857313156128,
      "rewards/int_reward_func": 0.2912946492433548,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7200134247541428,
      "step": 16
    },
    {
      "completion_length": 529.1428985595703,
      "epoch": 0.06367041198501873,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": -0.0,
      "reward": -0.12793974205851555,
      "reward_std": 1.207583248615265,
      "rewards/correctness_reward_func": 0.2544643022119999,
      "rewards/int_reward_func": 0.2678571492433548,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6502611860632896,
      "step": 17
    },
    {
      "completion_length": 532.3214569091797,
      "epoch": 0.06741573033707865,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": -0.0,
      "reward": -0.13686831016093493,
      "reward_std": 1.2532348036766052,
      "rewards/correctness_reward_func": 0.290178582072258,
      "rewards/int_reward_func": 0.2834821566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7105290591716766,
      "step": 18
    },
    {
      "completion_length": 520.7143096923828,
      "epoch": 0.07116104868913857,
      "grad_norm": 44.85335159301758,
      "kl": 0.0,
      "learning_rate": 3.7037037037037036e-08,
      "loss": 0.0,
      "reward": -0.061281259171664715,
      "reward_std": 1.1696560382843018,
      "rewards/correctness_reward_func": 0.2991071529686451,
      "rewards/int_reward_func": 0.2890625149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6494508683681488,
      "step": 19
    },
    {
      "completion_length": 521.7500152587891,
      "epoch": 0.0749063670411985,
      "grad_norm": 57.530582427978516,
      "kl": 0.0,
      "learning_rate": 7.407407407407407e-08,
      "loss": 0.0,
      "reward": -0.08753572031855583,
      "reward_std": 1.292522132396698,
      "rewards/correctness_reward_func": 0.357142873108387,
      "rewards/int_reward_func": 0.2890625149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7337411046028137,
      "step": 20
    },
    {
      "completion_length": 549.2143249511719,
      "epoch": 0.07865168539325842,
      "grad_norm": 58.32808303833008,
      "kl": 0.0003091096878051758,
      "learning_rate": 1.111111111111111e-07,
      "loss": 0.0,
      "reward": -0.011935262009501457,
      "reward_std": 1.229793220758438,
      "rewards/correctness_reward_func": 0.3705357387661934,
      "rewards/int_reward_func": 0.2946428656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6771138235926628,
      "step": 21
    },
    {
      "completion_length": 515.3928833007812,
      "epoch": 0.08239700374531835,
      "grad_norm": 46.207237243652344,
      "kl": 0.00029021501541137695,
      "learning_rate": 1.4814814814814815e-07,
      "loss": 0.0,
      "reward": -0.2325044833123684,
      "reward_std": 1.287639170885086,
      "rewards/correctness_reward_func": 0.2946428768336773,
      "rewards/int_reward_func": 0.2667410895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7938884347677231,
      "step": 22
    },
    {
      "completion_length": 520.4286041259766,
      "epoch": 0.08614232209737828,
      "grad_norm": 59.19898223876953,
      "kl": 0.0009493827819824219,
      "learning_rate": 1.8518518518518516e-07,
      "loss": 0.0,
      "reward": -0.27188840508461,
      "reward_std": 1.2708870768547058,
      "rewards/correctness_reward_func": 0.2366071566939354,
      "rewards/int_reward_func": 0.2801339402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.788629487156868,
      "step": 23
    },
    {
      "completion_length": 524.7143096923828,
      "epoch": 0.0898876404494382,
      "grad_norm": 51.78929901123047,
      "kl": 0.008197784423828125,
      "learning_rate": 2.222222222222222e-07,
      "loss": 0.0003,
      "reward": -0.1769486702978611,
      "reward_std": 1.3113940060138702,
      "rewards/correctness_reward_func": 0.2812500149011612,
      "rewards/int_reward_func": 0.2667410895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7249397337436676,
      "step": 24
    },
    {
      "completion_length": 522.7857360839844,
      "epoch": 0.09363295880149813,
      "grad_norm": 46.790557861328125,
      "kl": 0.06230926513671875,
      "learning_rate": 2.5925925925925923e-07,
      "loss": 0.0025,
      "reward": -0.05646652076393366,
      "reward_std": 1.2401413023471832,
      "rewards/correctness_reward_func": 0.3750000223517418,
      "rewards/int_reward_func": 0.2689732238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7004397809505463,
      "step": 25
    },
    {
      "completion_length": 519.3571624755859,
      "epoch": 0.09737827715355805,
      "grad_norm": 37.130306243896484,
      "kl": 0.0727081298828125,
      "learning_rate": 2.962962962962963e-07,
      "loss": 0.0029,
      "reward": -0.14675001706928015,
      "reward_std": 1.2979865074157715,
      "rewards/correctness_reward_func": 0.2991071492433548,
      "rewards/int_reward_func": 0.318080373108387,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7650536149740219,
      "step": 26
    },
    {
      "completion_length": 517.5000305175781,
      "epoch": 0.10112359550561797,
      "grad_norm": 39.826942443847656,
      "kl": 0.2943115234375,
      "learning_rate": 3.333333333333333e-07,
      "loss": 0.0118,
      "reward": -0.10758260171860456,
      "reward_std": 1.2460315823554993,
      "rewards/correctness_reward_func": 0.3080357201397419,
      "rewards/int_reward_func": 0.2734375074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6890558004379272,
      "step": 27
    },
    {
      "epoch": 0.10112359550561797,
      "eval_completion_length": 533.4601748817845,
      "eval_kl": 0.3318269428453947,
      "eval_loss": 0.013290046714246273,
      "eval_reward": -0.19538318767751517,
      "eval_reward_std": 1.2432113622364245,
      "eval_rewards/correctness_reward_func": 0.20874061106066955,
      "eval_rewards/int_reward_func": 0.27404449911494005,
      "eval_rewards/soft_format_reward_func": 0.00016447369156307297,
      "eval_rewards/strict_format_reward_func": 4.6992483303735133e-05,
      "eval_rewards/xmlcount_reward_func": -0.6783797549573999,
      "eval_runtime": 2477.4681,
      "eval_samples_per_second": 0.532,
      "eval_steps_per_second": 0.038,
      "step": 27
    },
    {
      "completion_length": 519.5000152587891,
      "epoch": 0.10486891385767791,
      "grad_norm": 36.47107696533203,
      "kl": 0.3983154296875,
      "learning_rate": 3.703703703703703e-07,
      "loss": 0.0159,
      "reward": -0.08669420145452023,
      "reward_std": 1.1603919565677643,
      "rewards/correctness_reward_func": 0.24107144493609667,
      "rewards/int_reward_func": 0.2812500149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6090156137943268,
      "step": 28
    },
    {
      "completion_length": 524.8214416503906,
      "epoch": 0.10861423220973783,
      "grad_norm": 30.607803344726562,
      "kl": 0.40972900390625,
      "learning_rate": 4.0740740740740737e-07,
      "loss": 0.0164,
      "reward": -0.06404018122702837,
      "reward_std": 1.2442965805530548,
      "rewards/correctness_reward_func": 0.3526785969734192,
      "rewards/int_reward_func": 0.2656250074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6823437958955765,
      "step": 29
    },
    {
      "completion_length": 525.4643096923828,
      "epoch": 0.11235955056179775,
      "grad_norm": 15.785079956054688,
      "kl": 0.9742431640625,
      "learning_rate": 4.444444444444444e-07,
      "loss": 0.039,
      "reward": -0.12213393254205585,
      "reward_std": 1.2745769321918488,
      "rewards/correctness_reward_func": 0.263392873108387,
      "rewards/int_reward_func": 0.2767857275903225,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6623125225305557,
      "step": 30
    },
    {
      "completion_length": 519.0714416503906,
      "epoch": 0.11610486891385768,
      "grad_norm": 12.257494926452637,
      "kl": 0.83203125,
      "learning_rate": 4.814814814814814e-07,
      "loss": 0.0333,
      "reward": -0.03814508765935898,
      "reward_std": 1.1726315319538116,
      "rewards/correctness_reward_func": 0.299107164144516,
      "rewards/int_reward_func": 0.2700892947614193,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6073415130376816,
      "step": 31
    },
    {
      "completion_length": 531.1428985595703,
      "epoch": 0.1198501872659176,
      "grad_norm": 13.690109252929688,
      "kl": 1.0924072265625,
      "learning_rate": 5.185185185185185e-07,
      "loss": 0.0437,
      "reward": -0.08791965199634433,
      "reward_std": 1.1981682777404785,
      "rewards/correctness_reward_func": 0.2410714440047741,
      "rewards/int_reward_func": 0.3247768059372902,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.653767928481102,
      "step": 32
    },
    {
      "completion_length": 527.7500152587891,
      "epoch": 0.12359550561797752,
      "grad_norm": 10.83491325378418,
      "kl": 0.9154052734375,
      "learning_rate": 5.555555555555555e-07,
      "loss": 0.0366,
      "reward": -0.1102946475148201,
      "reward_std": 1.2648592591285706,
      "rewards/correctness_reward_func": 0.2812500149011612,
      "rewards/int_reward_func": 0.2656250074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6571696400642395,
      "step": 33
    },
    {
      "completion_length": 520.0357513427734,
      "epoch": 0.12734082397003746,
      "grad_norm": 5.259093284606934,
      "kl": 1.219482421875,
      "learning_rate": 5.925925925925926e-07,
      "loss": 0.0488,
      "reward": -0.1045870566740632,
      "reward_std": 1.2835467457771301,
      "rewards/correctness_reward_func": 0.2857142984867096,
      "rewards/int_reward_func": 0.2912946566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6815960258245468,
      "step": 34
    },
    {
      "completion_length": 536.4286041259766,
      "epoch": 0.13108614232209737,
      "grad_norm": 4.761948108673096,
      "kl": 1.23583984375,
      "learning_rate": 6.296296296296296e-07,
      "loss": 0.0494,
      "reward": -0.1267544706352055,
      "reward_std": 1.181850403547287,
      "rewards/correctness_reward_func": 0.2723214328289032,
      "rewards/int_reward_func": 0.2823660895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6814420074224472,
      "step": 35
    },
    {
      "completion_length": 515.1428833007812,
      "epoch": 0.1348314606741573,
      "grad_norm": 4.87550687789917,
      "kl": 1.398681640625,
      "learning_rate": 6.666666666666666e-07,
      "loss": 0.0559,
      "reward": -0.21099777147173882,
      "reward_std": 1.4128418564796448,
      "rewards/correctness_reward_func": 0.2544642984867096,
      "rewards/int_reward_func": 0.3035714402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7690335363149643,
      "step": 36
    },
    {
      "completion_length": 521.6786041259766,
      "epoch": 0.13857677902621723,
      "grad_norm": 4.369849681854248,
      "kl": 1.700439453125,
      "learning_rate": 7.037037037037037e-07,
      "loss": 0.068,
      "reward": -0.0781383914873004,
      "reward_std": 1.3968295454978943,
      "rewards/correctness_reward_func": 0.3125000149011612,
      "rewards/int_reward_func": 0.279017873108387,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6707723736763,
      "step": 37
    },
    {
      "completion_length": 525.8214569091797,
      "epoch": 0.14232209737827714,
      "grad_norm": 4.071728706359863,
      "kl": 1.715087890625,
      "learning_rate": 7.407407407407406e-07,
      "loss": 0.0686,
      "reward": -0.18714285362511873,
      "reward_std": 1.3205263316631317,
      "rewards/correctness_reward_func": 0.2366071566939354,
      "rewards/int_reward_func": 0.3046875074505806,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.729553610086441,
      "step": 38
    },
    {
      "completion_length": 516.8214569091797,
      "epoch": 0.14606741573033707,
      "grad_norm": 4.24601411819458,
      "kl": 1.782470703125,
      "learning_rate": 7.777777777777778e-07,
      "loss": 0.0713,
      "reward": -0.1367053649155423,
      "reward_std": 1.3963970839977264,
      "rewards/correctness_reward_func": 0.3392857350409031,
      "rewards/int_reward_func": 0.2879464365541935,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7639375329017639,
      "step": 39
    },
    {
      "completion_length": 522.3214721679688,
      "epoch": 0.149812734082397,
      "grad_norm": 17.463598251342773,
      "kl": Infinity,
      "learning_rate": 8.148148148148147e-07,
      "loss": 0.1737,
      "reward": 0.012379469349980354,
      "reward_std": 1.293159008026123,
      "rewards/correctness_reward_func": 0.3883928656578064,
      "rewards/int_reward_func": 0.2857142984867096,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6628438010811806,
      "step": 40
    },
    {
      "completion_length": 519.8214569091797,
      "epoch": 0.15355805243445692,
      "grad_norm": 17.463598251342773,
      "kl": Infinity,
      "learning_rate": 8.148148148148147e-07,
      "loss": 0.1806,
      "reward": -0.09995536971837282,
      "reward_std": 1.3616773188114166,
      "rewards/correctness_reward_func": 0.3214285857975483,
      "rewards/int_reward_func": 0.3035714402794838,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7260714471340179,
      "step": 41
    },
    {
      "completion_length": 518.4643249511719,
      "epoch": 0.15730337078651685,
      "grad_norm": 13.264471054077148,
      "kl": Infinity,
      "learning_rate": 8.518518518518518e-07,
      "loss": 0.1821,
      "reward": -0.08204018184915185,
      "reward_std": 1.192356139421463,
      "rewards/correctness_reward_func": 0.2857143022119999,
      "rewards/int_reward_func": 0.3169642984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6847188025712967,
      "step": 42
    },
    {
      "completion_length": 522.6071624755859,
      "epoch": 0.16104868913857678,
      "grad_norm": 12.370172500610352,
      "kl": Infinity,
      "learning_rate": 8.888888888888888e-07,
      "loss": 0.1696,
      "reward": -0.16021204832941294,
      "reward_std": 1.268011212348938,
      "rewards/correctness_reward_func": 0.3080357313156128,
      "rewards/int_reward_func": 0.2500000074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.718247801065445,
      "step": 43
    },
    {
      "completion_length": 518.0000152587891,
      "epoch": 0.1647940074906367,
      "grad_norm": 4.8801093101501465,
      "kl": 3.1953125,
      "learning_rate": 9.259259259259259e-07,
      "loss": 0.1278,
      "reward": -0.21614731661975384,
      "reward_std": 1.3435205817222595,
      "rewards/correctness_reward_func": 0.258928582072258,
      "rewards/int_reward_func": 0.2700892947614193,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.7462812960147858,
      "step": 44
    },
    {
      "completion_length": 520.1786041259766,
      "epoch": 0.16853932584269662,
      "grad_norm": 3.4923696517944336,
      "kl": 2.48486328125,
      "learning_rate": 9.629629629629628e-07,
      "loss": 0.0994,
      "reward": -0.06906696478836238,
      "reward_std": 1.3961973786354065,
      "rewards/correctness_reward_func": 0.3839285895228386,
      "rewards/int_reward_func": 0.302455373108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7554509043693542,
      "step": 45
    },
    {
      "completion_length": 527.8571624755859,
      "epoch": 0.17228464419475656,
      "grad_norm": 3.2191402912139893,
      "kl": 2.7216796875,
      "learning_rate": 1e-06,
      "loss": 0.1089,
      "reward": -0.0912076011300087,
      "reward_std": 1.3214216530323029,
      "rewards/correctness_reward_func": 0.2857142984867096,
      "rewards/int_reward_func": 0.306919664144516,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6849575936794281,
      "step": 46
    },
    {
      "completion_length": 539.8928985595703,
      "epoch": 0.1760299625468165,
      "grad_norm": 4.932059288024902,
      "kl": 3.138671875,
      "learning_rate": 9.999571637870034e-07,
      "loss": 0.1255,
      "reward": -0.14713840186595917,
      "reward_std": 1.3085481524467468,
      "rewards/correctness_reward_func": 0.3794642984867096,
      "rewards/int_reward_func": 0.2857142984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8123169839382172,
      "step": 47
    },
    {
      "completion_length": 520.5357360839844,
      "epoch": 0.1797752808988764,
      "grad_norm": 2.9191691875457764,
      "kl": 3.26220703125,
      "learning_rate": 9.998286624877785e-07,
      "loss": 0.1305,
      "reward": -0.13023214414715767,
      "reward_std": 1.3650483787059784,
      "rewards/correctness_reward_func": 0.2991071566939354,
      "rewards/int_reward_func": 0.2734375149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7027768045663834,
      "step": 48
    },
    {
      "completion_length": 523.0714569091797,
      "epoch": 0.18352059925093633,
      "grad_norm": 3.485138177871704,
      "kl": 3.185546875,
      "learning_rate": 9.996145181203615e-07,
      "loss": 0.1274,
      "reward": -0.1737388470210135,
      "reward_std": 1.2116670310497284,
      "rewards/correctness_reward_func": 0.2321428693830967,
      "rewards/int_reward_func": 0.2801339477300644,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6860156804323196,
      "step": 49
    },
    {
      "completion_length": 522.857177734375,
      "epoch": 0.18726591760299627,
      "grad_norm": 3.4678406715393066,
      "kl": 3.61083984375,
      "learning_rate": 9.99314767377287e-07,
      "loss": 0.1445,
      "reward": 0.0006718746153637767,
      "reward_std": 1.2830963432788849,
      "rewards/correctness_reward_func": 0.3169642984867096,
      "rewards/int_reward_func": 0.2912946492433548,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6075870618224144,
      "step": 50
    },
    {
      "completion_length": 520.607177734375,
      "epoch": 0.19101123595505617,
      "grad_norm": 3.153886556625366,
      "kl": 3.3359375,
      "learning_rate": 9.989294616193017e-07,
      "loss": 0.1334,
      "reward": -0.0694642961025238,
      "reward_std": 1.3241977095603943,
      "rewards/correctness_reward_func": 0.2812500074505806,
      "rewards/int_reward_func": 0.3091517984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.6609821990132332,
      "step": 51
    },
    {
      "completion_length": 526.0357360839844,
      "epoch": 0.1947565543071161,
      "grad_norm": 3.5470564365386963,
      "kl": 3.62744140625,
      "learning_rate": 9.98458666866564e-07,
      "loss": 0.1451,
      "reward": -0.14027010113932192,
      "reward_std": 1.2881506383419037,
      "rewards/correctness_reward_func": 0.3660714402794838,
      "rewards/int_reward_func": 0.2678571566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7741986811161041,
      "step": 52
    },
    {
      "completion_length": 538.3571624755859,
      "epoch": 0.19850187265917604,
      "grad_norm": 3.714517831802368,
      "kl": 3.9716796875,
      "learning_rate": 9.979024637873308e-07,
      "loss": 0.1589,
      "reward": -0.1821674220263958,
      "reward_std": 1.214670181274414,
      "rewards/correctness_reward_func": 0.3303571678698063,
      "rewards/int_reward_func": 0.2812500074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7937746047973633,
      "step": 53
    },
    {
      "completion_length": 521.3214569091797,
      "epoch": 0.20224719101123595,
      "grad_norm": 2.559316396713257,
      "kl": 3.90087890625,
      "learning_rate": 9.972609476841365e-07,
      "loss": 0.156,
      "reward": -0.08583930134773254,
      "reward_std": 1.2739556729793549,
      "rewards/correctness_reward_func": 0.2544642984867096,
      "rewards/int_reward_func": 0.2734375074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6137410700321198,
      "step": 54
    },
    {
      "epoch": 0.20224719101123595,
      "eval_completion_length": 529.9684467516447,
      "eval_kl": Infinity,
      "eval_loss": Infinity,
      "eval_reward": -0.1981811836655987,
      "eval_reward_std": 1.2300641655921936,
      "eval_rewards/correctness_reward_func": 0.214567679245221,
      "eval_rewards/int_reward_func": 0.2699248248024991,
      "eval_rewards/soft_format_reward_func": 9.398496660747027e-05,
      "eval_rewards/strict_format_reward_func": 9.398496660747027e-05,
      "eval_rewards/xmlcount_reward_func": -0.6828616565779636,
      "eval_runtime": 2467.139,
      "eval_samples_per_second": 0.535,
      "eval_steps_per_second": 0.039,
      "step": 54
    },
    {
      "completion_length": 521.0357360839844,
      "epoch": 0.20599250936329588,
      "grad_norm": 7.153426170349121,
      "kl": 5.3447265625,
      "learning_rate": 9.965342284774631e-07,
      "loss": 0.2137,
      "reward": -0.24599106796085835,
      "reward_std": 1.428410142660141,
      "rewards/correctness_reward_func": 0.2544642984867096,
      "rewards/int_reward_func": 0.3013393059372902,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8029107004404068,
      "step": 55
    },
    {
      "completion_length": 518.3928833007812,
      "epoch": 0.20973782771535582,
      "grad_norm": 3.757683515548706,
      "kl": 4.50537109375,
      "learning_rate": 9.957224306869053e-07,
      "loss": 0.1802,
      "reward": -0.1148928627371788,
      "reward_std": 1.3824162483215332,
      "rewards/correctness_reward_func": 0.3214285895228386,
      "rewards/int_reward_func": 0.2756696604192257,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7119911015033722,
      "step": 56
    },
    {
      "completion_length": 533.7500305175781,
      "epoch": 0.21348314606741572,
      "grad_norm": 5.380308628082275,
      "kl": 5.0009765625,
      "learning_rate": 9.948256934098351e-07,
      "loss": 0.2001,
      "reward": -0.008803579956293106,
      "reward_std": 1.2614071667194366,
      "rewards/correctness_reward_func": 0.325892873108387,
      "rewards/int_reward_func": 0.297991082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.6338035762310028,
      "step": 57
    },
    {
      "completion_length": 534.8571624755859,
      "epoch": 0.21722846441947566,
      "grad_norm": 5.718860626220703,
      "kl": 5.1064453125,
      "learning_rate": 9.938441702975689e-07,
      "loss": 0.2042,
      "reward": -0.03438171138986945,
      "reward_std": 1.3422514498233795,
      "rewards/correctness_reward_func": 0.3214285895228386,
      "rewards/int_reward_func": 0.2901785895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6459888815879822,
      "step": 58
    },
    {
      "completion_length": 518.9643249511719,
      "epoch": 0.2209737827715356,
      "grad_norm": 3.3873369693756104,
      "kl": 4.580078125,
      "learning_rate": 9.927780295290389e-07,
      "loss": 0.1832,
      "reward": -0.13019196595996618,
      "reward_std": 1.2826339900493622,
      "rewards/correctness_reward_func": 0.2455357275903225,
      "rewards/int_reward_func": 0.2868303693830967,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.662558063864708,
      "step": 59
    },
    {
      "completion_length": 536.8928833007812,
      "epoch": 0.2247191011235955,
      "grad_norm": 2.9956631660461426,
      "kl": 4.05908203125,
      "learning_rate": 9.916274537819773e-07,
      "loss": 0.1624,
      "reward": -0.17484374437481165,
      "reward_std": 1.2421143352985382,
      "rewards/correctness_reward_func": 0.2232142984867096,
      "rewards/int_reward_func": 0.266741082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6647991389036179,
      "step": 60
    },
    {
      "completion_length": 531.4286041259766,
      "epoch": 0.22846441947565543,
      "grad_norm": 7.34979772567749,
      "kl": 5.10546875,
      "learning_rate": 9.90392640201615e-07,
      "loss": 0.2042,
      "reward": -0.07384821801679209,
      "reward_std": 1.2932099103927612,
      "rewards/correctness_reward_func": 0.2901785895228386,
      "rewards/int_reward_func": 0.3046875149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.6698303818702698,
      "step": 61
    },
    {
      "completion_length": 514.9643096923828,
      "epoch": 0.23220973782771537,
      "grad_norm": 5.556492328643799,
      "kl": 4.26611328125,
      "learning_rate": 9.890738003669027e-07,
      "loss": 0.1706,
      "reward": -0.13382366113364697,
      "reward_std": 1.174208790063858,
      "rewards/correctness_reward_func": 0.20982143748551607,
      "rewards/int_reward_func": 0.2845982238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6282433420419693,
      "step": 62
    },
    {
      "completion_length": 515.9643249511719,
      "epoch": 0.23595505617977527,
      "grad_norm": 3.788120985031128,
      "kl": 3.6513671875,
      "learning_rate": 9.876711602542563e-07,
      "loss": 0.146,
      "reward": -0.04057143209502101,
      "reward_std": 1.3794358968734741,
      "rewards/correctness_reward_func": 0.3348214477300644,
      "rewards/int_reward_func": 0.294642873108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.6711518168449402,
      "step": 63
    },
    {
      "completion_length": 517.1786041259766,
      "epoch": 0.2397003745318352,
      "grad_norm": 4.478960037231445,
      "kl": 3.5478515625,
      "learning_rate": 9.861849601988383e-07,
      "loss": 0.1419,
      "reward": -0.10594867006875575,
      "reward_std": 1.1321979463100433,
      "rewards/correctness_reward_func": 0.2723214402794838,
      "rewards/int_reward_func": 0.2801339402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6584040820598602,
      "step": 64
    },
    {
      "completion_length": 522.9643096923828,
      "epoch": 0.24344569288389514,
      "grad_norm": 3.61104416847229,
      "kl": 3.99169921875,
      "learning_rate": 9.846154548533772e-07,
      "loss": 0.1596,
      "reward": -0.13633259385824203,
      "reward_std": 1.269362598657608,
      "rewards/correctness_reward_func": 0.2723214477300644,
      "rewards/int_reward_func": 0.3013392984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.709993377327919,
      "step": 65
    },
    {
      "completion_length": 522.0000152587891,
      "epoch": 0.24719101123595505,
      "grad_norm": 8.631311416625977,
      "kl": 5.59375,
      "learning_rate": 9.82962913144534e-07,
      "loss": 0.2237,
      "reward": -0.053970987908542156,
      "reward_std": 1.3836995363235474,
      "rewards/correctness_reward_func": 0.4151785969734192,
      "rewards/int_reward_func": 0.2935267984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7626763731241226,
      "step": 66
    },
    {
      "completion_length": 518.2500152587891,
      "epoch": 0.250936329588015,
      "grad_norm": 12.603729248046875,
      "kl": 6.03515625,
      "learning_rate": 9.812276182268236e-07,
      "loss": 0.2414,
      "reward": -0.23663169657811522,
      "reward_std": 1.226292833685875,
      "rewards/correctness_reward_func": 0.2098214402794838,
      "rewards/int_reward_func": 0.2645089402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7109620720148087,
      "step": 67
    },
    {
      "completion_length": 518.8214721679688,
      "epoch": 0.2546816479400749,
      "grad_norm": 4.896903991699219,
      "kl": 4.85107421875,
      "learning_rate": 9.794098674340966e-07,
      "loss": 0.194,
      "reward": -0.036457577865803614,
      "reward_std": 1.3130376636981964,
      "rewards/correctness_reward_func": 0.3526785895228386,
      "rewards/int_reward_func": 0.2834821492433548,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6726183444261551,
      "step": 68
    },
    {
      "completion_length": 517.1428833007812,
      "epoch": 0.25842696629213485,
      "grad_norm": 2.8336970806121826,
      "kl": 4.263671875,
      "learning_rate": 9.775099722285934e-07,
      "loss": 0.1706,
      "reward": -0.20734822936356068,
      "reward_std": 1.3524066507816315,
      "rewards/correctness_reward_func": 0.258928582072258,
      "rewards/int_reward_func": 0.3002232313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7665000259876251,
      "step": 69
    },
    {
      "completion_length": 517.3214569091797,
      "epoch": 0.26217228464419473,
      "grad_norm": 3.12200665473938,
      "kl": 4.041015625,
      "learning_rate": 9.755282581475767e-07,
      "loss": 0.1616,
      "reward": 0.04342857655137777,
      "reward_std": 1.1121753752231598,
      "rewards/correctness_reward_func": 0.3035714365541935,
      "rewards/int_reward_func": 0.2790178656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.539160743355751,
      "step": 70
    },
    {
      "completion_length": 524.8928833007812,
      "epoch": 0.26591760299625467,
      "grad_norm": 2.3125550746917725,
      "kl": 3.697265625,
      "learning_rate": 9.73465064747553e-07,
      "loss": 0.1479,
      "reward": -0.2864709943532944,
      "reward_std": 1.2171598970890045,
      "rewards/correctness_reward_func": 0.15178572107106447,
      "rewards/int_reward_func": 0.2477678656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6860245764255524,
      "step": 71
    },
    {
      "completion_length": 526.4285888671875,
      "epoch": 0.2696629213483146,
      "grad_norm": 2.5901665687561035,
      "kl": 4.2919921875,
      "learning_rate": 9.713207455460892e-07,
      "loss": 0.1717,
      "reward": -0.0840669646859169,
      "reward_std": 1.3054583966732025,
      "rewards/correctness_reward_func": 0.3571428842842579,
      "rewards/int_reward_func": 0.305803582072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7470134347677231,
      "step": 72
    },
    {
      "completion_length": 527.3571624755859,
      "epoch": 0.27340823970037453,
      "grad_norm": 2.6217572689056396,
      "kl": 4.8828125,
      "learning_rate": 9.69095667961242e-07,
      "loss": 0.1953,
      "reward": -0.07796428725123405,
      "reward_std": 1.1255577206611633,
      "rewards/correctness_reward_func": 0.2812500111758709,
      "rewards/int_reward_func": 0.2879464402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6471607089042664,
      "step": 73
    },
    {
      "completion_length": 526.9286041259766,
      "epoch": 0.27715355805243447,
      "grad_norm": 11.475034713745117,
      "kl": Infinity,
      "learning_rate": 9.667902132486008e-07,
      "loss": 0.2543,
      "reward": -0.06999553460627794,
      "reward_std": 1.2003649473190308,
      "rewards/correctness_reward_func": 0.294642873108387,
      "rewards/int_reward_func": 0.2890625149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6537009179592133,
      "step": 74
    },
    {
      "completion_length": 517.8571624755859,
      "epoch": 0.2808988764044944,
      "grad_norm": 11.475034713745117,
      "kl": Infinity,
      "learning_rate": 9.667902132486008e-07,
      "loss": 0.2206,
      "reward": -0.1703772358596325,
      "reward_std": 1.2204445004463196,
      "rewards/correctness_reward_func": 0.2857142984867096,
      "rewards/int_reward_func": 0.2421875074505806,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6993951201438904,
      "step": 75
    },
    {
      "completion_length": 522.0714569091797,
      "epoch": 0.2846441947565543,
      "grad_norm": 2.270301580429077,
      "kl": 5.0947265625,
      "learning_rate": 9.644047764359621e-07,
      "loss": 0.2038,
      "reward": -0.13376785721629858,
      "reward_std": 1.3083040714263916,
      "rewards/correctness_reward_func": 0.2901785895228386,
      "rewards/int_reward_func": 0.2790178656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7029643207788467,
      "step": 76
    },
    {
      "completion_length": 523.9286041259766,
      "epoch": 0.2883895131086142,
      "grad_norm": 3.0469226837158203,
      "kl": 4.9599609375,
      "learning_rate": 9.619397662556433e-07,
      "loss": 0.1984,
      "reward": -0.1042410780210048,
      "reward_std": 1.188673734664917,
      "rewards/correctness_reward_func": 0.2410714440047741,
      "rewards/int_reward_func": 0.2611607313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6064732521772385,
      "step": 77
    },
    {
      "completion_length": 521.5714569091797,
      "epoch": 0.29213483146067415,
      "grad_norm": 2.852745294570923,
      "kl": 4.5966796875,
      "learning_rate": 9.593956050744492e-07,
      "loss": 0.1839,
      "reward": -0.15841741440817714,
      "reward_std": 1.2980459928512573,
      "rewards/correctness_reward_func": 0.2142857275903225,
      "rewards/int_reward_func": 0.291294664144516,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6639978140592575,
      "step": 78
    },
    {
      "completion_length": 525.6786041259766,
      "epoch": 0.2958801498127341,
      "grad_norm": 4.637290000915527,
      "kl": 4.21337890625,
      "learning_rate": 9.567727288213004e-07,
      "loss": 0.1685,
      "reward": -0.19912277371622622,
      "reward_std": 1.2618965804576874,
      "rewards/correctness_reward_func": 0.2633928656578064,
      "rewards/int_reward_func": 0.2600446529686451,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7225602865219116,
      "step": 79
    },
    {
      "completion_length": 522.5357360839844,
      "epoch": 0.299625468164794,
      "grad_norm": 3.2194876670837402,
      "kl": 3.90625,
      "learning_rate": 9.540715869125407e-07,
      "loss": 0.1563,
      "reward": -0.020919647999107838,
      "reward_std": 1.222971111536026,
      "rewards/correctness_reward_func": 0.2857142984867096,
      "rewards/int_reward_func": 0.277901791036129,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.584535762667656,
      "step": 80
    },
    {
      "completion_length": 529.8214416503906,
      "epoch": 0.30337078651685395,
      "grad_norm": 2.404968500137329,
      "kl": 3.91064453125,
      "learning_rate": 9.512926421749303e-07,
      "loss": 0.1564,
      "reward": -0.19220760464668274,
      "reward_std": 1.3635277450084686,
      "rewards/correctness_reward_func": 0.2812500111758709,
      "rewards/int_reward_func": 0.2845982313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7580558508634567,
      "step": 81
    },
    {
      "epoch": 0.30337078651685395,
      "eval_completion_length": 537.6496517783717,
      "eval_kl": Infinity,
      "eval_loss": Infinity,
      "eval_reward": -0.2110714678873161,
      "eval_reward_std": 1.246040644143757,
      "eval_rewards/correctness_reward_func": 0.21428572410031368,
      "eval_rewards/int_reward_func": 0.27151473525323366,
      "eval_rewards/soft_format_reward_func": 7.048872495560269e-05,
      "eval_rewards/strict_format_reward_func": 0.00016447369156307297,
      "eval_rewards/xmlcount_reward_func": -0.697106892497916,
      "eval_runtime": 2496.3791,
      "eval_samples_per_second": 0.528,
      "eval_steps_per_second": 0.038,
      "step": 81
    },
    {
      "completion_length": 525.0357513427734,
      "epoch": 0.30711610486891383,
      "grad_norm": 2.384713888168335,
      "kl": 5.0546875,
      "learning_rate": 9.484363707663441e-07,
      "loss": 0.2022,
      "reward": -0.17487278208136559,
      "reward_std": 1.2226494252681732,
      "rewards/correctness_reward_func": 0.2410714402794838,
      "rewards/int_reward_func": 0.266741082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6826852709054947,
      "step": 82
    },
    {
      "completion_length": 523.6785888671875,
      "epoch": 0.31086142322097376,
      "grad_norm": 4.885804653167725,
      "kl": 5.853515625,
      "learning_rate": 9.455032620941839e-07,
      "loss": 0.2341,
      "reward": -0.16699554910883307,
      "reward_std": 1.3849684596061707,
      "rewards/correctness_reward_func": 0.3125000074505806,
      "rewards/int_reward_func": 0.3002232238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7797188013792038,
      "step": 83
    },
    {
      "completion_length": 519.7500305175781,
      "epoch": 0.3146067415730337,
      "grad_norm": 5.602491855621338,
      "kl": 5.1337890625,
      "learning_rate": 9.424938187315209e-07,
      "loss": 0.2054,
      "reward": -0.13761384692043066,
      "reward_std": 1.31485316157341,
      "rewards/correctness_reward_func": 0.3437500149011612,
      "rewards/int_reward_func": 0.2533482275903225,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.734712079167366,
      "step": 84
    },
    {
      "completion_length": 528.607177734375,
      "epoch": 0.31835205992509363,
      "grad_norm": 3.0608065128326416,
      "kl": 4.798828125,
      "learning_rate": 9.394085563309826e-07,
      "loss": 0.1919,
      "reward": -0.3015223369002342,
      "reward_std": 1.3573426604270935,
      "rewards/correctness_reward_func": 0.20089286798611283,
      "rewards/int_reward_func": 0.2756696529686451,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.779200941324234,
      "step": 85
    },
    {
      "completion_length": 520.4286041259766,
      "epoch": 0.32209737827715357,
      "grad_norm": 3.1792619228363037,
      "kl": 4.05029296875,
      "learning_rate": 9.362480035363985e-07,
      "loss": 0.162,
      "reward": -0.09921875572763383,
      "reward_std": 1.2572968900203705,
      "rewards/correctness_reward_func": 0.2276785783469677,
      "rewards/int_reward_func": 0.302455373108387,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6304687634110451,
      "step": 86
    },
    {
      "completion_length": 524.8928833007812,
      "epoch": 0.3258426966292135,
      "grad_norm": 2.641871690750122,
      "kl": 4.3486328125,
      "learning_rate": 9.330127018922193e-07,
      "loss": 0.1739,
      "reward": -0.1489598285406828,
      "reward_std": 1.2619026154279709,
      "rewards/correctness_reward_func": 0.2812500223517418,
      "rewards/int_reward_func": 0.282366082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7125759124755859,
      "step": 87
    },
    {
      "completion_length": 515.8571624755859,
      "epoch": 0.3295880149812734,
      "grad_norm": 2.3113279342651367,
      "kl": 4.58251953125,
      "learning_rate": 9.297032057507264e-07,
      "loss": 0.1833,
      "reward": -0.15345759969204664,
      "reward_std": 1.2697344720363617,
      "rewards/correctness_reward_func": 0.330357164144516,
      "rewards/int_reward_func": 0.2444196529686451,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7282344102859497,
      "step": 88
    },
    {
      "completion_length": 522.4643096923828,
      "epoch": 0.3333333333333333,
      "grad_norm": 2.758471727371216,
      "kl": 5.03125,
      "learning_rate": 9.26320082177046e-07,
      "loss": 0.2012,
      "reward": -0.03455581283196807,
      "reward_std": 1.2184255123138428,
      "rewards/correctness_reward_func": 0.3839285969734192,
      "rewards/int_reward_func": 0.3046875149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0022321429569274187,
      "rewards/xmlcount_reward_func": -0.7254040837287903,
      "step": 89
    },
    {
      "completion_length": 515.5000305175781,
      "epoch": 0.33707865168539325,
      "grad_norm": 2.942793130874634,
      "kl": 4.84375,
      "learning_rate": 9.228639108519866e-07,
      "loss": 0.1937,
      "reward": -0.12176339235156775,
      "reward_std": 1.2805460095405579,
      "rewards/correctness_reward_func": 0.227678582072258,
      "rewards/int_reward_func": 0.282366082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.6329241469502449,
      "step": 90
    },
    {
      "completion_length": 518.8571624755859,
      "epoch": 0.3408239700374532,
      "grad_norm": 2.8008244037628174,
      "kl": 5.3193359375,
      "learning_rate": 9.19335283972712e-07,
      "loss": 0.2127,
      "reward": -0.05108258547261357,
      "reward_std": 1.26492041349411,
      "rewards/correctness_reward_func": 0.3348214477300644,
      "rewards/int_reward_func": 0.2745535857975483,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6604576334357262,
      "step": 91
    },
    {
      "completion_length": 517.5714569091797,
      "epoch": 0.3445692883895131,
      "grad_norm": 3.959707498550415,
      "kl": 5.1396484375,
      "learning_rate": 9.157348061512726e-07,
      "loss": 0.2055,
      "reward": -0.3041093908250332,
      "reward_std": 1.284685343503952,
      "rewards/correctness_reward_func": 0.2500000074505806,
      "rewards/int_reward_func": 0.279017873108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8331272155046463,
      "step": 92
    },
    {
      "completion_length": 525.4643096923828,
      "epoch": 0.34831460674157305,
      "grad_norm": 2.2119061946868896,
      "kl": 4.5244140625,
      "learning_rate": 9.120630943110077e-07,
      "loss": 0.181,
      "reward": -0.16022099647670984,
      "reward_std": 1.251663863658905,
      "rewards/correctness_reward_func": 0.2589285895228386,
      "rewards/int_reward_func": 0.2700892984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.6903549581766129,
      "step": 93
    },
    {
      "completion_length": 520.1071929931641,
      "epoch": 0.352059925093633,
      "grad_norm": 2.263819456100464,
      "kl": 4.20166015625,
      "learning_rate": 9.083207775808394e-07,
      "loss": 0.1681,
      "reward": -0.20089509710669518,
      "reward_std": 1.3503172099590302,
      "rewards/correctness_reward_func": 0.2901785857975483,
      "rewards/int_reward_func": 0.2723214328289032,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.7645111978054047,
      "step": 94
    },
    {
      "completion_length": 516.607177734375,
      "epoch": 0.35580524344569286,
      "grad_norm": 2.402242422103882,
      "kl": 4.662109375,
      "learning_rate": 9.045084971874737e-07,
      "loss": 0.1865,
      "reward": -0.2237209901213646,
      "reward_std": 1.2929619252681732,
      "rewards/correctness_reward_func": 0.2857143022119999,
      "rewards/int_reward_func": 0.2678571566939354,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7784085273742676,
      "step": 95
    },
    {
      "completion_length": 524.4285888671875,
      "epoch": 0.3595505617977528,
      "grad_norm": 2.300976276397705,
      "kl": 5.189453125,
      "learning_rate": 9.006269063455302e-07,
      "loss": 0.2076,
      "reward": -0.05375669337809086,
      "reward_std": 1.2477286458015442,
      "rewards/correctness_reward_func": 0.3080357313156128,
      "rewards/int_reward_func": 0.2979910857975483,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6597835123538971,
      "step": 96
    },
    {
      "completion_length": 525.0357360839844,
      "epoch": 0.36329588014981273,
      "grad_norm": 2.4251925945281982,
      "kl": 4.642578125,
      "learning_rate": 8.966766701456176e-07,
      "loss": 0.1857,
      "reward": -0.1464910740032792,
      "reward_std": 1.3847712874412537,
      "rewards/correctness_reward_func": 0.3526785895228386,
      "rewards/int_reward_func": 0.294642873108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7938125729560852,
      "step": 97
    },
    {
      "completion_length": 525.0000305175781,
      "epoch": 0.36704119850187267,
      "grad_norm": 6.032417297363281,
      "kl": 4.716796875,
      "learning_rate": 8.926584654403724e-07,
      "loss": 0.1886,
      "reward": -0.1682098340243101,
      "reward_std": 1.2604454159736633,
      "rewards/correctness_reward_func": 0.3035714402794838,
      "rewards/int_reward_func": 0.2968750149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7686562836170197,
      "step": 98
    },
    {
      "completion_length": 516.9643249511719,
      "epoch": 0.3707865168539326,
      "grad_norm": 2.159947633743286,
      "kl": 4.19580078125,
      "learning_rate": 8.885729807284854e-07,
      "loss": 0.1678,
      "reward": -0.15086161345243454,
      "reward_std": 1.2245171070098877,
      "rewards/correctness_reward_func": 0.227678582072258,
      "rewards/int_reward_func": 0.2801339402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6586741358041763,
      "step": 99
    },
    {
      "completion_length": 524.7857360839844,
      "epoch": 0.37453183520599254,
      "grad_norm": 2.146036148071289,
      "kl": 4.57275390625,
      "learning_rate": 8.844209160367298e-07,
      "loss": 0.1829,
      "reward": -0.24398661218583584,
      "reward_std": 1.3106769621372223,
      "rewards/correctness_reward_func": 0.2901785895228386,
      "rewards/int_reward_func": 0.2968750149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.831040233373642,
      "step": 100
    },
    {
      "completion_length": 516.7857360839844,
      "epoch": 0.3782771535580524,
      "grad_norm": 2.4187943935394287,
      "kl": 5.0068359375,
      "learning_rate": 8.802029828000155e-07,
      "loss": 0.2003,
      "reward": -0.10016518598422408,
      "reward_std": 1.3009901344776154,
      "rewards/correctness_reward_func": 0.3035714440047741,
      "rewards/int_reward_func": 0.279017873108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6827544867992401,
      "step": 101
    },
    {
      "completion_length": 516.1428833007812,
      "epoch": 0.38202247191011235,
      "grad_norm": 3.7174782752990723,
      "kl": 5.791015625,
      "learning_rate": 8.759199037394886e-07,
      "loss": 0.2317,
      "reward": -0.10657813027501106,
      "reward_std": 1.2493732273578644,
      "rewards/correctness_reward_func": 0.2991071566939354,
      "rewards/int_reward_func": 0.275669664144516,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6813549548387527,
      "step": 102
    },
    {
      "completion_length": 520.2143096923828,
      "epoch": 0.3857677902621723,
      "grad_norm": 9.145467758178711,
      "kl": Infinity,
      "learning_rate": 8.71572412738697e-07,
      "loss": 0.2629,
      "reward": -0.08905134350061417,
      "reward_std": 1.1978721618652344,
      "rewards/correctness_reward_func": 0.2678571529686451,
      "rewards/int_reward_func": 0.2578125111758709,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6147210150957108,
      "step": 103
    },
    {
      "completion_length": 514.3928833007812,
      "epoch": 0.3895131086142322,
      "grad_norm": 5.3565168380737305,
      "kl": 5.62890625,
      "learning_rate": 8.671612547178427e-07,
      "loss": 0.2251,
      "reward": -0.10154464282095432,
      "reward_std": 1.2342998683452606,
      "rewards/correctness_reward_func": 0.263392873108387,
      "rewards/int_reward_func": 0.2879464402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.652883917093277,
      "step": 104
    },
    {
      "completion_length": 527.7857360839844,
      "epoch": 0.39325842696629215,
      "grad_norm": 2.9622256755828857,
      "kl": 5.0244140625,
      "learning_rate": 8.626871855061437e-07,
      "loss": 0.201,
      "reward": -0.029102683067321777,
      "reward_std": 1.1962931752204895,
      "rewards/correctness_reward_func": 0.2723214402794838,
      "rewards/int_reward_func": 0.2779017984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5793259218335152,
      "step": 105
    },
    {
      "completion_length": 530.6071624755859,
      "epoch": 0.3970037453183521,
      "grad_norm": 2.1288747787475586,
      "kl": 4.4072265625,
      "learning_rate": 8.581509717123272e-07,
      "loss": 0.1763,
      "reward": -0.10718304198235273,
      "reward_std": 1.2426259815692902,
      "rewards/correctness_reward_func": 0.3571428805589676,
      "rewards/int_reward_func": 0.3035714402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.767897367477417,
      "step": 106
    },
    {
      "completion_length": 529.8571624755859,
      "epoch": 0.40074906367041196,
      "grad_norm": 2.16975474357605,
      "kl": 4.44921875,
      "learning_rate": 8.535533905932737e-07,
      "loss": 0.178,
      "reward": -0.08131696470081806,
      "reward_std": 1.2708694338798523,
      "rewards/correctness_reward_func": 0.325892873108387,
      "rewards/int_reward_func": 0.2968750149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7040848881006241,
      "step": 107
    },
    {
      "completion_length": 519.9643096923828,
      "epoch": 0.4044943820224719,
      "grad_norm": 2.3932907581329346,
      "kl": 4.525390625,
      "learning_rate": 8.488952299208401e-07,
      "loss": 0.181,
      "reward": -0.09825893118977547,
      "reward_std": 1.1982493996620178,
      "rewards/correctness_reward_func": 0.2723214477300644,
      "rewards/int_reward_func": 0.2845982238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6551786065101624,
      "step": 108
    },
    {
      "epoch": 0.4044943820224719,
      "eval_completion_length": 531.9955142372532,
      "eval_kl": Infinity,
      "eval_loss": Infinity,
      "eval_reward": -0.18952424128258014,
      "eval_reward_std": 1.2504213703306097,
      "eval_rewards/correctness_reward_func": 0.2154135453073602,
      "eval_rewards/int_reward_func": 0.27948779385340844,
      "eval_rewards/soft_format_reward_func": 7.048872495560269e-05,
      "eval_rewards/strict_format_reward_func": 9.398496660747027e-05,
      "eval_rewards/xmlcount_reward_func": -0.6845900519898064,
      "eval_runtime": 2483.5009,
      "eval_samples_per_second": 0.531,
      "eval_steps_per_second": 0.038,
      "step": 108
    },
    {
      "completion_length": 519.8214569091797,
      "epoch": 0.40823970037453183,
      "grad_norm": 2.8793561458587646,
      "kl": 4.71240234375,
      "learning_rate": 8.441772878468769e-07,
      "loss": 0.1885,
      "reward": -0.26077233999967575,
      "reward_std": 1.3895916938781738,
      "rewards/correctness_reward_func": 0.2901785857975483,
      "rewards/int_reward_func": 0.2678571566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8188081234693527,
      "step": 109
    },
    {
      "completion_length": 527.8928833007812,
      "epoch": 0.41198501872659177,
      "grad_norm": 2.0183770656585693,
      "kl": 4.36474609375,
      "learning_rate": 8.394003727664709e-07,
      "loss": 0.1746,
      "reward": -0.0459575979039073,
      "reward_std": 1.1931805312633514,
      "rewards/correctness_reward_func": 0.3437500149011612,
      "rewards/int_reward_func": 0.290178582072258,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.681002289056778,
      "step": 110
    },
    {
      "completion_length": 518.0714416503906,
      "epoch": 0.4157303370786517,
      "grad_norm": 2.79196834564209,
      "kl": 5.20703125,
      "learning_rate": 8.34565303179429e-07,
      "loss": 0.2083,
      "reward": 0.009531241841614246,
      "reward_std": 1.1548882126808167,
      "rewards/correctness_reward_func": 0.27678573317825794,
      "rewards/int_reward_func": 0.2901785783469677,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5585491135716438,
      "step": 111
    },
    {
      "completion_length": 526.0714569091797,
      "epoch": 0.41947565543071164,
      "grad_norm": 5.046426773071289,
      "kl": 5.162109375,
      "learning_rate": 8.296729075500343e-07,
      "loss": 0.2065,
      "reward": -0.1916651837527752,
      "reward_std": 1.2752774059772491,
      "rewards/correctness_reward_func": 0.2678571566939354,
      "rewards/int_reward_func": 0.314732164144516,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7742545157670975,
      "step": 112
    },
    {
      "completion_length": 540.5714416503906,
      "epoch": 0.4232209737827715,
      "grad_norm": 2.5626397132873535,
      "kl": 4.7333984375,
      "learning_rate": 8.247240241650917e-07,
      "loss": 0.1893,
      "reward": -0.046348221600055695,
      "reward_std": 1.2709387838840485,
      "rewards/correctness_reward_func": 0.3348214477300644,
      "rewards/int_reward_func": 0.2723214365541935,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6534911245107651,
      "step": 113
    },
    {
      "completion_length": 518.4643096923828,
      "epoch": 0.42696629213483145,
      "grad_norm": 2.389897584915161,
      "kl": 4.4033203125,
      "learning_rate": 8.197195009902923e-07,
      "loss": 0.1761,
      "reward": -0.08783036330714822,
      "reward_std": 1.316321700811386,
      "rewards/correctness_reward_func": 0.4107142984867096,
      "rewards/int_reward_func": 0.3024553656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8010000437498093,
      "step": 114
    },
    {
      "completion_length": 522.5000305175781,
      "epoch": 0.4307116104868914,
      "grad_norm": 2.389897584915161,
      "kl": Infinity,
      "learning_rate": 8.197195009902923e-07,
      "loss": 0.1755,
      "reward": -0.08958705811528489,
      "reward_std": 1.2088698744773865,
      "rewards/correctness_reward_func": 0.2991071566939354,
      "rewards/int_reward_func": 0.2912946566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6799888461828232,
      "step": 115
    },
    {
      "completion_length": 515.6071624755859,
      "epoch": 0.4344569288389513,
      "grad_norm": 1.8981910943984985,
      "kl": 4.796875,
      "learning_rate": 8.146601955249187e-07,
      "loss": 0.1919,
      "reward": -0.08243527729064226,
      "reward_std": 1.3043743073940277,
      "rewards/correctness_reward_func": 0.2946428656578064,
      "rewards/int_reward_func": 0.2712053656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6482834964990616,
      "step": 116
    },
    {
      "completion_length": 522.7500305175781,
      "epoch": 0.43820224719101125,
      "grad_norm": 2.014021873474121,
      "kl": 4.29296875,
      "learning_rate": 8.095469746549171e-07,
      "loss": 0.1717,
      "reward": -0.06306472327560186,
      "reward_std": 1.194327026605606,
      "rewards/correctness_reward_func": 0.2633928656578064,
      "rewards/int_reward_func": 0.2857142947614193,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6132879704236984,
      "step": 117
    },
    {
      "completion_length": 530.6428833007812,
      "epoch": 0.4419475655430712,
      "grad_norm": 2.255415439605713,
      "kl": 5.0439453125,
      "learning_rate": 8.043807145043603e-07,
      "loss": 0.2018,
      "reward": -0.22622546181082726,
      "reward_std": 1.2605921626091003,
      "rewards/correctness_reward_func": 0.1964285783469677,
      "rewards/int_reward_func": 0.266741082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6893951445817947,
      "step": 118
    },
    {
      "completion_length": 524.8571624755859,
      "epoch": 0.44569288389513106,
      "grad_norm": 2.743229627609253,
      "kl": 5.2587890625,
      "learning_rate": 7.991623002853294e-07,
      "loss": 0.2104,
      "reward": -0.18948884680867195,
      "reward_std": 1.2621938586235046,
      "rewards/correctness_reward_func": 0.2142857238650322,
      "rewards/int_reward_func": 0.2678571566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6716317236423492,
      "step": 119
    },
    {
      "completion_length": 517.5000305175781,
      "epoch": 0.449438202247191,
      "grad_norm": 3.164485454559326,
      "kl": 5.5810546875,
      "learning_rate": 7.938926261462365e-07,
      "loss": 0.2232,
      "reward": -0.21733705699443817,
      "reward_std": 1.231595516204834,
      "rewards/correctness_reward_func": 0.2187500149011612,
      "rewards/int_reward_func": 0.3058035895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.741890624165535,
      "step": 120
    },
    {
      "completion_length": 530.857177734375,
      "epoch": 0.45318352059925093,
      "grad_norm": 3.164485454559326,
      "kl": Infinity,
      "learning_rate": 7.938926261462365e-07,
      "loss": 0.234,
      "reward": -0.19310491904616356,
      "reward_std": 1.3157844841480255,
      "rewards/correctness_reward_func": 0.258928582072258,
      "rewards/int_reward_func": 0.2801339402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7321674823760986,
      "step": 121
    },
    {
      "completion_length": 517.7143249511719,
      "epoch": 0.45692883895131087,
      "grad_norm": 2.4260902404785156,
      "kl": 5.0986328125,
      "learning_rate": 7.885725950186168e-07,
      "loss": 0.2039,
      "reward": -0.11468750424683094,
      "reward_std": 1.3093117475509644,
      "rewards/correctness_reward_func": 0.3392857350409031,
      "rewards/int_reward_func": 0.2834821566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7374553978443146,
      "step": 122
    },
    {
      "completion_length": 522.8571624755859,
      "epoch": 0.4606741573033708,
      "grad_norm": 1.6100739240646362,
      "kl": 4.80859375,
      "learning_rate": 7.832031184624164e-07,
      "loss": 0.1923,
      "reward": -0.1980089396238327,
      "reward_std": 1.338472455739975,
      "rewards/correctness_reward_func": 0.2723214402794838,
      "rewards/int_reward_func": 0.2790178656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.7504643201828003,
      "step": 123
    },
    {
      "completion_length": 526.2857360839844,
      "epoch": 0.46441947565543074,
      "grad_norm": 1.831390380859375,
      "kl": 4.9580078125,
      "learning_rate": 7.777851165098011e-07,
      "loss": 0.1983,
      "reward": -0.10124107450246811,
      "reward_std": 1.3591813445091248,
      "rewards/correctness_reward_func": 0.3169643022119999,
      "rewards/int_reward_func": 0.2857143059372902,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7039196640253067,
      "step": 124
    },
    {
      "completion_length": 525.1428833007812,
      "epoch": 0.4681647940074906,
      "grad_norm": 1.9850410223007202,
      "kl": 4.2109375,
      "learning_rate": 7.723195175075135e-07,
      "loss": 0.1684,
      "reward": -0.10687501262873411,
      "reward_std": 1.1036955416202545,
      "rewards/correctness_reward_func": 0.2678571566939354,
      "rewards/int_reward_func": 0.2500000111758709,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.624732181429863,
      "step": 125
    },
    {
      "completion_length": 515.8571624755859,
      "epoch": 0.47191011235955055,
      "grad_norm": 2.7980728149414062,
      "kl": 4.34130859375,
      "learning_rate": 7.668072579578058e-07,
      "loss": 0.1737,
      "reward": -0.22420312091708183,
      "reward_std": 1.3303392231464386,
      "rewards/correctness_reward_func": 0.24107144214212894,
      "rewards/int_reward_func": 0.2912946566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7565692067146301,
      "step": 126
    },
    {
      "completion_length": 515.3928833007812,
      "epoch": 0.4756554307116105,
      "grad_norm": 1.5911880731582642,
      "kl": 5.1962890625,
      "learning_rate": 7.612492823579744e-07,
      "loss": 0.2078,
      "reward": -0.11461607180535793,
      "reward_std": 1.3034924268722534,
      "rewards/correctness_reward_func": 0.3303571566939354,
      "rewards/int_reward_func": 0.2968750074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7418482303619385,
      "step": 127
    },
    {
      "completion_length": 523.8928680419922,
      "epoch": 0.4794007490636704,
      "grad_norm": 2.3929038047790527,
      "kl": 5.15625,
      "learning_rate": 7.556465430385259e-07,
      "loss": 0.2062,
      "reward": -0.04959822096861899,
      "reward_std": 1.3580511510372162,
      "rewards/correctness_reward_func": 0.3482143059372902,
      "rewards/int_reward_func": 0.2957589440047741,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6935714930295944,
      "step": 128
    },
    {
      "completion_length": 514.5357513427734,
      "epoch": 0.48314606741573035,
      "grad_norm": 2.6407032012939453,
      "kl": 4.9072265625,
      "learning_rate": 7.5e-07,
      "loss": 0.1963,
      "reward": -0.11345758475363255,
      "reward_std": 1.2759953737258911,
      "rewards/correctness_reward_func": 0.2946428656578064,
      "rewards/int_reward_func": 0.2845982238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6926987171173096,
      "step": 129
    },
    {
      "completion_length": 515.6428985595703,
      "epoch": 0.4868913857677903,
      "grad_norm": 1.820491909980774,
      "kl": 4.42724609375,
      "learning_rate": 7.443106207484775e-07,
      "loss": 0.1771,
      "reward": -0.08515847939997911,
      "reward_std": 1.2134381830692291,
      "rewards/correctness_reward_func": 0.2946428693830967,
      "rewards/int_reward_func": 0.2656250074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6454263776540756,
      "step": 130
    },
    {
      "completion_length": 518.107177734375,
      "epoch": 0.49063670411985016,
      "grad_norm": 3.0650475025177,
      "kl": 4.8212890625,
      "learning_rate": 7.385793801298042e-07,
      "loss": 0.1929,
      "reward": -0.18281920308072586,
      "reward_std": 1.3228217363357544,
      "rewards/correctness_reward_func": 0.2544642984867096,
      "rewards/int_reward_func": 0.294642873108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.731926366686821,
      "step": 131
    },
    {
      "completion_length": 516.1786041259766,
      "epoch": 0.4943820224719101,
      "grad_norm": 2.8371877670288086,
      "kl": 5.048828125,
      "learning_rate": 7.328072601625557e-07,
      "loss": 0.202,
      "reward": -0.03922321368008852,
      "reward_std": 1.0432914346456528,
      "rewards/correctness_reward_func": 0.2098214365541935,
      "rewards/int_reward_func": 0.2912946566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.540339283645153,
      "step": 132
    },
    {
      "completion_length": 530.3571624755859,
      "epoch": 0.49812734082397003,
      "grad_norm": 2.9099578857421875,
      "kl": 4.8974609375,
      "learning_rate": 7.269952498697734e-07,
      "loss": 0.1959,
      "reward": -0.1908794827759266,
      "reward_std": 1.1966271996498108,
      "rewards/correctness_reward_func": 0.1741071492433548,
      "rewards/int_reward_func": 0.2767857238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6417723447084427,
      "step": 133
    },
    {
      "completion_length": 537.1786193847656,
      "epoch": 0.50187265917603,
      "grad_norm": 1.9121432304382324,
      "kl": 4.748046875,
      "learning_rate": 7.211443451095006e-07,
      "loss": 0.1899,
      "reward": -0.07099777227267623,
      "reward_std": 1.2925509214401245,
      "rewards/correctness_reward_func": 0.2991071566939354,
      "rewards/int_reward_func": 0.282366082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6524710208177567,
      "step": 134
    },
    {
      "completion_length": 531.8214569091797,
      "epoch": 0.5056179775280899,
      "grad_norm": 2.4540979862213135,
      "kl": 5.0126953125,
      "learning_rate": 7.152555484041475e-07,
      "loss": 0.2005,
      "reward": -0.19015179947018623,
      "reward_std": 1.273418515920639,
      "rewards/correctness_reward_func": 0.2500000074505806,
      "rewards/int_reward_func": 0.2823660895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7225178927183151,
      "step": 135
    },
    {
      "epoch": 0.5056179775280899,
      "eval_completion_length": 534.5113062808388,
      "eval_kl": Infinity,
      "eval_loss": Infinity,
      "eval_reward": -0.1988269823652349,
      "eval_reward_std": 1.2561254689567967,
      "eval_rewards/correctness_reward_func": 0.2170739448384235,
      "eval_rewards/int_reward_func": 0.2772791479763232,
      "eval_rewards/soft_format_reward_func": 4.6992483303735133e-05,
      "eval_rewards/strict_format_reward_func": 2.3496241651867567e-05,
      "eval_rewards/xmlcount_reward_func": -0.6932505579371201,
      "eval_runtime": 2478.2667,
      "eval_samples_per_second": 0.532,
      "eval_steps_per_second": 0.038,
      "step": 135
    },
    {
      "completion_length": 517.3571472167969,
      "epoch": 0.5093632958801498,
      "grad_norm": 2.260101079940796,
      "kl": 4.990234375,
      "learning_rate": 7.09329868768714e-07,
      "loss": 0.1996,
      "reward": -0.14810045702324715,
      "reward_std": 1.2247705459594727,
      "rewards/correctness_reward_func": 0.2410714402794838,
      "rewards/int_reward_func": 0.2812500149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.6715379506349564,
      "step": 136
    },
    {
      "completion_length": 520.4286041259766,
      "epoch": 0.5131086142322098,
      "grad_norm": 1.816656231880188,
      "kl": 5.271484375,
      "learning_rate": 7.033683215379002e-07,
      "loss": 0.2109,
      "reward": -0.2372276969254017,
      "reward_std": 1.2999887466430664,
      "rewards/correctness_reward_func": 0.2500000074505806,
      "rewards/int_reward_func": 0.2745535895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.7628973424434662,
      "step": 137
    },
    {
      "completion_length": 516.1428833007812,
      "epoch": 0.5168539325842697,
      "grad_norm": 3.60597825050354,
      "kl": 5.40625,
      "learning_rate": 6.973719281921336e-07,
      "loss": 0.2163,
      "reward": -0.1252053566277027,
      "reward_std": 1.2438539862632751,
      "rewards/correctness_reward_func": 0.3035714402794838,
      "rewards/int_reward_func": 0.2812500111758709,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7100268006324768,
      "step": 138
    },
    {
      "completion_length": 518.8928833007812,
      "epoch": 0.5205992509363296,
      "grad_norm": 2.287635326385498,
      "kl": 5.2548828125,
      "learning_rate": 6.913417161825449e-07,
      "loss": 0.2102,
      "reward": -0.1885022409260273,
      "reward_std": 1.2898488640785217,
      "rewards/correctness_reward_func": 0.2812500149011612,
      "rewards/int_reward_func": 0.2879464402794838,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7588147670030594,
      "step": 139
    },
    {
      "completion_length": 516.5357360839844,
      "epoch": 0.5243445692883895,
      "grad_norm": 1.85598623752594,
      "kl": 4.8173828125,
      "learning_rate": 6.852787187549181e-07,
      "loss": 0.1926,
      "reward": -0.08186161960475147,
      "reward_std": 1.207483857870102,
      "rewards/correctness_reward_func": 0.2946428693830967,
      "rewards/int_reward_func": 0.266741082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6432455480098724,
      "step": 140
    },
    {
      "completion_length": 519.7500305175781,
      "epoch": 0.5280898876404494,
      "grad_norm": 2.6306166648864746,
      "kl": 4.4951171875,
      "learning_rate": 6.7918397477265e-07,
      "loss": 0.1798,
      "reward": -0.12313839327543974,
      "reward_std": 1.155171424150467,
      "rewards/correctness_reward_func": 0.2232142984867096,
      "rewards/int_reward_func": 0.2656250149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6119777262210846,
      "step": 141
    },
    {
      "completion_length": 516.3928833007812,
      "epoch": 0.5318352059925093,
      "grad_norm": 2.6810405254364014,
      "kl": 5.0771484375,
      "learning_rate": 6.730585285387465e-07,
      "loss": 0.2031,
      "reward": -0.022080346941947937,
      "reward_std": 1.2572471797466278,
      "rewards/correctness_reward_func": 0.3660714402794838,
      "rewards/int_reward_func": 0.2979910895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6861428916454315,
      "step": 142
    },
    {
      "completion_length": 522.3214569091797,
      "epoch": 0.5355805243445693,
      "grad_norm": 3.570694923400879,
      "kl": 4.92724609375,
      "learning_rate": 6.669034296168854e-07,
      "loss": 0.1971,
      "reward": -0.18164288252592087,
      "reward_std": 1.2775076925754547,
      "rewards/correctness_reward_func": 0.3035714477300644,
      "rewards/int_reward_func": 0.266741082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7519554197788239,
      "step": 143
    },
    {
      "completion_length": 533.3571624755859,
      "epoch": 0.5393258426966292,
      "grad_norm": 1.905169129371643,
      "kl": 4.7041015625,
      "learning_rate": 6.607197326515807e-07,
      "loss": 0.1881,
      "reward": -0.1941227694042027,
      "reward_std": 1.394089788198471,
      "rewards/correctness_reward_func": 0.325892873108387,
      "rewards/int_reward_func": 0.2600446492433548,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7800603359937668,
      "step": 144
    },
    {
      "completion_length": 533.7500152587891,
      "epoch": 0.5430711610486891,
      "grad_norm": 1.7941057682037354,
      "kl": 4.5029296875,
      "learning_rate": 6.545084971874736e-07,
      "loss": 0.1801,
      "reward": -0.1683393083512783,
      "reward_std": 1.3101599216461182,
      "rewards/correctness_reward_func": 0.28571430779993534,
      "rewards/int_reward_func": 0.274553582072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.728607177734375,
      "step": 145
    },
    {
      "completion_length": 521.607177734375,
      "epoch": 0.5468164794007491,
      "grad_norm": 2.9071526527404785,
      "kl": 4.875,
      "learning_rate": 6.482707874877854e-07,
      "loss": 0.1949,
      "reward": -0.04411383857950568,
      "reward_std": 1.3005799651145935,
      "rewards/correctness_reward_func": 0.3169642984867096,
      "rewards/int_reward_func": 0.3035714440047741,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6646495461463928,
      "step": 146
    },
    {
      "completion_length": 544.9643096923828,
      "epoch": 0.550561797752809,
      "grad_norm": 2.726607322692871,
      "kl": Infinity,
      "learning_rate": 6.420076723519614e-07,
      "loss": 0.22,
      "reward": -0.0854665283113718,
      "reward_std": 1.307901293039322,
      "rewards/correctness_reward_func": 0.401785746216774,
      "rewards/int_reward_func": 0.277901791036129,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7662701457738876,
      "step": 147
    },
    {
      "completion_length": 532.1428833007812,
      "epoch": 0.5543071161048689,
      "grad_norm": 2.9170711040496826,
      "kl": 5.365234375,
      "learning_rate": 6.357202249325371e-07,
      "loss": 0.2146,
      "reward": -0.1308973114937544,
      "reward_std": 1.3350401818752289,
      "rewards/correctness_reward_func": 0.3035714402794838,
      "rewards/int_reward_func": 0.2790178656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7134866267442703,
      "step": 148
    },
    {
      "completion_length": 518.2857360839844,
      "epoch": 0.5580524344569289,
      "grad_norm": 2.2603960037231445,
      "kl": 5.1748046875,
      "learning_rate": 6.294095225512604e-07,
      "loss": 0.207,
      "reward": -0.18937501218169928,
      "reward_std": 1.3142652213573456,
      "rewards/correctness_reward_func": 0.2767857238650322,
      "rewards/int_reward_func": 0.305803582072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7719643265008926,
      "step": 149
    },
    {
      "completion_length": 531.0000457763672,
      "epoch": 0.5617977528089888,
      "grad_norm": 2.677884578704834,
      "kl": 5.2392578125,
      "learning_rate": 6.230766465144965e-07,
      "loss": 0.2096,
      "reward": -0.1788214324042201,
      "reward_std": 1.2578088641166687,
      "rewards/correctness_reward_func": 0.2455357238650322,
      "rewards/int_reward_func": 0.2667410857975483,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.69221431016922,
      "step": 150
    },
    {
      "completion_length": 523.357177734375,
      "epoch": 0.5655430711610487,
      "grad_norm": 2.8838837146759033,
      "kl": 5.416015625,
      "learning_rate": 6.167226819279527e-07,
      "loss": 0.2166,
      "reward": -0.17594197671860456,
      "reward_std": 1.438678652048111,
      "rewards/correctness_reward_func": 0.325892873108387,
      "rewards/int_reward_func": 0.305803582072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8076384514570236,
      "step": 151
    },
    {
      "completion_length": 528.3571624755859,
      "epoch": 0.5692883895131086,
      "grad_norm": 4.809299945831299,
      "kl": 5.3623046875,
      "learning_rate": 6.103487175107507e-07,
      "loss": 0.2145,
      "reward": -0.18464063061401248,
      "reward_std": 1.3211232721805573,
      "rewards/correctness_reward_func": 0.3035714402794838,
      "rewards/int_reward_func": 0.314732164144516,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8029441684484482,
      "step": 152
    },
    {
      "completion_length": 536.7500457763672,
      "epoch": 0.5730337078651685,
      "grad_norm": 1.6935267448425293,
      "kl": 4.90625,
      "learning_rate": 6.039558454088795e-07,
      "loss": 0.1962,
      "reward": -0.1592745715752244,
      "reward_std": 1.2062239050865173,
      "rewards/correctness_reward_func": 0.28125001955777407,
      "rewards/int_reward_func": 0.3002232238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7407478094100952,
      "step": 153
    },
    {
      "completion_length": 515.2500152587891,
      "epoch": 0.5767790262172284,
      "grad_norm": 2.8796603679656982,
      "kl": 4.6162109375,
      "learning_rate": 5.975451610080642e-07,
      "loss": 0.1846,
      "reward": -0.14308037795126438,
      "reward_std": 1.2807807624340057,
      "rewards/correctness_reward_func": 0.2767857238650322,
      "rewards/int_reward_func": 0.2656250149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6854911148548126,
      "step": 154
    },
    {
      "completion_length": 518.0714569091797,
      "epoch": 0.5805243445692884,
      "grad_norm": 2.875056505203247,
      "kl": 4.8203125,
      "learning_rate": 5.911177627460738e-07,
      "loss": 0.1928,
      "reward": -0.14478794857859612,
      "reward_std": 1.3489331901073456,
      "rewards/correctness_reward_func": 0.316964291036129,
      "rewards/int_reward_func": 0.286830373108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7485826313495636,
      "step": 155
    },
    {
      "completion_length": 526.2143096923828,
      "epoch": 0.5842696629213483,
      "grad_norm": 2.2693309783935547,
      "kl": 5.466796875,
      "learning_rate": 5.846747519245122e-07,
      "loss": 0.2187,
      "reward": -0.18839732877677307,
      "reward_std": 1.3681853413581848,
      "rewards/correctness_reward_func": 0.3437500223517418,
      "rewards/int_reward_func": 0.286830373108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8189777284860611,
      "step": 156
    },
    {
      "completion_length": 518.7143096923828,
      "epoch": 0.5880149812734082,
      "grad_norm": 3.1548948287963867,
      "kl": 5.4326171875,
      "learning_rate": 5.782172325201155e-07,
      "loss": 0.2173,
      "reward": -0.11439956957474351,
      "reward_std": 1.17717644572258,
      "rewards/correctness_reward_func": 0.294642873108387,
      "rewards/int_reward_func": 0.2767857238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6858281642198563,
      "step": 157
    },
    {
      "completion_length": 528.7143096923828,
      "epoch": 0.5917602996254682,
      "grad_norm": 2.895555019378662,
      "kl": 5.6181640625,
      "learning_rate": 5.717463109955895e-07,
      "loss": 0.2247,
      "reward": -0.08654686715453863,
      "reward_std": 1.2630547881126404,
      "rewards/correctness_reward_func": 0.2678571566939354,
      "rewards/int_reward_func": 0.2801339402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6345379799604416,
      "step": 158
    },
    {
      "completion_length": 519.7500305175781,
      "epoch": 0.5955056179775281,
      "grad_norm": 2.325378894805908,
      "kl": 5.177734375,
      "learning_rate": 5.652630961100258e-07,
      "loss": 0.2071,
      "reward": -0.05345983011648059,
      "reward_std": 1.290926218032837,
      "rewards/correctness_reward_func": 0.4107143133878708,
      "rewards/int_reward_func": 0.2834821566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7476563155651093,
      "step": 159
    },
    {
      "completion_length": 518.357177734375,
      "epoch": 0.599250936329588,
      "grad_norm": 2.012712240219116,
      "kl": 5.064453125,
      "learning_rate": 5.587686987289189e-07,
      "loss": 0.2026,
      "reward": -0.20887724682688713,
      "reward_std": 1.2406348288059235,
      "rewards/correctness_reward_func": 0.227678582072258,
      "rewards/int_reward_func": 0.2767857164144516,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.713341549038887,
      "step": 160
    },
    {
      "completion_length": 525.3928833007812,
      "epoch": 0.602996254681648,
      "grad_norm": 2.0301644802093506,
      "kl": 4.5810546875,
      "learning_rate": 5.522642316338268e-07,
      "loss": 0.1832,
      "reward": -0.15497322753071785,
      "reward_std": 1.2159148454666138,
      "rewards/correctness_reward_func": 0.2232142984867096,
      "rewards/int_reward_func": 0.2723214402794838,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.651625007390976,
      "step": 161
    },
    {
      "completion_length": 525.3571472167969,
      "epoch": 0.6067415730337079,
      "grad_norm": 2.222111463546753,
      "kl": 4.67431640625,
      "learning_rate": 5.457508093317013e-07,
      "loss": 0.1869,
      "reward": -0.09700893424451351,
      "reward_std": 1.1505461633205414,
      "rewards/correctness_reward_func": 0.2098214365541935,
      "rewards/int_reward_func": 0.2500000074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5568303763866425,
      "step": 162
    },
    {
      "epoch": 0.6067415730337079,
      "eval_completion_length": 531.0331080386513,
      "eval_kl": Infinity,
      "eval_loss": Infinity,
      "eval_reward": -0.19856291947102075,
      "eval_reward_std": 1.2412375688552857,
      "eval_rewards/correctness_reward_func": 0.2193609140421215,
      "eval_rewards/int_reward_func": 0.27259556368777627,
      "eval_rewards/soft_format_reward_func": 0.00018796993321494053,
      "eval_rewards/strict_format_reward_func": 0.00011748120825933783,
      "eval_rewards/xmlcount_reward_func": -0.6908248430804202,
      "eval_runtime": 2481.1198,
      "eval_samples_per_second": 0.532,
      "eval_steps_per_second": 0.038,
      "step": 162
    },
    {
      "completion_length": 521.6428680419922,
      "epoch": 0.6104868913857678,
      "grad_norm": 1.9306232929229736,
      "kl": 4.5283203125,
      "learning_rate": 5.392295478639225e-07,
      "loss": 0.1812,
      "reward": -0.2653035670518875,
      "reward_std": 1.287449061870575,
      "rewards/correctness_reward_func": 0.2232142947614193,
      "rewards/int_reward_func": 0.2834821566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7720000445842743,
      "step": 163
    },
    {
      "completion_length": 520.1428833007812,
      "epoch": 0.6142322097378277,
      "grad_norm": 2.5203397274017334,
      "kl": 5.091796875,
      "learning_rate": 5.327015646150716e-07,
      "loss": 0.2037,
      "reward": 0.042263399343937635,
      "reward_std": 1.2937269806861877,
      "rewards/correctness_reward_func": 0.3437500149011612,
      "rewards/int_reward_func": 0.2924107313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.593897320330143,
      "step": 164
    },
    {
      "completion_length": 519.8214569091797,
      "epoch": 0.6179775280898876,
      "grad_norm": 1.5657882690429688,
      "kl": 4.630859375,
      "learning_rate": 5.26167978121472e-07,
      "loss": 0.1852,
      "reward": -0.22304466180503368,
      "reward_std": 1.2659655511379242,
      "rewards/correctness_reward_func": 0.2633928768336773,
      "rewards/int_reward_func": 0.2767857238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7632232308387756,
      "step": 165
    },
    {
      "completion_length": 524.0357360839844,
      "epoch": 0.6217228464419475,
      "grad_norm": 2.093134880065918,
      "kl": 4.900390625,
      "learning_rate": 5.196299078795343e-07,
      "loss": 0.196,
      "reward": -0.27018305100500584,
      "reward_std": 1.325235664844513,
      "rewards/correctness_reward_func": 0.2723214365541935,
      "rewards/int_reward_func": 0.2845982313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8271027356386185,
      "step": 166
    },
    {
      "completion_length": 524.6428833007812,
      "epoch": 0.6254681647940075,
      "grad_norm": 2.1078877449035645,
      "kl": 5.291015625,
      "learning_rate": 5.130884741539366e-07,
      "loss": 0.2116,
      "reward": -0.12524107471108437,
      "reward_std": 1.3548560738563538,
      "rewards/correctness_reward_func": 0.2991071566939354,
      "rewards/int_reward_func": 0.3191964402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.743544690310955,
      "step": 167
    },
    {
      "completion_length": 520.4643096923828,
      "epoch": 0.6292134831460674,
      "grad_norm": 3.086167812347412,
      "kl": 5.5380859375,
      "learning_rate": 5.065447977856722e-07,
      "loss": 0.2215,
      "reward": -0.21059376862831414,
      "reward_std": 1.2520826160907745,
      "rewards/correctness_reward_func": 0.21428572107106447,
      "rewards/int_reward_func": 0.2656250111758709,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6905045062303543,
      "step": 168
    },
    {
      "completion_length": 526.7143096923828,
      "epoch": 0.6329588014981273,
      "grad_norm": 1.7510619163513184,
      "kl": 5.1826171875,
      "learning_rate": 5e-07,
      "loss": 0.2073,
      "reward": -0.1397678554058075,
      "reward_std": 1.1866637468338013,
      "rewards/correctness_reward_func": 0.2232142947614193,
      "rewards/int_reward_func": 0.2734375111758709,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6364196613430977,
      "step": 169
    },
    {
      "completion_length": 514.8571624755859,
      "epoch": 0.6367041198501873,
      "grad_norm": 2.065687894821167,
      "kl": 4.8115234375,
      "learning_rate": 4.934552022143279e-07,
      "loss": 0.1925,
      "reward": -0.3015558011829853,
      "reward_std": 1.4401615262031555,
      "rewards/correctness_reward_func": 0.2991071566939354,
      "rewards/int_reward_func": 0.2801339477300644,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8807969093322754,
      "step": 170
    },
    {
      "completion_length": 517.0357208251953,
      "epoch": 0.6404494382022472,
      "grad_norm": 2.2018256187438965,
      "kl": 4.5322265625,
      "learning_rate": 4.869115258460634e-07,
      "loss": 0.1813,
      "reward": -0.06244420446455479,
      "reward_std": 1.273977816104889,
      "rewards/correctness_reward_func": 0.294642873108387,
      "rewards/int_reward_func": 0.2756696529686451,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.632756732404232,
      "step": 171
    },
    {
      "completion_length": 531.3214569091797,
      "epoch": 0.6441947565543071,
      "grad_norm": 2.4541852474212646,
      "kl": 4.6171875,
      "learning_rate": 4.803700921204658e-07,
      "loss": 0.1847,
      "reward": 0.0766785740852356,
      "reward_std": 1.2054692655801773,
      "rewards/correctness_reward_func": 0.3526785895228386,
      "rewards/int_reward_func": 0.2991071566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5751071497797966,
      "step": 172
    },
    {
      "completion_length": 526.0714569091797,
      "epoch": 0.6479400749063671,
      "grad_norm": 2.2811288833618164,
      "kl": 4.4521484375,
      "learning_rate": 4.7383202187852804e-07,
      "loss": 0.1781,
      "reward": -0.08707589283585548,
      "reward_std": 1.3148585855960846,
      "rewards/correctness_reward_func": 0.3705357387661934,
      "rewards/int_reward_func": 0.3046875074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.762299120426178,
      "step": 173
    },
    {
      "completion_length": 517.8214416503906,
      "epoch": 0.651685393258427,
      "grad_norm": 5.915732383728027,
      "kl": 5.205078125,
      "learning_rate": 4.672984353849284e-07,
      "loss": 0.2081,
      "reward": -0.024636156857013702,
      "reward_std": 1.2718493938446045,
      "rewards/correctness_reward_func": 0.4107143096625805,
      "rewards/int_reward_func": 0.3125000074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7478504627943039,
      "step": 174
    },
    {
      "completion_length": 517.0714569091797,
      "epoch": 0.6554307116104869,
      "grad_norm": 2.853475332260132,
      "kl": 5.4931640625,
      "learning_rate": 4.6077045213607755e-07,
      "loss": 0.2197,
      "reward": -0.31054020673036575,
      "reward_std": 1.394624948501587,
      "rewards/correctness_reward_func": 0.2455357201397419,
      "rewards/int_reward_func": 0.2979910895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8540670275688171,
      "step": 175
    },
    {
      "completion_length": 525.5714569091797,
      "epoch": 0.6591760299625468,
      "grad_norm": 1.7146692276000977,
      "kl": 4.740234375,
      "learning_rate": 4.542491906682988e-07,
      "loss": 0.1896,
      "reward": -0.13947322126477957,
      "reward_std": 1.1648866534233093,
      "rewards/correctness_reward_func": 0.2053571566939354,
      "rewards/int_reward_func": 0.2566964328289032,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6015268266201019,
      "step": 176
    },
    {
      "completion_length": 521.2143249511719,
      "epoch": 0.6629213483146067,
      "grad_norm": 1.6346399784088135,
      "kl": 4.8623046875,
      "learning_rate": 4.477357683661733e-07,
      "loss": 0.1945,
      "reward": -0.0936138320248574,
      "reward_std": 1.2343957126140594,
      "rewards/correctness_reward_func": 0.3392857313156128,
      "rewards/int_reward_func": 0.286830373108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7197299301624298,
      "step": 177
    },
    {
      "completion_length": 522.5000305175781,
      "epoch": 0.6666666666666666,
      "grad_norm": 1.850631594657898,
      "kl": 5.2626953125,
      "learning_rate": 4.412313012710812e-07,
      "loss": 0.2105,
      "reward": -0.10290402628015727,
      "reward_std": 1.2204317450523376,
      "rewards/correctness_reward_func": 0.2008928656578064,
      "rewards/int_reward_func": 0.2578125037252903,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5616094022989273,
      "step": 178
    },
    {
      "completion_length": 534.1785888671875,
      "epoch": 0.6704119850187266,
      "grad_norm": 2.1959705352783203,
      "kl": 5.53125,
      "learning_rate": 4.347369038899743e-07,
      "loss": 0.2212,
      "reward": -0.37369197979569435,
      "reward_std": 1.324585199356079,
      "rewards/correctness_reward_func": 0.2053571529686451,
      "rewards/int_reward_func": 0.2622767984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8413259387016296,
      "step": 179
    },
    {
      "completion_length": 520.8571624755859,
      "epoch": 0.6741573033707865,
      "grad_norm": 1.6929233074188232,
      "kl": 5.0546875,
      "learning_rate": 4.2825368900441037e-07,
      "loss": 0.2022,
      "reward": -0.1256919801235199,
      "reward_std": 1.4035832583904266,
      "rewards/correctness_reward_func": 0.3348214477300644,
      "rewards/int_reward_func": 0.2979910895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7585045099258423,
      "step": 180
    },
    {
      "completion_length": 528.6786041259766,
      "epoch": 0.6779026217228464,
      "grad_norm": 1.5895494222640991,
      "kl": 4.9208984375,
      "learning_rate": 4.2178276747988444e-07,
      "loss": 0.1968,
      "reward": -0.05114509118720889,
      "reward_std": 1.2217027842998505,
      "rewards/correctness_reward_func": 0.3348214477300644,
      "rewards/int_reward_func": 0.2935267984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.679493322968483,
      "step": 181
    },
    {
      "completion_length": 527.107177734375,
      "epoch": 0.6816479400749064,
      "grad_norm": 1.8839842081069946,
      "kl": 4.43359375,
      "learning_rate": 4.153252480754877e-07,
      "loss": 0.1774,
      "reward": -0.16289733722805977,
      "reward_std": 1.2423360645771027,
      "rewards/correctness_reward_func": 0.2812500149011612,
      "rewards/int_reward_func": 0.2879464440047741,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7332098335027695,
      "step": 182
    },
    {
      "completion_length": 516.9285888671875,
      "epoch": 0.6853932584269663,
      "grad_norm": 2.7923595905303955,
      "kl": 4.537109375,
      "learning_rate": 4.0888223725392624e-07,
      "loss": 0.1815,
      "reward": -0.19774777255952358,
      "reward_std": 1.229084461927414,
      "rewards/correctness_reward_func": 0.2232142984867096,
      "rewards/int_reward_func": 0.2689732201397419,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6899353265762329,
      "step": 183
    },
    {
      "completion_length": 522.357177734375,
      "epoch": 0.6891385767790262,
      "grad_norm": 1.9454573392868042,
      "kl": 4.69140625,
      "learning_rate": 4.0245483899193586e-07,
      "loss": 0.1877,
      "reward": -0.15699554327875376,
      "reward_std": 1.1800599992275238,
      "rewards/correctness_reward_func": 0.2633928693830967,
      "rewards/int_reward_func": 0.3147321566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7351205945014954,
      "step": 184
    },
    {
      "completion_length": 533.1786041259766,
      "epoch": 0.6928838951310862,
      "grad_norm": 1.6554526090621948,
      "kl": 4.5224609375,
      "learning_rate": 3.960441545911204e-07,
      "loss": 0.181,
      "reward": -0.07045312505215406,
      "reward_std": 1.3237174451351166,
      "rewards/correctness_reward_func": 0.3660714402794838,
      "rewards/int_reward_func": 0.290178582072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7267031669616699,
      "step": 185
    },
    {
      "completion_length": 518.2500305175781,
      "epoch": 0.6966292134831461,
      "grad_norm": 2.1547718048095703,
      "kl": 4.7314453125,
      "learning_rate": 3.896512824892495e-07,
      "loss": 0.1893,
      "reward": -0.05010267719626427,
      "reward_std": 1.0627744495868683,
      "rewards/correctness_reward_func": 0.2544642984867096,
      "rewards/int_reward_func": 0.2890625149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5936295166611671,
      "step": 186
    },
    {
      "completion_length": 520.6428833007812,
      "epoch": 0.700374531835206,
      "grad_norm": 2.404644250869751,
      "kl": 4.72265625,
      "learning_rate": 3.8327731807204744e-07,
      "loss": 0.1889,
      "reward": -0.00044643133878707886,
      "reward_std": 1.2751690447330475,
      "rewards/correctness_reward_func": 0.3660714402794838,
      "rewards/int_reward_func": 0.2991071566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6656250506639481,
      "step": 187
    },
    {
      "completion_length": 527.7143096923828,
      "epoch": 0.704119850187266,
      "grad_norm": 2.1762359142303467,
      "kl": 4.845703125,
      "learning_rate": 3.7692335348550346e-07,
      "loss": 0.1938,
      "reward": -0.019415173679590225,
      "reward_std": 1.4023870527744293,
      "rewards/correctness_reward_func": 0.4375000074505806,
      "rewards/int_reward_func": 0.3013392984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7582544833421707,
      "step": 188
    },
    {
      "completion_length": 516.1428833007812,
      "epoch": 0.7078651685393258,
      "grad_norm": 1.9605624675750732,
      "kl": 4.767578125,
      "learning_rate": 3.7059047744873955e-07,
      "loss": 0.1907,
      "reward": -0.28092633839696646,
      "reward_std": 1.2700144350528717,
      "rewards/correctness_reward_func": 0.2455357238650322,
      "rewards/int_reward_func": 0.2857142984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8121763616800308,
      "step": 189
    },
    {
      "epoch": 0.7078651685393258,
      "eval_completion_length": 534.5684467516447,
      "eval_kl": Infinity,
      "eval_loss": Infinity,
      "eval_reward": -0.19735878194976403,
      "eval_reward_std": 1.251351103029753,
      "eval_rewards/correctness_reward_func": 0.2117481308548074,
      "eval_rewards/int_reward_func": 0.2752584706795843,
      "eval_rewards/soft_format_reward_func": 4.6992483303735133e-05,
      "eval_rewards/strict_format_reward_func": 7.048872495560269e-05,
      "eval_rewards/xmlcount_reward_func": -0.6844828577418076,
      "eval_runtime": 2493.3052,
      "eval_samples_per_second": 0.529,
      "eval_steps_per_second": 0.038,
      "step": 189
    },
    {
      "completion_length": 519.7143096923828,
      "epoch": 0.7116104868913857,
      "grad_norm": 2.0175857543945312,
      "kl": 4.990234375,
      "learning_rate": 3.642797750674629e-07,
      "loss": 0.1996,
      "reward": -0.04893304128199816,
      "reward_std": 1.230804204940796,
      "rewards/correctness_reward_func": 0.3258928693830967,
      "rewards/int_reward_func": 0.3058035895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6806294620037079,
      "step": 190
    },
    {
      "completion_length": 524.7857360839844,
      "epoch": 0.7153558052434457,
      "grad_norm": 1.9237942695617676,
      "kl": 5.12890625,
      "learning_rate": 3.5799232764803867e-07,
      "loss": 0.2052,
      "reward": -0.05077901855111122,
      "reward_std": 1.1569238305091858,
      "rewards/correctness_reward_func": 0.2232142984867096,
      "rewards/int_reward_func": 0.2767857275903225,
      "rewards/soft_format_reward_func": 0.0022321429569274187,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5530111789703369,
      "step": 191
    },
    {
      "completion_length": 515.8928833007812,
      "epoch": 0.7191011235955056,
      "grad_norm": 2.954789876937866,
      "kl": 5.5703125,
      "learning_rate": 3.517292125122145e-07,
      "loss": 0.2228,
      "reward": -0.22044420626480132,
      "reward_std": 1.381562888622284,
      "rewards/correctness_reward_func": 0.366071455180645,
      "rewards/int_reward_func": 0.2712053656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8577210158109665,
      "step": 192
    },
    {
      "completion_length": 518.7143096923828,
      "epoch": 0.7228464419475655,
      "grad_norm": 2.220942735671997,
      "kl": 5.3818359375,
      "learning_rate": 3.454915028125263e-07,
      "loss": 0.2153,
      "reward": -0.010651785880327225,
      "reward_std": 1.300764262676239,
      "rewards/correctness_reward_func": 0.3526785895228386,
      "rewards/int_reward_func": 0.302455373108387,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6669018268585205,
      "step": 193
    },
    {
      "completion_length": 523.6071624755859,
      "epoch": 0.7265917602996255,
      "grad_norm": 1.9517362117767334,
      "kl": 4.9345703125,
      "learning_rate": 3.392802673484193e-07,
      "loss": 0.1974,
      "reward": -0.18442187644541264,
      "reward_std": 1.2702946960926056,
      "rewards/correctness_reward_func": 0.2723214440047741,
      "rewards/int_reward_func": 0.2700893022119999,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7268326282501221,
      "step": 194
    },
    {
      "completion_length": 544.4643249511719,
      "epoch": 0.7303370786516854,
      "grad_norm": 2.2297322750091553,
      "kl": 4.875,
      "learning_rate": 3.330965703831146e-07,
      "loss": 0.195,
      "reward": -0.08661830611526966,
      "reward_std": 1.2692251205444336,
      "rewards/correctness_reward_func": 0.3348214477300644,
      "rewards/int_reward_func": 0.2901785895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7116183340549469,
      "step": 195
    },
    {
      "completion_length": 521.3928833007812,
      "epoch": 0.7340823970037453,
      "grad_norm": 3.919473648071289,
      "kl": 5.24609375,
      "learning_rate": 3.269414714612534e-07,
      "loss": 0.2098,
      "reward": -0.2262209877371788,
      "reward_std": 1.326324224472046,
      "rewards/correctness_reward_func": 0.196428582072258,
      "rewards/int_reward_func": 0.2500000074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6726495623588562,
      "step": 196
    },
    {
      "completion_length": 517.5714569091797,
      "epoch": 0.7378277153558053,
      "grad_norm": 2.9488518238067627,
      "kl": 5.2353515625,
      "learning_rate": 3.2081602522734985e-07,
      "loss": 0.2094,
      "reward": -0.3113616332411766,
      "reward_std": 1.211115837097168,
      "rewards/correctness_reward_func": 0.2008928656578064,
      "rewards/int_reward_func": 0.2712053656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7834598571062088,
      "step": 197
    },
    {
      "completion_length": 515.5000152587891,
      "epoch": 0.7415730337078652,
      "grad_norm": 6.636002063751221,
      "kl": 4.9560546875,
      "learning_rate": 3.147212812450818e-07,
      "loss": 0.1982,
      "reward": -0.09657144173979759,
      "reward_std": 1.3694335520267487,
      "rewards/correctness_reward_func": 0.4151785895228386,
      "rewards/int_reward_func": 0.3158482313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8275982290506363,
      "step": 198
    },
    {
      "completion_length": 532.4643096923828,
      "epoch": 0.7453183520599251,
      "grad_norm": 2.0196456909179688,
      "kl": 4.9970703125,
      "learning_rate": 3.086582838174551e-07,
      "loss": 0.1998,
      "reward": -0.273979929741472,
      "reward_std": 1.3600831031799316,
      "rewards/correctness_reward_func": 0.2455357201397419,
      "rewards/int_reward_func": 0.2611607275903225,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7806763797998428,
      "step": 199
    },
    {
      "completion_length": 532.4286041259766,
      "epoch": 0.7490636704119851,
      "grad_norm": 1.798177719116211,
      "kl": 4.6240234375,
      "learning_rate": 3.026280718078664e-07,
      "loss": 0.185,
      "reward": -0.26691294088959694,
      "reward_std": 1.3088684678077698,
      "rewards/correctness_reward_func": 0.2500000111758709,
      "rewards/int_reward_func": 0.2633928656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7803058326244354,
      "step": 200
    },
    {
      "completion_length": 516.8928985595703,
      "epoch": 0.7528089887640449,
      "grad_norm": 1.977175235748291,
      "kl": 5.1142578125,
      "learning_rate": 2.9663167846209996e-07,
      "loss": 0.2046,
      "reward": -0.11764062941074371,
      "reward_std": 1.2654018104076385,
      "rewards/correctness_reward_func": 0.2455357275903225,
      "rewards/int_reward_func": 0.2734375149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6366139054298401,
      "step": 201
    },
    {
      "completion_length": 516.0357360839844,
      "epoch": 0.7565543071161048,
      "grad_norm": 2.0420053005218506,
      "kl": 4.185546875,
      "learning_rate": 2.906701312312861e-07,
      "loss": 0.1674,
      "reward": -0.1846384033560753,
      "reward_std": 1.2015181183815002,
      "rewards/correctness_reward_func": 0.2678571566939354,
      "rewards/int_reward_func": 0.2723214402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7248169928789139,
      "step": 202
    },
    {
      "completion_length": 527.607177734375,
      "epoch": 0.7602996254681648,
      "grad_norm": 2.0275211334228516,
      "kl": 4.865234375,
      "learning_rate": 2.847444515958523e-07,
      "loss": 0.1946,
      "reward": -0.0009955335408449173,
      "reward_std": 1.2537928223609924,
      "rewards/correctness_reward_func": 0.3794643059372902,
      "rewards/int_reward_func": 0.2979910746216774,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6795670092105865,
      "step": 203
    },
    {
      "completion_length": 532.7143096923828,
      "epoch": 0.7640449438202247,
      "grad_norm": 1.5860823392868042,
      "kl": 4.7626953125,
      "learning_rate": 2.7885565489049946e-07,
      "loss": 0.1905,
      "reward": -0.25017189234495163,
      "reward_std": 1.346979558467865,
      "rewards/correctness_reward_func": 0.2857142984867096,
      "rewards/int_reward_func": 0.286830373108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.822716549038887,
      "step": 204
    },
    {
      "completion_length": 514.9286041259766,
      "epoch": 0.7677902621722846,
      "grad_norm": 1.611587405204773,
      "kl": 4.9775390625,
      "learning_rate": 2.730047501302266e-07,
      "loss": 0.1991,
      "reward": -0.009238844271749258,
      "reward_std": 1.2853399515151978,
      "rewards/correctness_reward_func": 0.3660714477300644,
      "rewards/int_reward_func": 0.2667410895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6420513689517975,
      "step": 205
    },
    {
      "completion_length": 517.7500305175781,
      "epoch": 0.7715355805243446,
      "grad_norm": 2.1895291805267334,
      "kl": 5.3173828125,
      "learning_rate": 2.671927398374443e-07,
      "loss": 0.2127,
      "reward": -0.02679464779794216,
      "reward_std": 1.2472732365131378,
      "rewards/correctness_reward_func": 0.3526785895228386,
      "rewards/int_reward_func": 0.2912946566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6707678735256195,
      "step": 206
    },
    {
      "completion_length": 524.6071624755859,
      "epoch": 0.7752808988764045,
      "grad_norm": 2.1895291805267334,
      "kl": Infinity,
      "learning_rate": 2.671927398374443e-07,
      "loss": 0.2269,
      "reward": -0.11460491642355919,
      "reward_std": 1.2216467559337616,
      "rewards/correctness_reward_func": 0.2812500149011612,
      "rewards/int_reward_func": 0.2823660895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6782210320234299,
      "step": 207
    },
    {
      "completion_length": 564.3928680419922,
      "epoch": 0.7790262172284644,
      "grad_norm": 1.4329153299331665,
      "kl": 5.193359375,
      "learning_rate": 2.6142061987019574e-07,
      "loss": 0.2077,
      "reward": -0.15532590076327324,
      "reward_std": 1.2209243476390839,
      "rewards/correctness_reward_func": 0.2544642984867096,
      "rewards/int_reward_func": 0.2645089402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6742991209030151,
      "step": 208
    },
    {
      "completion_length": 535.3571624755859,
      "epoch": 0.7827715355805244,
      "grad_norm": 2.543517827987671,
      "kl": 5.3876953125,
      "learning_rate": 2.556893792515227e-07,
      "loss": 0.2155,
      "reward": -0.10556028177961707,
      "reward_std": 1.288367360830307,
      "rewards/correctness_reward_func": 0.3035714402794838,
      "rewards/int_reward_func": 0.297991082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7071228325366974,
      "step": 209
    },
    {
      "completion_length": 525.5714416503906,
      "epoch": 0.7865168539325843,
      "grad_norm": 2.5897624492645264,
      "kl": 5.2900390625,
      "learning_rate": 2.500000000000001e-07,
      "loss": 0.2116,
      "reward": -0.1855446556583047,
      "reward_std": 1.2734140753746033,
      "rewards/correctness_reward_func": 0.2276785783469677,
      "rewards/int_reward_func": 0.2890625074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7022857666015625,
      "step": 210
    },
    {
      "completion_length": 516.7142944335938,
      "epoch": 0.7902621722846442,
      "grad_norm": 4.0968732833862305,
      "kl": Infinity,
      "learning_rate": 2.44353456961474e-07,
      "loss": 0.2274,
      "reward": -0.24795536883175373,
      "reward_std": 1.2348434031009674,
      "rewards/correctness_reward_func": 0.2232142984867096,
      "rewards/int_reward_func": 0.2712053656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7423750311136246,
      "step": 211
    },
    {
      "completion_length": 529.4643096923828,
      "epoch": 0.7940074906367042,
      "grad_norm": 1.8898731470108032,
      "kl": 4.9951171875,
      "learning_rate": 2.387507176420256e-07,
      "loss": 0.1998,
      "reward": -0.2701317183673382,
      "reward_std": 1.333281546831131,
      "rewards/correctness_reward_func": 0.1741071566939354,
      "rewards/int_reward_func": 0.2533482201397419,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.6987031251192093,
      "step": 212
    },
    {
      "completion_length": 522.2857360839844,
      "epoch": 0.797752808988764,
      "grad_norm": 1.5559784173965454,
      "kl": 4.8896484375,
      "learning_rate": 2.3319274204219424e-07,
      "loss": 0.1956,
      "reward": -0.15001339837908745,
      "reward_std": 1.1915283203125,
      "rewards/correctness_reward_func": 0.2232143022119999,
      "rewards/int_reward_func": 0.2845982238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6578259021043777,
      "step": 213
    },
    {
      "completion_length": 523.5357513427734,
      "epoch": 0.8014981273408239,
      "grad_norm": 1.6422719955444336,
      "kl": 4.6875,
      "learning_rate": 2.2768048249248644e-07,
      "loss": 0.1875,
      "reward": -0.14155804365873337,
      "reward_std": 1.2657607197761536,
      "rewards/correctness_reward_func": 0.3080357387661934,
      "rewards/int_reward_func": 0.2578125111758709,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7074063122272491,
      "step": 214
    },
    {
      "completion_length": 541.3928833007812,
      "epoch": 0.8052434456928839,
      "grad_norm": 1.9782863855361938,
      "kl": 4.970703125,
      "learning_rate": 2.2221488349019902e-07,
      "loss": 0.1988,
      "reward": -0.03004240826703608,
      "reward_std": 1.232962042093277,
      "rewards/correctness_reward_func": 0.325892873108387,
      "rewards/int_reward_func": 0.2823660746216774,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6383013874292374,
      "step": 215
    },
    {
      "completion_length": 519.107177734375,
      "epoch": 0.8089887640449438,
      "grad_norm": 1.8962812423706055,
      "kl": 5.2587890625,
      "learning_rate": 2.167968815375837e-07,
      "loss": 0.2104,
      "reward": -0.05005356844048947,
      "reward_std": 1.2248910963535309,
      "rewards/correctness_reward_func": 0.325892873108387,
      "rewards/int_reward_func": 0.3013392984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6772857159376144,
      "step": 216
    },
    {
      "epoch": 0.8089887640449438,
      "eval_completion_length": 529.0346140008223,
      "eval_kl": Infinity,
      "eval_loss": Infinity,
      "eval_reward": -0.20390458432712444,
      "eval_reward_std": 1.2542201882914492,
      "eval_rewards/correctness_reward_func": 0.21278196471302133,
      "eval_rewards/int_reward_func": 0.2732456263742949,
      "eval_rewards/soft_format_reward_func": 4.6992483303735133e-05,
      "eval_rewards/strict_format_reward_func": 0.00018796993321494053,
      "eval_rewards/xmlcount_reward_func": -0.6901671490386913,
      "eval_runtime": 2401.7698,
      "eval_samples_per_second": 0.549,
      "eval_steps_per_second": 0.04,
      "step": 216
    },
    {
      "completion_length": 530.4643249511719,
      "epoch": 0.8127340823970037,
      "grad_norm": 1.53620183467865,
      "kl": 5.0556640625,
      "learning_rate": 2.1142740498138324e-07,
      "loss": 0.2023,
      "reward": -0.06143304280703887,
      "reward_std": 1.3513036668300629,
      "rewards/correctness_reward_func": 0.3348214402794838,
      "rewards/int_reward_func": 0.2890625149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6853170022368431,
      "step": 217
    },
    {
      "completion_length": 521.9286041259766,
      "epoch": 0.8164794007490637,
      "grad_norm": 1.607338786125183,
      "kl": 4.9990234375,
      "learning_rate": 2.0610737385376348e-07,
      "loss": 0.1999,
      "reward": -0.2871718890964985,
      "reward_std": 1.2107819616794586,
      "rewards/correctness_reward_func": 0.2008928656578064,
      "rewards/int_reward_func": 0.2265625111758709,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7146272361278534,
      "step": 218
    },
    {
      "completion_length": 523.6428833007812,
      "epoch": 0.8202247191011236,
      "grad_norm": 3.4551496505737305,
      "kl": 5.6904296875,
      "learning_rate": 2.0083769971467045e-07,
      "loss": 0.2276,
      "reward": -0.09260490955784917,
      "reward_std": 1.119924932718277,
      "rewards/correctness_reward_func": 0.2544642984867096,
      "rewards/int_reward_func": 0.2801339328289032,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6272031292319298,
      "step": 219
    },
    {
      "completion_length": 521.1428833007812,
      "epoch": 0.8239700374531835,
      "grad_norm": 1.8328137397766113,
      "kl": 4.896484375,
      "learning_rate": 1.9561928549563966e-07,
      "loss": 0.1958,
      "reward": -0.06695759034482762,
      "reward_std": 1.2224089801311493,
      "rewards/correctness_reward_func": 0.3035714402794838,
      "rewards/int_reward_func": 0.3069196566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6774486601352692,
      "step": 220
    },
    {
      "completion_length": 519.1071624755859,
      "epoch": 0.8277153558052435,
      "grad_norm": 2.1587209701538086,
      "kl": 5.431640625,
      "learning_rate": 1.9045302534508295e-07,
      "loss": 0.2173,
      "reward": -0.10098438337445259,
      "reward_std": 1.4179500043392181,
      "rewards/correctness_reward_func": 0.325892873108387,
      "rewards/int_reward_func": 0.2712053656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6980826258659363,
      "step": 221
    },
    {
      "completion_length": 515.1428833007812,
      "epoch": 0.8314606741573034,
      "grad_norm": 2.184795618057251,
      "kl": 5.4384765625,
      "learning_rate": 1.8533980447508135e-07,
      "loss": 0.2176,
      "reward": -0.19461161736398935,
      "reward_std": 1.3318405449390411,
      "rewards/correctness_reward_func": 0.258928582072258,
      "rewards/int_reward_func": 0.2633928693830967,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7169330567121506,
      "step": 222
    },
    {
      "completion_length": 520.8928985595703,
      "epoch": 0.8352059925093633,
      "grad_norm": 1.745571494102478,
      "kl": 5.236328125,
      "learning_rate": 1.8028049900970767e-07,
      "loss": 0.2094,
      "reward": -0.16562947491183877,
      "reward_std": 1.4269579648971558,
      "rewards/correctness_reward_func": 0.3125000111758709,
      "rewards/int_reward_func": 0.2723214402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7504509389400482,
      "step": 223
    },
    {
      "completion_length": 521.5,
      "epoch": 0.8389513108614233,
      "grad_norm": 1.687516689300537,
      "kl": 4.83203125,
      "learning_rate": 1.7527597583490823e-07,
      "loss": 0.1933,
      "reward": -0.18002455797977746,
      "reward_std": 1.2918389737606049,
      "rewards/correctness_reward_func": 0.2366071492433548,
      "rewards/int_reward_func": 0.2600446529686451,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6766763776540756,
      "step": 224
    },
    {
      "completion_length": 533.3571472167969,
      "epoch": 0.8426966292134831,
      "grad_norm": 1.9606373310089111,
      "kl": 4.9736328125,
      "learning_rate": 1.7032709244996556e-07,
      "loss": 0.1989,
      "reward": -0.16109822131693363,
      "reward_std": 1.2961285710334778,
      "rewards/correctness_reward_func": 0.2633928656578064,
      "rewards/int_reward_func": 0.2555803582072258,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6811875551939011,
      "step": 225
    },
    {
      "completion_length": 524.8928680419922,
      "epoch": 0.846441947565543,
      "grad_norm": 1.9700647592544556,
      "kl": 5.3662109375,
      "learning_rate": 1.6543469682057104e-07,
      "loss": 0.2147,
      "reward": -0.21540626091882586,
      "reward_std": 1.4037460684776306,
      "rewards/correctness_reward_func": 0.258928582072258,
      "rewards/int_reward_func": 0.2477678656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7221027463674545,
      "step": 226
    },
    {
      "completion_length": 527.5357208251953,
      "epoch": 0.850187265917603,
      "grad_norm": 1.6679848432540894,
      "kl": 4.9638671875,
      "learning_rate": 1.605996272335291e-07,
      "loss": 0.1985,
      "reward": -0.08088393486104906,
      "reward_std": 1.2692644596099854,
      "rewards/correctness_reward_func": 0.3839285895228386,
      "rewards/int_reward_func": 0.2834821566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7482946962118149,
      "step": 227
    },
    {
      "completion_length": 519.2857360839844,
      "epoch": 0.8539325842696629,
      "grad_norm": 1.6768486499786377,
      "kl": 5.38671875,
      "learning_rate": 1.5582271215312293e-07,
      "loss": 0.2154,
      "reward": -0.09632367175072432,
      "reward_std": 1.2886937856674194,
      "rewards/correctness_reward_func": 0.2901785895228386,
      "rewards/int_reward_func": 0.2756696492433548,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6621718853712082,
      "step": 228
    },
    {
      "completion_length": 515.8571624755859,
      "epoch": 0.8576779026217228,
      "grad_norm": 1.442134976387024,
      "kl": 4.7998046875,
      "learning_rate": 1.5110477007916e-07,
      "loss": 0.192,
      "reward": -0.2380736656486988,
      "reward_std": 1.2216798663139343,
      "rewards/correctness_reward_func": 0.1964285783469677,
      "rewards/int_reward_func": 0.2712053656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7057076245546341,
      "step": 229
    },
    {
      "completion_length": 538.9643096923828,
      "epoch": 0.8614232209737828,
      "grad_norm": 2.484066963195801,
      "kl": 4.9287109375,
      "learning_rate": 1.4644660940672627e-07,
      "loss": 0.1971,
      "reward": -0.11506248824298382,
      "reward_std": 1.112929493188858,
      "rewards/correctness_reward_func": 0.2767857238650322,
      "rewards/int_reward_func": 0.274553582072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6664018034934998,
      "step": 230
    },
    {
      "completion_length": 516.0357208251953,
      "epoch": 0.8651685393258427,
      "grad_norm": 2.529921770095825,
      "kl": 5.560546875,
      "learning_rate": 1.4184902828767286e-07,
      "loss": 0.2224,
      "reward": -0.07746875449083745,
      "reward_std": 1.3153048753738403,
      "rewards/correctness_reward_func": 0.299107164144516,
      "rewards/int_reward_func": 0.3013392984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6779152303934097,
      "step": 231
    },
    {
      "completion_length": 517.7500152587891,
      "epoch": 0.8689138576779026,
      "grad_norm": 2.0569961071014404,
      "kl": 5.0791015625,
      "learning_rate": 1.3731281449385628e-07,
      "loss": 0.2032,
      "reward": -0.2263192073442042,
      "reward_std": 1.2717112004756927,
      "rewards/correctness_reward_func": 0.2500000111758709,
      "rewards/int_reward_func": 0.2756696566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.751988872885704,
      "step": 232
    },
    {
      "completion_length": 516.8214569091797,
      "epoch": 0.8726591760299626,
      "grad_norm": 1.546913743019104,
      "kl": 4.626953125,
      "learning_rate": 1.3283874528215733e-07,
      "loss": 0.185,
      "reward": -0.0691049168817699,
      "reward_std": 1.3089016377925873,
      "rewards/correctness_reward_func": 0.3035714402794838,
      "rewards/int_reward_func": 0.286830373108387,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.660622775554657,
      "step": 233
    },
    {
      "completion_length": 515.107177734375,
      "epoch": 0.8764044943820225,
      "grad_norm": 3.8171000480651855,
      "kl": 5.3759765625,
      "learning_rate": 1.284275872613028e-07,
      "loss": 0.215,
      "reward": -0.04519642610102892,
      "reward_std": 1.1741289049386978,
      "rewards/correctness_reward_func": 0.3169643059372902,
      "rewards/int_reward_func": 0.3203125223517418,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6824732571840286,
      "step": 234
    },
    {
      "completion_length": 522.7500152587891,
      "epoch": 0.8801498127340824,
      "grad_norm": 1.9495872259140015,
      "kl": 4.9345703125,
      "learning_rate": 1.2408009626051135e-07,
      "loss": 0.1974,
      "reward": -0.058752238750457764,
      "reward_std": 1.273261696100235,
      "rewards/correctness_reward_func": 0.3214285895228386,
      "rewards/int_reward_func": 0.3046875074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6848683282732964,
      "step": 235
    },
    {
      "completion_length": 518.7500305175781,
      "epoch": 0.8838951310861424,
      "grad_norm": 1.8630801439285278,
      "kl": 5.265625,
      "learning_rate": 1.1979701719998454e-07,
      "loss": 0.2106,
      "reward": -0.20418081060051918,
      "reward_std": 1.4417231380939484,
      "rewards/correctness_reward_func": 0.3392857387661934,
      "rewards/int_reward_func": 0.3147321566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8581986874341965,
      "step": 236
    },
    {
      "completion_length": 532.1071624755859,
      "epoch": 0.8876404494382022,
      "grad_norm": 1.293250560760498,
      "kl": 4.8095703125,
      "learning_rate": 1.1557908396327026e-07,
      "loss": 0.1924,
      "reward": -0.08567634131759405,
      "reward_std": 1.2106836140155792,
      "rewards/correctness_reward_func": 0.2991071566939354,
      "rewards/int_reward_func": 0.2712053693830967,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.655988872051239,
      "step": 237
    },
    {
      "completion_length": 526.3928833007812,
      "epoch": 0.8913857677902621,
      "grad_norm": 2.8587074279785156,
      "kl": 5.3740234375,
      "learning_rate": 1.1142701927151454e-07,
      "loss": 0.215,
      "reward": -0.1162053607404232,
      "reward_std": 1.1973454058170319,
      "rewards/correctness_reward_func": 0.2589285857975483,
      "rewards/int_reward_func": 0.2968750149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6720089465379715,
      "step": 238
    },
    {
      "completion_length": 517.2500305175781,
      "epoch": 0.8951310861423221,
      "grad_norm": 1.921934962272644,
      "kl": 4.974609375,
      "learning_rate": 1.0734153455962763e-07,
      "loss": 0.199,
      "reward": -0.16254911944270134,
      "reward_std": 1.2848596572875977,
      "rewards/correctness_reward_func": 0.3125000223517418,
      "rewards/int_reward_func": 0.2946428656578064,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.769691988825798,
      "step": 239
    },
    {
      "completion_length": 517.1071624755859,
      "epoch": 0.898876404494382,
      "grad_norm": 2.530561685562134,
      "kl": 5.380859375,
      "learning_rate": 1.0332332985438247e-07,
      "loss": 0.2153,
      "reward": -0.12155580706894398,
      "reward_std": 1.2704256772994995,
      "rewards/correctness_reward_func": 0.3214285895228386,
      "rewards/int_reward_func": 0.3046875149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7476719170808792,
      "step": 240
    },
    {
      "completion_length": 519.7143249511719,
      "epoch": 0.9026217228464419,
      "grad_norm": 2.2845098972320557,
      "kl": 4.5966796875,
      "learning_rate": 9.937309365446971e-08,
      "loss": 0.1839,
      "reward": 0.07114956201985478,
      "reward_std": 1.152601718902588,
      "rewards/correctness_reward_func": 0.3660714477300644,
      "rewards/int_reward_func": 0.2968750074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5917969048023224,
      "step": 241
    },
    {
      "completion_length": 520.1786041259766,
      "epoch": 0.9063670411985019,
      "grad_norm": 2.1960318088531494,
      "kl": 5.0009765625,
      "learning_rate": 9.549150281252632e-08,
      "loss": 0.2,
      "reward": -0.03549107629805803,
      "reward_std": 1.2035468220710754,
      "rewards/correctness_reward_func": 0.2723214402794838,
      "rewards/int_reward_func": 0.2812500074505806,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5890625417232513,
      "step": 242
    },
    {
      "completion_length": 527.7143096923828,
      "epoch": 0.9101123595505618,
      "grad_norm": 2.1960318088531494,
      "kl": Infinity,
      "learning_rate": 9.549150281252632e-08,
      "loss": 0.2017,
      "reward": -0.17003126139752567,
      "reward_std": 1.2424619793891907,
      "rewards/correctness_reward_func": 0.21875000931322575,
      "rewards/int_reward_func": 0.2912946566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6800759434700012,
      "step": 243
    },
    {
      "epoch": 0.9101123595505618,
      "eval_completion_length": 537.0225817228618,
      "eval_kl": Infinity,
      "eval_loss": Infinity,
      "eval_reward": -0.19749192681751754,
      "eval_reward_std": 1.2450736566593772,
      "eval_rewards/correctness_reward_func": 0.21118422009442983,
      "eval_rewards/int_reward_func": 0.2718750116072203,
      "eval_rewards/soft_format_reward_func": 9.398496660747027e-05,
      "eval_rewards/strict_format_reward_func": 7.048872495560269e-05,
      "eval_rewards/xmlcount_reward_func": -0.6807156267919039,
      "eval_runtime": 2424.2347,
      "eval_samples_per_second": 0.544,
      "eval_steps_per_second": 0.039,
      "step": 243
    },
    {
      "completion_length": 515.9643096923828,
      "epoch": 0.9138576779026217,
      "grad_norm": 1.9732158184051514,
      "kl": 4.8046875,
      "learning_rate": 9.167922241916054e-08,
      "loss": 0.1922,
      "reward": -0.22770760208368301,
      "reward_std": 1.2481779158115387,
      "rewards/correctness_reward_func": 0.2366071566939354,
      "rewards/int_reward_func": 0.2712053768336773,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7355201095342636,
      "step": 244
    },
    {
      "completion_length": 531.1071624755859,
      "epoch": 0.9176029962546817,
      "grad_norm": 1.794622540473938,
      "kl": 5.0888671875,
      "learning_rate": 8.793690568899215e-08,
      "loss": 0.2035,
      "reward": -0.12151339650154114,
      "reward_std": 1.2451187074184418,
      "rewards/correctness_reward_func": 0.3303571678698063,
      "rewards/int_reward_func": 0.2845982275903225,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7364687621593475,
      "step": 245
    },
    {
      "completion_length": 529.2143096923828,
      "epoch": 0.9213483146067416,
      "grad_norm": 3.917875051498413,
      "kl": 5.42578125,
      "learning_rate": 8.426519384872732e-08,
      "loss": 0.217,
      "reward": -0.20378349348902702,
      "reward_std": 1.318730652332306,
      "rewards/correctness_reward_func": 0.258928582072258,
      "rewards/int_reward_func": 0.2700892984867096,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7328013777732849,
      "step": 246
    },
    {
      "completion_length": 527.3571624755859,
      "epoch": 0.9250936329588015,
      "grad_norm": 1.6186972856521606,
      "kl": 4.9921875,
      "learning_rate": 8.066471602728803e-08,
      "loss": 0.1997,
      "reward": -0.17118750512599945,
      "reward_std": 1.3723264038562775,
      "rewards/correctness_reward_func": 0.2366071566939354,
      "rewards/int_reward_func": 0.2845982313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6923928707838058,
      "step": 247
    },
    {
      "completion_length": 521.7143096923828,
      "epoch": 0.9288389513108615,
      "grad_norm": 1.6485563516616821,
      "kl": 4.7802734375,
      "learning_rate": 7.71360891480134e-08,
      "loss": 0.1912,
      "reward": -0.2930625192821026,
      "reward_std": 1.3668555617332458,
      "rewards/correctness_reward_func": 0.2232142984867096,
      "rewards/int_reward_func": 0.2723214402794838,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7897143363952637,
      "step": 248
    },
    {
      "completion_length": 528.6071624755859,
      "epoch": 0.9325842696629213,
      "grad_norm": 2.551079034805298,
      "kl": 5.6396484375,
      "learning_rate": 7.36799178229539e-08,
      "loss": 0.2256,
      "reward": -0.09009598195552826,
      "reward_std": 1.4010609686374664,
      "rewards/correctness_reward_func": 0.3348214402794838,
      "rewards/int_reward_func": 0.2991071566939354,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7240245938301086,
      "step": 249
    },
    {
      "completion_length": 530.5714569091797,
      "epoch": 0.9363295880149812,
      "grad_norm": 1.623828649520874,
      "kl": 5.0751953125,
      "learning_rate": 7.029679424927365e-08,
      "loss": 0.203,
      "reward": -0.2168437521904707,
      "reward_std": 1.3051793575286865,
      "rewards/correctness_reward_func": 0.2767857238650322,
      "rewards/int_reward_func": 0.313616082072258,
      "rewards/soft_format_reward_func": 0.0022321429569274187,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8094777017831802,
      "step": 250
    },
    {
      "completion_length": 529.1428985595703,
      "epoch": 0.9400749063670412,
      "grad_norm": 2.090834379196167,
      "kl": 5.0615234375,
      "learning_rate": 6.698729810778064e-08,
      "loss": 0.2024,
      "reward": -0.35498661547899246,
      "reward_std": 1.3341231644153595,
      "rewards/correctness_reward_func": 0.1830357238650322,
      "rewards/int_reward_func": 0.282366082072258,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.8203884065151215,
      "step": 251
    },
    {
      "completion_length": 541.7143096923828,
      "epoch": 0.9438202247191011,
      "grad_norm": 1.5824670791625977,
      "kl": 5.072265625,
      "learning_rate": 6.375199646360141e-08,
      "loss": 0.2029,
      "reward": -0.16221206076443195,
      "reward_std": 1.3483150601387024,
      "rewards/correctness_reward_func": 0.2678571566939354,
      "rewards/int_reward_func": 0.2834821566939354,
      "rewards/soft_format_reward_func": 0.0011160714784637094,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7146674692630768,
      "step": 252
    },
    {
      "completion_length": 521.357177734375,
      "epoch": 0.947565543071161,
      "grad_norm": 2.750608205795288,
      "kl": 5.5068359375,
      "learning_rate": 6.059144366901736e-08,
      "loss": 0.2202,
      "reward": -0.1930067086359486,
      "reward_std": 1.1880230009555817,
      "rewards/correctness_reward_func": 0.2455357238650322,
      "rewards/int_reward_func": 0.2611607313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6997031569480896,
      "step": 253
    },
    {
      "completion_length": 519.3214569091797,
      "epoch": 0.951310861423221,
      "grad_norm": 1.7069954872131348,
      "kl": 5.3154296875,
      "learning_rate": 5.750618126847912e-08,
      "loss": 0.2126,
      "reward": -0.04345982416998595,
      "reward_std": 1.2727106213569641,
      "rewards/correctness_reward_func": 0.3214285857975483,
      "rewards/int_reward_func": 0.2890625149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6539509445428848,
      "step": 254
    },
    {
      "completion_length": 517.6786041259766,
      "epoch": 0.9550561797752809,
      "grad_norm": 1.5457656383514404,
      "kl": 4.9775390625,
      "learning_rate": 5.44967379058161e-08,
      "loss": 0.1991,
      "reward": -0.09097097720950842,
      "reward_std": 1.3295612633228302,
      "rewards/correctness_reward_func": 0.3348214477300644,
      "rewards/int_reward_func": 0.302455373108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7282477915287018,
      "step": 255
    },
    {
      "completion_length": 529.6428833007812,
      "epoch": 0.9588014981273408,
      "grad_norm": 3.4232678413391113,
      "kl": Infinity,
      "learning_rate": 5.156362923365587e-08,
      "loss": 0.2322,
      "reward": -0.2118973322212696,
      "reward_std": 1.3026173412799835,
      "rewards/correctness_reward_func": 0.2455357275903225,
      "rewards/int_reward_func": 0.2901785857975483,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0011160714784637094,
      "rewards/xmlcount_reward_func": -0.7487277090549469,
      "step": 256
    },
    {
      "completion_length": 526.2500305175781,
      "epoch": 0.9625468164794008,
      "grad_norm": 1.854992389678955,
      "kl": 4.8408203125,
      "learning_rate": 4.870735782506979e-08,
      "loss": 0.1936,
      "reward": -0.1605692015727982,
      "reward_std": 1.3640064895153046,
      "rewards/correctness_reward_func": 0.2901785783469677,
      "rewards/int_reward_func": 0.2812500149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7319977879524231,
      "step": 257
    },
    {
      "completion_length": 516.5357208251953,
      "epoch": 0.9662921348314607,
      "grad_norm": 4.538949489593506,
      "kl": 5.3310546875,
      "learning_rate": 4.592841308745932e-08,
      "loss": 0.2132,
      "reward": -0.15284822694957256,
      "reward_std": 1.2363348603248596,
      "rewards/correctness_reward_func": 0.2500000111758709,
      "rewards/int_reward_func": 0.3058035895228386,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7086518108844757,
      "step": 258
    },
    {
      "completion_length": 514.5714569091797,
      "epoch": 0.9700374531835206,
      "grad_norm": 1.904927372932434,
      "kl": 4.98828125,
      "learning_rate": 4.322727117869951e-08,
      "loss": 0.1995,
      "reward": -0.04242856800556183,
      "reward_std": 1.3000076413154602,
      "rewards/correctness_reward_func": 0.3526785932481289,
      "rewards/int_reward_func": 0.2845982313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6797053962945938,
      "step": 259
    },
    {
      "completion_length": 528.6071472167969,
      "epoch": 0.9737827715355806,
      "grad_norm": 1.8089706897735596,
      "kl": 5.0556640625,
      "learning_rate": 4.06043949255509e-08,
      "loss": 0.2022,
      "reward": -0.1290022386237979,
      "reward_std": 1.231601744890213,
      "rewards/correctness_reward_func": 0.2901785895228386,
      "rewards/int_reward_func": 0.2700893022119999,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.689270094037056,
      "step": 260
    },
    {
      "completion_length": 517.4643096923828,
      "epoch": 0.9775280898876404,
      "grad_norm": 2.000513792037964,
      "kl": 5.0791015625,
      "learning_rate": 3.806023374435663e-08,
      "loss": 0.2031,
      "reward": -0.10831697471439838,
      "reward_std": 1.2132568657398224,
      "rewards/correctness_reward_func": 0.2500000074505806,
      "rewards/int_reward_func": 0.2924107238650322,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.650727704167366,
      "step": 261
    },
    {
      "completion_length": 527.6428833007812,
      "epoch": 0.9812734082397003,
      "grad_norm": 1.606877088546753,
      "kl": 5.140625,
      "learning_rate": 3.559522356403788e-08,
      "loss": 0.2056,
      "reward": -0.1139330342411995,
      "reward_std": 1.2683680653572083,
      "rewards/correctness_reward_func": 0.325892873108387,
      "rewards/int_reward_func": 0.3113839402794838,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7512098550796509,
      "step": 262
    },
    {
      "completion_length": 526.2143249511719,
      "epoch": 0.9850187265917603,
      "grad_norm": 1.3300163745880127,
      "kl": 4.8916015625,
      "learning_rate": 3.3209786751399184e-08,
      "loss": 0.1956,
      "reward": -0.23550447449088097,
      "reward_std": 1.2275131940841675,
      "rewards/correctness_reward_func": 0.16964286379516125,
      "rewards/int_reward_func": 0.2332589365541935,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6384062841534615,
      "step": 263
    },
    {
      "completion_length": 526.7143096923828,
      "epoch": 0.9887640449438202,
      "grad_norm": 5.495514392852783,
      "kl": 5.5947265625,
      "learning_rate": 3.0904332038757974e-08,
      "loss": 0.2238,
      "reward": -0.03097990620881319,
      "reward_std": 1.2771217823028564,
      "rewards/correctness_reward_func": 0.3035714477300644,
      "rewards/int_reward_func": 0.3125000149011612,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6470513641834259,
      "step": 264
    },
    {
      "completion_length": 521.1428680419922,
      "epoch": 0.9925093632958801,
      "grad_norm": 1.869722604751587,
      "kl": 5.494140625,
      "learning_rate": 2.8679254453910785e-08,
      "loss": 0.2197,
      "reward": -0.08227232204808388,
      "reward_std": 1.3968145847320557,
      "rewards/correctness_reward_func": 0.4017857387661934,
      "rewards/int_reward_func": 0.3158482313156128,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.7999062538146973,
      "step": 265
    },
    {
      "completion_length": 534.107177734375,
      "epoch": 0.9962546816479401,
      "grad_norm": 1.7442686557769775,
      "kl": 4.9443359375,
      "learning_rate": 2.653493525244721e-08,
      "loss": 0.1978,
      "reward": -0.18733705952763557,
      "reward_std": 1.244984209537506,
      "rewards/correctness_reward_func": 0.2410714440047741,
      "rewards/int_reward_func": 0.271205373108387,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.6996138542890549,
      "step": 266
    },
    {
      "completion_length": 517.25,
      "epoch": 1.0,
      "grad_norm": 1.4125308990478516,
      "kl": 4.6435546875,
      "learning_rate": 2.4471741852423233e-08,
      "loss": 0.1817,
      "reward": -0.36143749207258224,
      "reward_std": 1.4109329879283905,
      "rewards/correctness_reward_func": 0.0,
      "rewards/int_reward_func": 0.21875,
      "rewards/soft_format_reward_func": 0.0,
      "rewards/strict_format_reward_func": 0.0,
      "rewards/xmlcount_reward_func": -0.5801874995231628,
      "step": 267
    },
    {
      "epoch": 1.0,
      "step": 267,
      "total_flos": 0.0,
      "train_loss": 0.16846829135939004,
      "train_runtime": 42505.033,
      "train_samples_per_second": 0.176,
      "train_steps_per_second": 0.006
    }
  ],
  "logging_steps": 1,
  "max_steps": 267,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 54,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}